Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stecf.org/documents/isr/ISR_JWST_2003-01.pdf Äàòà èçìåíåíèÿ: Thu May 11 17:33:36 2006 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 21:12:51 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: âå÷íûé êàëåíäàðü

Throughput & image quality of a MEMS spectrograph

ST-ECF Instrument Science Report JWST 2003-01

Calibration concept for the JWST Near-infrared Spectrograph (NIRSpec)

Harald Kuntschner, Robert Fosbury, Wolfram Freudling & Stefano Cristiani

--1--


ST-ECF ISR JWST 2003-01 V1.2 June 18, 2003

Calibration concept for the JWST Near-infrared Sp ectrograph (NIRSp ec)
Harald Kuntschner, Rob ert A.E. Fosbury, Wolfram Freudling ST European Coordinating Facility, European Southern Observatory Karl-Schwarzschild-Strasse 2, D-85748 Garching bei Munchen ¨ Stefano Cristiani Istituto Nazionale di Astrofisica, Osservatorio Astronomico di Trieste, Via Tiepolo 11, 34131 Trieste, Italy ABSTRACT This do cument provides a time consumption for the Near version assumes that slits for electromechanical system and lamps. discussion of the basic calibration needs and calibration Infrared Sp ectrograph (NIRSp ec) on JWST. The current multi ob ject sp ectroscopy will b e formed via a microthat NIRSp ec will carry internal continuum and line

1.

Intro duction

The James Webb Space Telescop e (hereafter JWST) has b een conceived to carry out breakthrough observations in the infrared (0.6 - 28 µm) bringing ma jor progress in areas such as: · Cosmology and the Structure of the Universe · The Origin and Evolution of Galaxies · The History of the Milky Way and its Neighb ours · The Birth and Formation of Stars · The Origin and Evolution of Planetary Systems In particular these themes have b een expanded by the JWST Ad Ho c Science Working Group (ASWG) into a set of p otential scientific observations comprising the JWST Design Reference Mission (DRM; http://www.ngst.nasa.gov/science/drm.html), that is a representative science program elab orated in sufficient detail to aid in the development of functional requirements for the JWST mission. In its revised version (http://www.ngst.stsci.edu/studies/drmv2.3/) the


­2­ DRM is made up of 18 programs, requiring approximately half of the nominal JWST 5-year mission life to complete. In order to carry out the DRM, a complex range of instrumentation is required with imaging and sp ectroscopic capabilities over a large wavelength range (0.6 - 28 µm). The JWST instrument suite will consist of three science instruments: a Near Infrared Camera (NIRCam), a Near Infrared Sp ectrograph (NIRSp ec) and a Mid Infrared Instrument (MIRI). Unlike the Hubble Space Telescop e, the JWST will b e in a second Lagrange p oint orbit and will not b e serviceable. Therefore, these will b e the only instruments JWST will ever have. This do cument fo cuses on the basic outline of tasks needed to provide the calibration for the NIRSp ec; emphasis is given to the astrophysical requirements, the time consumption of the calibrations and the general feasibility of the strategy. The calibration requirements derived here are based on the current status of the instrument design and will need to b e refined when more details of the instrument and its op eration concept b ecome available.

2.

Baseline of the Near Infrared Sp ectrograph

The Near Infrared Sp ectrograph (NIRSp ec) will b e the sp ectrograph in the wavelength range of 0.6 to 5 µm. The study of galaxy formation, clustering, chemical abundances, star formation, and kinematics, as well as active galactic nuclei, gamma-ray-bursters, sup ernovae, young stellar clusters, and measurements of the initial mass function of stars (IMF) require a near-infrared sp ectrograph. NIRSp ec in its current design provides users of JWST with the ability to obtain simultaneous sp ectra of more than 100 ob jects in a 9 (3. 4 â 3. 4) square arc-minute field of view. The sp ectra cover the 0.6 to 5 µm wavelength range with a resolving p ower of 100, and 1 to 5 µm with 1000. The baseline sp ectrograph will probably take advantage of a micro-electromechanical system to provide dynamic ap erture shutter masks on a fixed grid. The micro shutter array (MSA) will feature approximately 800 â 400 shutters, with each 200 â 450 mas (TBC) slit width and slit length resp ectively (see also Figure 1). Observations will b e p erformed with 200 mas wide slits while the slit length is variable and will b e constructed by op ening several adjacent shutters. The 100 mo de will b e covered by one prism (0.6 to 5 µm). For 1000 the observations are separated in three wavelength ranges (1.0 - 1.8 µm, 1.7 - 3.0 µm, 2.9 - 5.0 µm) where long pass filters are used to avoid order overlapping. The ap erture fo cal plane (200 mas slit width, slit length 3000 mo de covering the (3 â 3 arcsec FOV) to b e used
1

(AFP) will also carry a small numb er of fixed, conventional slits 4 ). In combination with the fixed slits there will b e an additional 1.7 to 5 µm range in two steps 1 . Furthermore an optional IFU, with 3000, may b e incorp orated into the ap erture fo cal plane.
3000.

Note, that the wavelength range below 2 µm is thought to be covered by ground based instruments with


­3­

Fig. 1.-- Schematic view of MSA shutters (shaded regions) configured with a 1 â 3 slit pro jected onto the detector (fine grid). The disp ersion direction is along the x-axis. The effective shutter width and height is 200 mas and 447 mas resp ectively, while the pitch size is 247 mas and 494 mas, resp ectively. The detector is sampled at 100 mas p er pixel. The schematic layout of the sp ectrograph showing the sequence of the basic sp ectral comp onents is presented in Figure 2. Current designs forsee eight filter wheel p ositions: three order sorting filters for 1000 & 3000, one transparent p osition for 100, one closed p osition with a reflective diffuser for internal calibration lamps, and three broad band filters (JKL). The grating wheel will carry one 100 prism covering the full wavelength range, three 1000 gratings & two 3000 gratings. Additionally there is one mirror for direct imaging (for target acquisition, and calibration purp oses). Thus there are 14 sp ectral elements in total provided. Not all of the elements will b e used indep endently. Most imp ortantly there are ten standard grating/prism and order blo cking filter combinations to calibrate (see Table 1). If the IFU is implemented then there are two more mo des to b e considered in the calibration concept. Additionally at least four imaging mo des (for target acquisition, hereafter TA) need to b e calibrated. This assumes that the same filters are used for TA as for the sp ectroscopic mo des. The use of the broad band filters will add another three imaging mo des (TBC). The 100, 1000 mo des can b e used with the MSA and with the conventional slits. The 3000 mo de will b e only used with the conventional slits and the IFU. Non-standard combinations such as the 100 prism with a 1000 order sorting


­4­ filter are p ossible but not considered in this do cument. We note that there is a p otentially large numb er of non-standard disp ersing element and filter combinations to b e calibrated if they are b eing offered to the general user. Not all mo des listed in Table 1 need a full set of individual calibrations. For example, all internal lamp calibrations carried out for the MOS mo des will deliver a signal for the conventional slits as well. It is assumed that the IFU needs individual calibrations. In summary, for the internal lamp calibrations we consider eight different sp ectroscopic op erational mo des and four imaging mo des. For p ointed observations we assume that each mo de has to b e calibrated individually (12 sp ectroscopic mo des and four imaging mo des). In order to allow for an efficient calibration strategy we assume that the NIRSp ec can b e built in such a way that every combination of the filter and grating wheel p ositions share the same fo cus.
Tel. focus Fore optics with filter wheel and diffuser for internal lamps Collimator Camera AFP with MSA, slits Grating wheel Detector

Fig. 2.-- Schematic layout of sp ectrograph We assume that the MSA p osition is fixed with resp ect to the detectors. Therefore the detectors cannot b e illuminated directly but all light has to go through the MSA. To optimise sensitivity, detector pixels with a relatively large pro jected size on the sky will b e used (100 mas, see also Arribas et al. 2002). The fo cal plane assembly will b e covered by a mosaic of detectors (TBD). Detectors will b e either InSb or HgCdTe; b oth technologies are actively under study.

3.

The Scientific Needs

The DRM observations which involve the NIRSp ec instrument are summarised in Table 3 (adapted from ASCI I text form, version 2.4; August 6, 2001). The S/N requested for all observations is equal 10 or lower. More details (including texts of the prop osals) can b e found at http://www.stsci.edu/jwst/science/drm/programs.html. From a scientific p oint of view the following items need to b e addressed in the calibration of the NIRSp ec. 1. Knowledge of wavelength calibration 2. Knowledge of sp ectrophotometric calibration


­5­

Table 1: Main op erational science mo des Mo de wavelength Filter wheel Imaging 0.6 - 5 µm transparent Imaging > 1.0 µm long pass I Imaging > 1.7 µm long pass I I Imaging > 2.9 µm long pass I I I Imaging 1.2 µm J Imaging 2.2 µm K 3.8 µm L Imaging Sp ectroscopy 0.6 - 5 µm transparent Sp ectroscopy 0.6 - 5 µm transparent Sp ectroscopy 1.0 - 1.8 µm long pass I Sp ectroscopy 1.7 - 3 µm long pass I I Sp ectroscopy 2.9 - 5 µm long pass I I I Sp ectroscopy 1.0 - 1.8 µm long pass I Sp ectroscopy 1.7 - 3 µm long pass I I Sp ectroscopy 2.9 - 5 µm long pass I I I Sp ectroscopy 1.7 - 3 µm long pass I I Sp ectroscopy 2.9 - 5 µm long pass I I I Sp ectroscopy 1.7 - 3 µm long pass I I Sp ectroscopy 2.9 - 5 µm long pass I I I

AFP TA configuration TA configuration TA configuration TA configuration TA configuration TA configuration TA configuration 200 mas MOS 200 mas slit 200 mas MOS 200 mas MOS 200 mas MOS 200 mas slit 200 mas slit 200 mas slit 200 mas slit 200 mas slit IFU IFU

Grating wheel mirror mirror mirror mirror mirror mirror mirror R=100 R=100 R=1000 R=1000 R=1000 R=1000 R=1000 R=1000 R=3000 R=3000 R=3000 R=3000


­6­

Table 2: Summary of DRM NIRSp ec observations (version 2.4) Prop osal Title S/N mag diam µm (AB) (") P004 P015 P018 P026 P008 P013 P020 P026 P015 P015 P016 P017 P013 P018 P017 Low Resolution (100) IGM 1.26 10 29 Form. & Evol. Galaxies I I 3.5 10 29.4 GRB 2.1 7 27.5 SN 3.5 10 28 Low Resolution (300) Kuip er Belt 3 10 26.4 Sub-stellar Mass Ob jects 3 10 27.5 Circumstellar Disks 3 10 26.5 SN 1.6 10 27 Mo derate Resolution (1000) Form. & Evol. Galaxies I I 3.5 10 25.4 Form. & Evol. Galaxies I I 3.5 10 23.8 Form. & Evol. Galaxies I I I 3.5 10 25.4 Form. & Evol. Galaxies IV 3.5 10 25.45 High Resolution (3000) Sub-stellar Mass Ob jects 3 10 25.3 GRB 2.1 10 25 Form. & Evol. Galaxies IV 3.5 5 24.45 0.5 0.2 0.005 0.005 0.5 2 0.5 0.5 0.5

# p er sq.arcm. 112 100 112 13 110 12 110


­7­ 3. Knowledge of astrometric p osition along spatial extent of slit, or IFU 4. Knowledge of spatial and sp ectral PSF In order to quantify the scientific requirements for sp ectroscopy it is useful to classify the sp ectroscopic observations in three broad categories: 1. Redshifts 2. Line Diagnostics 3. Dynamics

3.1.

Redshifts

The determination of redshifts for galaxies, SNe, gamma-ray bursters at the faintest magnitudes is a ma jor part of the DRM. Typically it will b e carried out at a resolution of 100. Not all the astrophysical applications require an extremely accurate absolute wavelength calibration (for example when the emphasis is on line ratios). There are however cases in which an accurate absolute calibration is mandatory. Requirements on the wavelength accuracy, absolute and relative, are derived from studies of large scale structure (clustering) and of the correlations b etween luminous matter (for example galaxies) and gas in the intergalactic medium (IGM, see, for example Steidel et al. 2002; Adelb erger et al. 2002). A plausible upp er limit can b e set by the disp ersion of the systemic velo cities observed in Lyman-break galaxies, that show pronounced differences dep ending on the lines used to determine the redshift: nebular emission from gas around stars as opp osed to absorption of stellar continua by outflowing interstellar gas or multiply-scattered Lyman- emission. A typical value is v = 300 km/s. Another imp ortant case to b e envisaged is the comparison of the systemic velo cities of line-systems observed with NIRSp ec with resp ect to other instruments (e.g. ALMA). In such cases it is required that the wavelength determination is limited by the photon noise of the astrophysical source rather than by systematics in the zero p oint determination. Therefore, in the following we will assume as a goal for the accuracy of the wavelength calibration (combined effects of systematic and relative errors) in the 100 mo de 1/10 (rms) of a resolution element (FWHM; e.g., FWHM sampled by 2.3 pixels, i.e. 0.23 detector pixel accuracy or v 300 km/s). The wavelength calibration accuracy sets a lower limit on the redshift accuracy. The final accuracy which can b e achieved for a given ob ject critically dep ends on the numb er and typ e of sp ectroscopic features which can b e used for the redshift determination. Therefore, in practice the redshift error can b e significantly larger than that given by the wavelength calibration accuracy. Sp ectrophotometric calibration is not a critical issue for the ab ove mentioned observations. In general it can b e assumed that in many cases, when required, the SED will b e re-calibrated


­8­ post-facto on the basis of broad-band photometry. Exceptions are provided by the DRM prop osals "Probing the IGM out to the re-ionisation ep o ch" and "Measuring Cosmological Parameters with High-z Sup ernovae and the Evolution of the Cosmic Sup ernova Rate". In the former program an accurate measure of the damping wing of the IGM absorption marking the transition to a neutral IGM requires at least a relative accuracy of 10% in order to determine reliable optical depths. The relative sp ectrophotometric accuracy for SN observations needs to b e of the order of 5% while the absolute accuracy can b e relaxed to 15% (B. Leibundgut, private conversation). For 100 we are not aware of any stringent requirements on the spatial information in the sp ectra. The overall goal is to observe sources as faint as p ossible which naturally requires the collection and extraction of as much of the available flux as p ossible. A basic requirement for the spatial information is that the sp ectra of neighb ouring slits, separated by two closed shutters in spacial direction, do not overlap on the detectors. In order to minimize detector read-noise contributions the signal should b e concentrated on as few detector pixels as p ossible. We assume that diffraction limited spatial imaging quality at 3 µm (Strehl ratio > 0.8) and longer wavelength can b e achieved at the detector level. A reasonable knowledge of the spatial PSF shap e (at all wavelengths) will b e needed to facilitate an optimal data extraction. The spatial PSF shap e is defined as the relative intensity distribution of a p oint source observed with NIRSp ec along the spatial direction of the slit at the detector level. The FWHM of the spatial PSF will b e known and mappable with a low order p olynomial over at least 95% of the FOV to b etter than 5% (rms) at all wavelengths. A go o d knowledge of the spectral PSF shap e as a function of wavelength (i.e. line spread function) is needed to minimize systematic shifts intro duced by template mismatch in e.g., cross correlation techniques used for redshift determinations. The sp ectral PSF shap e is defined as the relative intensity distribution of a delta function observed with NIRSp ec along the disp ersion direction at detector level. We assume that the knowledge of the first moments of the sp ectral shap e (i.e. FWHM + asymmetry mo delled by e.g., a Gauss-Hermite series) to b etter than 5% (rms) will b e sufficient. In order to compare NIRSp ec observations to other data sources such as radio maps it is imp ortant to have a go o d knowledge of the astrometric p osition of the observed sp ectra (i.e. along the spatial extent of the slit). This should b e limited by the detector pixel sampling only. Thus we assume that an astrometric accuracy of 1/5 of a detector pixel (i.e. 20 mas, rms) or b etter can b e achieved. The astrometric p osition accuracy is defined with resp ect to the co ordinate frame given by the target acquisition reference ob jects.

3.2.

Line Diagnostics

The 1000 mo de will b e customarily used for astrophysical diagnostics such as determination of stellar ages, metallicities, temp eratures, densities, stellar surface gravity, and classification


­9­ of ob jects. A typical example is shown in Figure 3, illustrating a diagram to discriminate AGN from Liners and Starburst galaxies.

Fig. 3.-- Examples of line ratio diagrams used for the classification of ultra-luminous infrared galaxies. From Veilleux et al. (1999). Typical lines used in this typ e of study will b e: H, H , [OI I]3727, [OI I I]4363, 4959 + 5007, [NI I]6548, 6583. As can b e inferred from Figure 3, an accuracy b etter than 10%, absolute, over p ossibly large sp ectral ranges will b e required for a correct application of the diagnostic tests. This sets tight constraints on the absolute sp ectrophotometric calibration. The measurement of characteristic continuum shap es (e.g., 4000 ° break) can b e used to conA


­ 10 ­ strain the star-formation history of stellar p opulations in galaxies through sp ectral synthesis mo dels. Particularly interesting are diagnostics which can prob e the first Gyr after a star-burst. For example, in order to determine the difference b etween a 0.9 and 0.7 Gyr old stellar p opulation with the 4000 ° break, we require a relative sp ectrophotometric accuracy of the order of 5% (see Figure 4). A

Fig. 4.-- Stellar p opulation mo dels of an instantaneous starburst of age = 0.1, 0.3, 0.5, 0.7, 0.9 Gyr (from top to b ottom) and a constant metallicity of Z=0.008 are shown (mo dels from STARBURST99). The mo dels are normalized around 4000 ° A. Requirements on the stability of the wavelength scale can b e derived from the ability to clearly separate and analyse neighb ouring emission lines such as H and [NI I] ( 15 ° A combined A). absolute and relative wavelength scale accuracy of 1/5 (rms) of a resolution element (FWHM) app ears sufficient. For the 1000 mo de spatial information will b e of scientific interest since diagnostic properties are likely to change with p osition in extended ob jects. Since the ob jects are typically small


­ 11 ­ we require diffraction limited spatial imaging quality at 3 µm (Strehl ratio > 0.8) and longer wavelength. In order to investigate for example winds in star-formation regions the stability and rep eatability of the line-spread function is of scientific interest. In order to detect the exp ected asymmetric line profiles (e.g., a outflow from a few 10 km/s to 1000 km/s) we estimate that the knowledge of the sp ectral and spatial PSF shap e (defined in Section 3.1) to b etter than 5% (rms) is sufficient. However, a more detailed knowledge of the spatial PSF shap e will b e useful in the data-reduction of bright sources. It is to b e exp ected that some observations in the 1000 mo de are carried out at a higher S/N than the standard DRM observations. The astrometric requirements are similar to the ones given in Section 3.1.

3.3.

Dynamics

Studies of the dynamics of clusters of galaxies and individual galaxies at high redshift are another ma jor comp onent of the DRM, aiming at the determination of masses and the relation b etween visible and dark matter. A typical observation of cluster galaxies will require multi-ob ject sp ectroscopy (MOS) of ab out 100 ob jects p er cluster with the 1000 mo de. An absolute and relative precision for the wavelength calibration of 1/10 (rms) of a resolution element is sufficient for the determination of velo city disp ersions and substructure of clusters (v 30 km/s). The value of v 30 km/s is to b e interpreted as a minimum characteristic since in reality all gratings/prisms will show a wavelength dep endent sp ectral resolution. Internal dynamics of individual galaxies require a higher resolution mo de, 3000, to b e implemented in the form of an Integral Field Unit (IFU) or with a set of gratings in combination with the fixed slits. The observations are likely to b e carried out at a higher S/N (e.g., 30). Features like the Mg5177 line, the CaI I triplet at 8500 ° or the CO bands at 2.3 µm will b e A observed. The requested accuracy of the wavelength calibration (relative + absolute) is of the order of 1/10 of a resolution element (i.e. v 10km/s). Since this mo de will predominantly b e used to determine internal galaxy dynamics the absolute sp ectrophotometric accuracy can b e relaxed to 15%, while the relative sp ectrophotometric accuracy should b e on the same level as 1000 grating (i.e. b etter than 5% rms). For high S/N observations the line spread for the function needs to b e very well determined and repro ducable in order to measure accurate line-ofsight velo city distributions (LOSVD). The LOSVDs can b e used together with mo dels to investigate the dynamical state of the target. The knowledge of the sp ectral and spatial PSF shap e (defined in Section 3.1) to b etter than 5% (rms) is sufficient for the successful op eration of this sp ectroscopic mo de. For the 1000, 3000 mo de spatial information will b e of scientific interest since dynamics are


­ 12 ­ likely to change with p osition in extended ob jects. Since the ob jects are typically small we require diffraction limited (at 3 µm and longer wavelength, Strehl ratio > 0.8) spatial imaging quality. The astrometric requirements are similar to the ones given in Section 3.1.

3.4.

Summary of the scientific requirements

Almost all DRM observations with 100 or 1000 require a S/N of 10. The GRB observations 3000 mo de is likely to b e used with higher are an exception with a requested S/N = 7. The S/N ratios (S/N 25). The core calibration requirements derived from the ma jor science areas are listed b elow. Individual observations may need a less accurate calibration, while some observers may also desire a more accurate calibration for sp ecific programs. Where a requirement is given as a function of wavelength, the relevant sampling is assumed to b e the size of one resolution element (FWHM) of the disp ersive element. For imaging it is the band-width of the filters. For spatial requirements the sampling is assumed to b e the pixel size at the detector level. All requirements are applicable to fully reduced and calibrated data, i.e. the end-pro duct of a data reduction pip eline. We note, that in order to meet the following requirements, a typical, fully calibrated sp ectrum of a given target may need to b e built up of multiple exp osures. For example, in order to meet the sp ectrophotometric throughput one may need to dither individual exp osures. Requirements for 100, 1000, 3000

· The combination of absolute and relative wavelength calibration errors will b e smaller than 1/8 (rms, goal 1/10) of the resolution element (FWHM) of a given disp ersive element at all wavelengths and over at least 95% of the FOV. · Assuming no Poisson noise in the signal, multiple observations of the same target with different MSA or single slit configurations will provide a rep eatability for the overall throughput (with resp ect to a given standard source 2 ) of b etter than 5% (rms) for the full FOV. The throughput uncertainty as a function of wavelength will b e b elow 5% (rms). · The imaging quality at the fo cal plane assembly will b e diffraction limited for 3 µm and longer wavelengths (Strehl ratio > 0.8). The width (FWHM) of the spatial PSF at detector level will b e known and mappable with a low order p olynomial to b etter than 5% (rms) over at least 95% of the FOV and at all wavelengths. The first moments of the spectral PSF shap e (i.e. FWHM + deviations from a Gaussian mo delled by e.g., a Gauss-Hermite series) will known and mappable with a low order p olynomial over at least 95% of the FOV to b etter than 5% (rms) at all wavelengths.
We assume that suitable standard stars calibrated in absolute terms will be available for the wavelength range of NIRSpec.
2


­ 13 ­ · The astrometric p osition along the spatial extent of the slit will b e known and mappable over at least 95% of the FOV to b etter than 20 mas (rms) with resp ect to the co ordinate frame defined by the target acquisition reference ob jects for all wavelengths. We require that the fixed slits and the IFU are placed in a region of the FOV where the knowledge of the parameters stated ab ove is not degraded by vignetting or similar effects.

4.

Calibration Pro cedures

We follow largely the general overview of calibration pro cedures given by Casertano (2001) and Henry & Casertano (2002), having in mind the sp ecific needs of the NIR-sp ectrograph. We also make use of the "JWST-NIRSp ec Op erations Concept Do cument" (Regan et al. 2003) which describ es the overall op erations concept. A numb er of uncertainties affect the design of the JWST NIRSp ec calibration strategy. Ma jor ones are the level of stability and predictability of the optical train of the telescop e and the detectors, the detailed consequences of the diffraction prop erties of the MSA, and the availability of internal calibration sources in spite of the tight limits on p ower dissipation. Furthermore, the effects and the frequency of the re-phasing of the primary mirrors is currently not well established. We assume that monthly re-phasing will b e carried out. It is likely that the re-phasing will change not only the fo cus of the instrument but also the PSF and the overall distortion mo del. Therefore the calibration unit and the calibration concept have to provide for efficient and accurate re-calibrations of fo cus, PSF and distortion mo del. Continuum micro-lamps which reach temp eratures of 1000 K have b een rep orted by the GSFC group (Moseley, private communication). A useful calibration lamp for JWST should probably achieve somewhat higher temp eratures (1300 K, see Figure 5). Prop er emission line sources have not yet b een investigated. If no cold line sources are available at the wavelength of interest for NIRSp ec, an alternative p ossibility is filtering the continuum source using a Fabry-Perot etalon or similar for the internal wavelength calibration. For the determination of the zero p oint this pro cedure should provide a few emission lines in the wavelength range of each of the gratings. In order to determine the disp ersion function more lines are necessary. Typically > 10 unambiguously identifiable lines (line/continuum contrast > 80%) covering the wavelength range of each disp ersive element uniformly are sufficient to meet the requirements stated in Section 3.4. This assumes that the true disp ersion function of the sp ectrograph can b e describ ed by a simple function derived from the sp ectrograph optical mo del (detailed ground based measurements with >100 lines p er sp ectral band are needed). The line-spread function of the lines pro duced by the lamp (e.g., via Fabry-Perot etalon) should b e such that one can easily determine the central wavelength of the lines. This sets tight constraints on the stability of the lamp itself and the line-shap es of individual lines. We estimate that the lamps need to b e stable to b etter than < 0.025 FWHM (rms) of sp ectral element for at least a month (goal one year) in order to allow for efficient op erations.


­ 14 ­

° Fig. 5.-- Black b o dy curves for T=1000 K and T=1300 K. The flux is given in ergs/s/cm 2 /A. Sp ecifically note the low flux at wavelengths < 1.3 µm. In the following analysis we assume that go o d internal line lamps are available. The line calibration lamp exp osure time is assumed to b e 60 sec. For this exp osure time we exp ect a S/N of > 30 (p er detector pixel) at the p eaks of the lines. In order to achieve a pixel-to-pixel flat-field accuracy of b etter than 1-2% (rms) the continuum lamps need to pro duce a S/N > 100. We assume that this signal can b e achieved within 1000 sec. We note, that accurate flat-fields particularly at wavelength b elow 1.3 µm are probably very challenging (see also Figure 5). The details dep end on the actual temp erature achieved by the lamp which is TBC. We note that the internal lamps are not required to cover the wavelength range 0.6 - 1 µm, but it should b e a goal. Furthermore, we assume that the calibration lamps are adequately redundant systems. For on sky calibrations we assume generic exp osure times of 100 sec for bright sources and 500 or 1000 sec for faint sources. We give only the on-source exp osure times and do not include overheads for read-out time, telescop e slews etc. A further critical p oint of uncertainty are the diffraction prop erties of the MSA and their implications for the sp ectroscopic observations. The intro duction of the MSA is driven by the requirement to observe 100 ob jects simultaneously. However, its prop erties are quite different from conventional multi-slit observations for two main reasons: (a) The individual slits will b e constructed by op ening adjacent shutters which leads to diffraction effects by the MSA supp ort structure in the spatial direction; (b) Due to the random distribution of targets on the sky, only one science target p er mask can b e placed at an optimal p osition within its slit. All other ob jects within the same MSA mask will b e slightly misplaced by up to half the shutter size in sp ectral


­ 15 ­ direction. The final calibration strategy for many NIRSp ec comp onents may dep end on these diffraction prop erties. Therefore detailed instrument mo delling is needed. First results have b een obtained (see Freudling 2002) on which this do cument is based. We further note that the MSA prop erties, particularly the influences on the sp ectrophotometric accuracy, may determine the overall observing strategy (e.g., dithering, see Section 4.1.3) which in turn has an impact on the calibration needs. Overall the on b oard calibration concept should provide for some redundancy. For example it should b e still p ossible to acquire a target with the fixed slits if the MSA or one detector fails.

4.1.

Characteristics of the Calibrations

The goal of this section is to establish the need for and approximate time consumption of calibrations to ensure the successful op eration of NIRSp ec. The typ e and frequency of calibrations is determined by the scientific requirements, the instrument design, p erformance & stability and the observing strategy. All of these items are not finalized yet, hence this do cument can only describ e the calibration needs with resp ect to our current knowledge. It is exp ected that some of the requirements will change when the final instrument design is known. It is useful to separate calibrations by themes such as detector, optics, sensitivity, geometric and sp ectroscopic calibrations. Furthermore the calibrations can b e classified into three broad time categories: pre-flight, b eginning of mission and regular calibrations during science op erations. In this do cument we concentrate on the regular calibrations during mission. Pre-flight and b eginning of mission calibrations are referred to where deemed necessary, but are not describ ed in detail. Throughout this do cument it is assumed that calibrations taken in orbit can built on extensive ground-based calibrations and an accurate instrument mo del. Therefore the character of the inorbit calibrations is one of "verification" rather than "determination". However, the instrument and in particular the calibration unit design should provide the means to determine the individual calibrations without the input of the ground based mo del wherever this is p ossible. The tables in the following sub-sections summarize the routine calibrations needed during a typical year of observations. We follow the nomenclature of Henry & Casertano (2002). Sp ecifically the "Typ e" of calibration is defined as: · Pointed calibrations. These require dedicated p ointings. · Sky calibrations. Here the ob jective is to observe the sky background; the p ointing generally do esn't matter, although crowded fields or bright ob jects need to b e avoided. · Lamp calibrations. These observations make use of internal (continuum and line) lamps.


­ 16 ­ · Dark calibrations. Observations with the shutter closed. · Auto calibrations. Here the calibration information is extracted via iterative metho ds directly for the science data. Although no spacecraft time will b e necessary some metho ds will require substantial analysis in order to provide the required information. · Opp ortunistic calibrations. Here calibration information is extracted from a set of suitable science exp osures (separate from the observation b eing calibrated). The third table column gives the frequency of the calibration p er year. This dep ends on the stability of the quantity b eing measured. For auto calibrations and opp ortunistic calibrations this is given as 0. The fourth column shows the numb er of separate calibrations needed to p erform the measurement. An individual calibration is defined as a set of exp osures that generates a particular piece of information or reference file. The numb er of calibrations dep ends on how the measurement must b e iterated over parameters such as readout pattern, filters etc. to pro duce a complete set of data. The numb er of separate exp osures is given in the 5th column. Often this numb er is one but some calibrations need dithering or summing to pro duce an adequate signal. The last column of each table gives the effective cost in % of available NIRSp ec time, assuming that NIRSp ec will b e used 1/3 of the time during one year.

4.1.1.

Detector characteristics calibrations

A thorough ground based characterisation will b e needed for the whole set of detector calibrations summarised in Table 3. In orbit, only verification and monitoring of p otential changes should b e required. The frequency of these checks dep ends strongly on the stability of the detectors and their aging prop erties (e.g., increase in numb er of hot pixels). The detector typ e and its prop erties are TBC, so only general calibration pro cedures and time estimates can b e describ ed at the moment. Detector characteristics have also significant implications on the flat-field determination which will b e discussed in Section 4.1.3. Current designs of NIRSp ec foresee one detector readout mo de (MULTIACCUM) with user selectable exp osure times ranging from 50 to 20000 sec in steps of 50 sec and a single gain setting ( 1.5 e- /ADU). The 50 sec intervals are determined by the maximum available bandwidth to down-link the data. The exp osure times are given for full frame mo des. There is likely to b e a sub-array mo de for bright ob ject observations with much shorter exp osure times (minimum 40 ms). These bright-source mo des need to b e considered in the calibration strategy but are not exp ected to significantly contribute to the overall calibration time consumption. Furthermore, sp ecific readoutpatterns may b e used for frequent calibrations such as the target acquisition image, the mirror


­ 17 ­

Table 3: Detector characteristics calibrations
Measurement Bias drift Dark current RN determination RN verification Gain Linearity Bad pixels Hot pixels Image persistence Sum
a

Type auto dark dark dark c. lamp pointed opp. dark pointed

Frequency per year 0 2 2 52 2 1 0 120 6

# Calibrations 1 8 1 1 3 10 1 1 10

# Exposure (exp. time) - (-) 20 (1000) 10 (1000) 1 (500) 5 (60) 1 (500) - (-) 1 (1000) 1 (1000)

Total time ksec/year 0.0 320.0 20.0 26.0 1.8 5.0 0.0 120.0 60.0

Eff. cost % time 0.00 3.04 0.00a 0.00a 0.02 0.05 0.00 0.00a 0.57 3.68

It is assumed that the dark observations can be shared with the main dark current determination program.

p osition calibration, the through-slit image and the wavelength zero-p oint (see Sections 4.1.4 & 4.1.5 for details). Casertano (2002) recommends auto calibration for the bias drift. The NIRSp ec detectors are likely to carry reference pixels which can b e used to track effects such as bias drifts. The dark current can b e calibrated with long exp osures taken with the external light path blo cked (diffuser selected in filter wheel). In order to achieve accurate dark subtraction (dark noise 0.7 e- /frame) a total exp osure time of 20 ksec is estimated p er exp osure time setting (assuming a detector noise of 3 e- rms p er frame). Since most of the JWST science targets are very faint sources it is essential to minimise the noise contributions from bias and dark subtractions. Therefore we see the estimated dark exp osure time (20 ksec) listed in Table 3 as minimum requirement. Casertano (2002) recommends calibrations twice a year. However, dark calibrations will probably b e run on a daily basis to track short term changes. It is envisaged that a sup er dark for the most imp ortant exp osure time settings (we assume 8 different ones) is determined at the b eginning of mission and then adjusted accordingly with the daily dark calibrations. Dark calibrations for intermediate exp osure times can b e extracted from the long dark exp osures since all science exp osures will have the same sampling, i.e. every 50 sec one full frame will b e send down to the ground station. Since the dark calibrations will take up a significant amount of time (> 3%), this calibration would greatly b enefit from a parallel capability of NIRSp ec, i.e. obtaining darks while another instrument on JWST is observing astrophysical targets. The read noise (RN) can b e obtained from a series of short darks and should b e determined twice a year, while RN verifications should b e run on a weekly basis. We exp ect that these observations can b e shared with the dark calibrations prop er. For the gain calibration Casertano (2002) suggests examination of the noise statistics of high illumination images. This calibration can b e p erformed most efficiently with a bright internal


­ 18 ­ continuum lamp and the mirror selected in the grating wheel. Casertano (2002) p oints out that the effective gain may b e wavelength dep endent if each photon can pro duce multiple electrons. This will impact the noise mo del. Careful ground testing of this effect is mandatory since in orbit an illumination of the detector in different wavelength bandpasses with internal lamps is difficult. Dep ending on the design, the internal calibration lamps may carry filters which can b e used to calibrate the wavelength dep endence in orbit. Extended astronomical sources (nearby galaxies, nebulae) are also viable gain calibration targets but their use makes the pro cedure longer and more complex. Currently we assume that the gain calibration will b e p erformed with internal continuum lamps in three filters twice a year and that 5 exp osures of 60 sec are sufficient to determine the gain. The linearity of the detectors will b e determined on the ground and the resulting mo del can b e checked by p ointed observations of bright astronomical targets on a yearly basis. Nearby elliptical galaxies may b e go o d targets to prob e a large dynamic range with one ob ject. We assume that 10 exp osures are sufficient to validate the ground based mo del. If the internal calibration lamp output do es show a high degree of stability then the calibration may b e p erformed with the continuum lamp. The p osition of bad pixels can b e frequently monitored by making use of suitable science exp osures. The monitoring of hot pixels probably requires frequent (e.g., daily) darks which accounts for a significant amount of exp osure time, however, it is envisaged that observations can b e shared with the dark calibration prop er. A thorough ground characterisation of latent images (p ersistence) is essential, since precise verification from space can b e time-consuming. The in orbit calibration strategy requires observations of bright targets followed by a series of darks. Latent images from cosmic rays can b e verified from darks and do not require dedicated observations. However, the b ehaviour during solar events pro ducing a rise in cosmic ray frequency needs to b e carefully calibrated at the b eginning of mission, if p ossible. Ma jor solar events (6 p er year) will require image p ersistence calibrations. Another imp ortant issue in connection with image p ersistence are observations of bright objects or highly illuminated calibration frames (e.g., bright emission lines). Current (conservative) estimates are of the order of 0.3% image p ersistence in the next exp osure. For example, an emission line with a S/N of 100 will pro duce a latent image of 30 electrons on the next exp osure. The data-reduction pip eline may b e able to remove the remaining signal if one can predict the exact time dep endence of the remaining latent image. This demands careful ground based testing and calibration of the detector system. Additionally, in orbit verification is necessary to monitor changes. We estimate that a half yearly verification is adequate.


­ 19 ­ 4.1.2. Optics calibrations

The optics calibrations are summarised in Table 4 and involve most imp ortantly the fo cus of the instrument and the determination of the PSF. In order to allow for an efficient calibration strategy we require that NIRSp ec can b e built in such a way that every combination of the filter & grating wheel p ositions with the MSA, the conventional slits and the IFU share the same fo cus. Due to the frequent (monthly, TBC) re-phasing of the primary mirror segments we exp ect that the basic fo cus and PSF determinations have to b e carried out on a monthly basis. Table 4: Optics calibrations
Measurement Focus F. field dependence PSF PSF verification PSF field dependence Image anomalies Sum Type p p p p p p ointed ointed ointed ointed ointed ointed Frequency per year 12 2 2 12 2 1 # Calibrations 1 1 16 1 16 16 # Exposure (exp. time) 7 (100) 25 (100) 16 (100) 16 (100) 16 (100) 5 (500) Total time ksec/year 8.4 5.0 51.2 19.2 51.2 40.0 Eff. cost % time 0.08 0.05 0.49 0.18 0.49 0.38 1.66

The determination of the fo cus in orbit requires a go o d, a priori, knowledge of the PSF shap e as function of fo cus p osition. Therefore extensive ground based calibrations and instrument mo delling is required. At the b eginning of mission a p ossible dep endence of the fo cus on orbital conditions needs to b e established. Having NIRSp ec in fo cus requires two steps: a) fo cusing of the instrument from the MSA to the detector; b) fo cusing of the telescop e on the MSA. Whether b oth fo ci are actively controlled in orbit is TBC. The first step can b e accomplished with internal continuum and line sources. The second fo cus step requires the observation of astronomical p oint-sources placing a mirror in the grating wheel. The b est fo cus needs to optimise the PSF in the spatial as well as in the sp ectral direction (i.e. line-spread function, see Section 4.1.5). We note that due to the diffraction effects on the MSA supp ort structure the PSF shap e variations may not b e trivial. Therefore, an instrument mo de which would allow access to a clear ap erture would b e very useful to establish the b est fo cus. The ap erture size should b e such that the full PSF of the telescop e is sampled (e.g., 3 arcsec diameter). Such a clear ap erture could b e incorp orated in the ap erture fo cal plane next to the conventional slits. There should b e at least two clear ap ertures (one on each detector) in order to allow for some redundancy. The clear ap ertures can also b e used for target acquisition if the MSA should fail. We assume that 7 images of a bright source (100 sec exp osure time) will b e sufficient to determine the b est instrument fo cus after each re-phasing of the primary mirrors. The accuracy of the fo cus adjustment should b e such that the requirements on the knowledge of the PSF shap e listed in Section 3.4 can b e easily achieved.


­ 20 ­ Fo cus field dep endence can b e investigated by observing, for example, a star cluster with all shutters op en and the mirror selected in the grating wheel. At the b eginning of mission extensive, dedicated observations of astronomical targets are necessary. We assume that the routine fo cus field dep endence calibration need to b e carried out only twice a year and that five p ointings with five dither p ositions of each 100 sec exp osure time are sufficient. If the re-phasing pro cedure changes the fo cus field dep endence, this calibration step has to b e carried out after each re-phasing. Much of the monitoring and control of the PSF pro duced by the telescop e itself will b e carried out by the Wavefront Sensing and Control (WSC) subsystem. The stability requirement is 2% (rms, 24h). Although the exact description of the PSF shap e pro duced by the combination of telescop e and NIRSp ec may b e complex, we exp ect a high stability. We assume currently that the PSF has to b e re-calibrated after each re-phasing, while the PSF field dep endence will need to b e re-calibrated only twice a year. The PSF calibration can b e p erformed with observations of astronomical p oint sources. For a full calibration, the spatial PSF has to b e determined as a function of wavelength for at least 12 sp ectroscopic instrument configurations and four imaging mo des (see Table 1). It is assumed that a full calibration is only needed twice a year while the PSF after re-phasing will b e determined for only one disp ersive element (PSF verification). The observations should deliver at least a S/N of 200 (p eak) for the combined images. The calibration of the PSF in the wavelength direction (line-spread function) is describ ed in Section 4.1.5. Each individual calibration may require dithers in order to account for sub-pixel accuracy and diffraction effects of the MSA. The numb er of dithers could b e large (e.g., 16; see Section 4.1.3 for details). However, we exp ect that bright targets can b e used with exp osure times of 100 sec. The requirements on the knowledge of the PSF shap e are listed in Section 3.4. FOV variations of the PSF can b e determined by observations of e.g., a star cluster while it is imp ortant to avoid crowding and therefore overlapping of PSFs. Stray light, ghost images, images of stars outside FOV can b e largely predicted on the basis of the instrument mo del but are typically identified on the basis of science data. The exp ected calibration time required for these image anomalies is limited to short (e.g., 500 sec) observations of appropriate astronomical fields on a yearly basis. We assume that 5 exp osures p er calibration will b e sufficient to determine the signal.

4.1.3.

Sensitivity Calibrations

Sensitivity calibrations are a complex issue and need to b e p erformed in principle for a prohibitively large set of instrument configurations (e.g., 800 â 400 shutters in MSA, and 12 sp ectral mo des). Key to an efficient calibration strategy is a well b ehaved sp ectrograph where sensitivity parameters change only smo othly across the FOV and thus a rather coarse sampling is sufficient to calibrate the full system. We require that the overall system sensitivity (excluding MSA diffraction effects) do es not change more than 10% (TBC) over the full FOV. The changes should b e known


­ 21 ­ and mappable to an accuracy of b etter than 2% (rms) with a low order p olynomial. A precise determination of the system parameters on the mission are required. We assume that the photometric resp onse the MSA) is stable so a yearly re-calibration will b e sufficient to b etter than 1% p er year). The necessary sensitivity calibrations Table 5: Sensitivity calibrations
Measurement Shutter throughput Slit throughput IFU throughput Phot. resp. MSA Phot. resp. slit Phot. resp. IFU Phot. resp. filters Background Small scale FFb Large scale FF Sum
a

ground and at the b eginning of of the optical elements (excluding characterise the system (stability are summarised in Table 5.

Type c. lamp c. lamp c. lamp pointed pointed pointed pointed opp. c. lamp sky

Frequency per year 12 2 2 1 2 2 1 0 12 2

# Calibrations 1 1 1 4 6 2 6 0 5 9

# Exposure (exp. time) 10 (60) 10 (60) 10 (60) 256 (100) 2 (100) 1 (100) 16 (100) 0 10 (60) 10 (1000)

Total time ksec/year 7.2 1.2 1.2 102.4 2.4 0.4 9.6 0.0 36.0 180.0

Eff. cost % time 0.07 0.00a 0.01 0.97 0.02 0.00 0.09 0.00 0.34 1.71 3.23

It is assumed that the throughput calibrations of the MSA shutters and the conventional slits can be shared. b A complete determination of the small scale MSA related flat field (FF) in orbit seems not feasible at the moment. The FF will heavily rely on a ground based model and occasional verifications in orbit.

The sp ectroscopic throughput is a pro duct of a) throughput for each ap erture (slit made of several shutters), b) throughput of the disp ersive element, and c) the sensitivity of the detector. We assume that an overall sensitivity mo del can b e constructed by determining the individual comp onents: a) Ap erture throughput: The variations in principle optics and shutter throughput (i.e. exact size on the sky) should b e measured on the ground to a high precision (knowledge <1% rms over full FOV) and can b e verified in orbit with exp osures of an internal continuum lamp with the mirror selected in the grating wheel. We assume that the illumination of the internal lamp is uniform (deviations <1% rms) over scales of 5 . This allows for illumination calibrations of the fixed slits and the IFU without the need for external, p ointed observations of background sources. For the conventional slits and the IFU a half-yearly verification seems adequate. We assume that ten exp osures of 60 sec are sufficient to deliver the information. For the MSA, ten exp osures are sufficient to map the entire array if the MSA mask is configured with a sp ecific pattern; for example op ening one micro-shutter every ten in the x direction and shifting the x p osition of the op ened micro-shutters by three units for each increment in y direction. With this pro cedure one can


­ 22 ­ also monitor MSA failures and partial op enings which can severely affect the throughput. We recommend monthly checks. However, dep ending on the exact p osition (sub-shutter accuracy) of a p oint source within the slit the throughput can vary significantly. For example, at 1.4 µm we estimate slit efficiencies b etween 55 and 65% within the nominal acceptance zone of one shutter (see W. Freudling 2002, for details). Therefore it is quite challenging to meet the scientific requirements in sp ectrophotometric throughput accuracy (see Section 3.4). Two themes may b e envisaged: (i) The actual throughput correction for each sp ectrum can b e mo delled theoretically if the exact ob ject p osition, characteristics of shutters and the incoming effective PSF is known for each ob ject [active calibration]. (ii) The throughput variations are smo othed out by an appropriate dither pattern which ensures a semi-uniform illumination of the slit in the disp ersion direction [passive calibration]. The first option implies an accurate knowledge of the ob jects parameters which may b e difficult or even imp ossible to obtain. For example, the emission line emitting regions for a given target may not b e at the same lo cation as the target p osition seen in the broad-band acquisition image. The second (passive) strategy will partly smo oth out these effects. Preliminary simulations show that reasonable dither patterns can deliver a mean absolute and relative throughput accuracy of 2% (e.g., 16 dither p ositions with 4 MSA re-configurations). Dithering is also necessary to achieve a full wavelength coverage, b ecause there will b e substantial gaps b etween the detectors in wavelength direction (see Section 4.1.4). An optimal dither pattern, which p ossibly dep ends on the wavelength band is TBD. We note that almost all p ointed calibration observations may have to b e dithered in order to eliminate the MSA diffraction effects. This will imp ose a significant overhead in configuration and read-out time particularly if exp osure times are short. b) Disp ersive element throughput: At least 12 disp ersive elements and order blo cking filter combinations need to b e calibrated as a function of wavelength and p osition within the FOV. For the MSA mo des the photometric resp onse can b e determined by observations of a sp ectrophotometric standard star placed at e.g., a 4x4 grid on the array (16 dithers of 16 p ositions gives 256 observations in total). By interp olating b etween grid p ositions one can determine a mo del for the entire area. The b ehaviour of the disp ersive element and the required accuracy determine the numb er of grid p oints needed to achieve a successful calibration. We estimate that a 2% (rms) accuracy is needed. For the IFU and the conventional slits we assume that the photometric resp onse can b e determined from one-p oint spatial sampling (spatial information can b e derived from continuum lamp exp osures). We assume a bright source can b e used (exp osure time 100 sec). The MSA mo de calibrations will b e p erformed on a yearly time scale, while the conventional slits and the IFU will b e rep eated every half year. We exp ect to achieve a higher quality for the fixed slits and the IFU calibrations. We note that the currently available sp ectrophotometric standard stars in the wavelength range


­ 23 ­ 1-5 µm are probably not suitable for NIRSp ec calibrations since they are very bright. Ground based observations need to provide "secondary standards" by the time JWST will take up its op erations. It would b e desirable to calibrate individual filter throughput changes as well. For this the mirror needs to b e selected in the grating wheel. We assume that the observation of one standard star field with 16 dithers in each of the filters is sufficient to calibrate the system. We assume high stability of the throughput so only yearly dedicated observations are necessary. Variations in the sky background as a function of wavelength and orbital conditions will b e determined at the b eginning of mission. Further determinations can probably b e extracted from science data. c) Detector sensitivity: The determination of the flat fields is the last and p ossibly most challenging step in this sequence. There are two fundamental strategies to obtain the flat-fields: (i) Individual calibration of each instrument set-up (grating+MSA configuration) with an internal calibration lamp. This obviously carries a heavy load on the calibration time needed (1000s) for each MSA reconfiguration but will deliver accurate flat-fields; (ii) Alternatively, one can aim to built up a global mo del of the detector flat-field where the wavelength dep endent resp onse for each pixel is known. Assuming a stable detector this mo del should b e verified only once a year. We exp ect that the first metho d may b e used for some individual high S/N while the second metho d will b e used for the DRM programs. The two conventional IFU, which are presumably used for high S/N observations, can b e calibrated every with dedicated lamp observations. The information from the MSA observations can guideline. observations slits and the other month b e used as a

The second metho d will minimise the calibration needs during normal observations. In order to evaluate the feasibility of such a global mo del a detailed knowledge of the detector is needed. For example, information on the following asp ects would b e useful: wavelength variation of small & large scale flat-field, spatial scale variation, temp oral stability and p erhaps most critically, fringing. Assuming favourable detector characteristics, a p ossible calibration strategy, following largely the ACS slitless grism mo de (Pirzkal, Pasquali & Walsh 2002), is outlined in the following: On the ground the wavelength dep endence of pixel-to-pixel variations can b e calibrated with narrow-band filters (a p ossible temp erature dep endence needs to b e investigated). Here the detectors need to b e illuminated directly (MSA removed). Possible fringing needs to b e investigated. Fringing effects may require a mono chromatic illumination of the detector. Medium-scale (20 pixel) flat-field variations intro duced by the detector should also b e established on the ground. In orbit regular flat-field calibrations will b e p erformed with a subset of the p ossible MSA configurations. These in orbit calibrations can b e used to validate the ground based mo del. The stability of the detector flat-field determines the frequency of these in orbit calibrations. We estimate, that monthly checks of 5 configurations with 10 times 100s exp osures are sufficient to validate the ground based mo del. We note, that unstable detector flat-fields can demand much more time intensive calibrations. A complete re-calibration of the wavelength dep endent pixel-to-pixel variations in orbit app ears


­ 24 ­ with the current design very difficult (800 shutter columns, each 1000s exp osure for all disp ersing elements). The large scale flat-field will b e thoroughly calibrated at b eginning of mission with exp osures of astronomical targets and sky observations. Regular in-orbit calibrations (half-yearly frequency) will b e used to verify the validity of the large scale flat-field. As describ ed in Section 3 the requested total sp ectrophotometric accuracy needs to b e as go o d as < 5% for some observing mo des, based on astrophysical considerations (e.g. line diagnostics) which sets tight constraints on the overall flat field accuracy of the order of 1-2% (rms). In order to achieve this accuracy we require that the large scale flat-field variations are <15% over the full FOV and that the variations can b e mapp ed with a low order p olynomial to b etter than 2% (rms).

4.1.4.

Geometric calibrations

The astrometric calibration includes the overall telescop e plate scale solution, the p osition of NIRSp ec in the fo cal plane, and the small scale distortions within NIRSp ec. The general shap e and amount of geometric distortion within NIRSp ec will b e largely known from the instrument mo del. However, precise in orbit validation of the mo del is necessary. The initial characterisation must b e carried out as a function of Fast Steering Mirror and any other parameters that may affect the geometric solution, such as wavelength, temp erature, orientation, fo cus etc. See Table 6 for a summary of the geometric calibrations. The accuracy of the geometric calibration is driven by the target acquisition pro cedure demands. Here a knowledge of the image distortions (sky to detector and MSA to detector) of b etter than 5 mas (rms) is required at all times. In order to allow for efficient op erations a high stability of the distortion mapping is needed. Frequent (monthly, TBC) re-phasing of the telescop e mirrors may have an effect on the distortion mo del. At the moment it is assumed that the distortion mo del will need to b e re-calibrated after each re-phasing pro cedure. In order to meet the stringent demands of the target acquisition we estimate that the distortion maps need to b e determined to an accuracy b etter than 4.5 mas (rms) while the stability of the distortion mo del in b etween re-phasings is assumed to b e b etter than 2 mas (rms). The geometric distortion calibration of the NIRSp ec is a two-step pro cess, requiring (i) measurements of the mapping of the ap erture fo cal plane (carrying the MSA, fixed slits & IFU) onto the detector grid and (ii) of the telescop e fo cal plane on the MSA. The first step can b e accomplished with exp osures of an internal continuum lamp with the mirror selected in the grating wheel and configuring the MSA mask with a sp ecific pattern. For example, a pin hole mask. In order to allow for efficient calibrations in orbit the distortions within the sp ectrographic stage of the NIRSp ec should b e smaller than 5% over the full FOV. The distortions should b e known and easily mappable (i.e. low order p olynomial) to an accuracy of


­ 25 ­

Table 6: Geometric calibrations
Measurement Geometric dist. Geometric dist. verf. Geometric dist. Mirror position TA image Focal plane pos. Detector gaps AFP rel. position Sum
a

Type pointed pointed c. lamp c. lamp pointed pointed pointed pointed

Frequency per year 1 12 2 1000 1000 2 2 2

# Calibrations 3 1 3 1 1 1 1 1

# Exposure (exp. time) 25(100) 25(100) 5(60) 5(12) 5(24) 5(100) 5(100) 5(100)

Total time ksec/year 7.5 30.0 1.8 60.0 120.0 1.0 1.0 1.0

Eff. cost % time 0.07 0.29 0.02 0.57 1.14 0.00a 0.00a 0.00a 2.09

It is assumed that theses calibrations can be derived from the geometric calibrations.

b etter than 4.5 mas (rms) over the full FOV. For these measurements we exp ect a S/N > 50 (p er pixel) to b e necessary which can b e achieved with the internal continuum lamp within 60s. In order to check for chromatic trends at least three different filters should b e used, where we assume that 5 MSA configurations p er filter are sufficient. Since the re-phasing of the primary mirrors will have little affect on this calibration step we estimate that calibrations will b e carried out only twice a year. The second step requires the observation of astrometric fields. For example, a star-cluster where the exact p ositions of individual stars are known. Note that a relative astrometric accuracy of < 5 mas is required (ACS images will b e able to provide this accuracy). Here the "imaging" characteristics of the MSA array can play an imp ortant role. First simulations show that the diffraction on the MSA supp ort structure and the exact p osition of a p oint source within the slit can result in small systematic displacements (1/7 detector pixel, i.e. 15 mas). Dithering of the astrometric target will b e needed to deliver a b etter accuracy on the p osition. Since individual star p ositions will not b e known accurately enough we assume that 10 stars can b e combined to deliver one indep endent p oint in the distortion map fit. We assume that 5 p ointings, with 5 dither steps of each 100 sec in three filters will b e sufficient to meet the requirements. While the full calibration is carried out once a year, we currently assume that after re-phasing distortion calibrations in one filter will b e needed. Similar observations can also b e used to determine the exact fo cal plane p osition of NIRSp ec and the monitor the gaps b etween detectors. The distortion mo del which describ es the optical path from the MSA to the sky can b e determined from the two steps describ ed ab ove. We assume that an accuracy of b etter than 5.02 + 5.02 mas 7.1 mas (rms) can b e achieved. Current instrument designs show that the rep eatability of the grating wheel (carrying the mirror for direct imaging) is a critical issue. Should the designs not pro duce a rep eatability of


­ 26 ­ the wheel p osition to b etter than 5 mas (rms) at the detector level then a contemp oraneous "zerop oint" calibration is necessary. In order to avoid that the whole distortion mo del is affected, we require that the wheel moves to b etter than one detector pixel accuracy (i.e. < 100 mas rms). Once at a a given p osition we assume that the disp ersive elements or the mirror do not move by more than 1/60 (rms) of a resolution element. If a "zero-p oint" calibration of the grating wheel should b e necessary this will b e achieved with sp ecial "L-shap ed" slits in the ap erture fo cal plane. Exp osure times of 5 â 12 sec (5 exp osures for cosmic ray rejection) seem sufficient since the light is not disp ersed. However, this calibration is needed for each target acquisition. The sp ecial slits should b e designed such that neither a MSA failure nor a failure of one detector impacts on the grating wheel calibration. With the help of these calibrations one can also track relative motions of comp onents in the ap erture fo cal plane (e.g., MSA unit with resp ect to fixed slits) or drifts of the detector p ositions. Overall, we exp ect these motions to b e small and the ap erture fo cal plane and the fo cal plane assembly to b e highly stable (i.e. < 1% change p er year over full FOV). Accurate ground based measurements are needed to characterize the ap erture fo cal plane and fo cal plane assembly geometry. In order to p erform the target acquisition (TA), i.e. the exact alignment of the AFP with the science targets, images (mirror selected in grating wheel) of the target field will b e taken. We assume that relatively bright reference targets can b e used thus total exp osure times of 120 sec will b e sufficient. We recommend to take 5 individual images of 24 sec in order to allow for cosmic ray removal. We assume in this do cument 1000 TAs p er year (TBC).

4.1.5.

Spectroscopic calibrations

In this section we discuss the sp ectroscopic calibrations which ensure that the observations meet the scientific requirements. Table 7 summarises the calibrations. An imp ortant assumption for this section is that not all p ossible MSA configurations need to b e calibrated indep endently, but a coarse sampling is sufficient to built up a global mo del. Furthermore we assume that the information on sp ectral trace, disp ersion solution and zero p oints derived from the internal lamps can b e directly applied to observations of astronomical ob jects. Note, that the light of the calibration lamps do es not go through the filters used for external sources. However, the internal lamps may carry their own filters to avoid e.g., second order contamination. Typical targets for NIRSp ec are going to b e faint and not clearly visible on individual exp osures. Furthermore, the sp ectra of emission line ob jects with no detectable continuum flux will b e difficult to trace on the detector. In order to determine the p osition of the sp ectrum from each slit element a priori, an exp osure with an internal continuum source is required ("trace" calibration). The instrument stability should b e such that one can built up a global mo del of the sp ectral trace and re-calibrate it on a yearly basis (deviations from the mo del < 0.1 pixel/year rms at detector


­ 27 ­

Table 7: Sp ectroscopic calibrations
Measurement Spectral trace Dispersion sol. Zero points Through slit image Internal line lamp Line spread func. Shutter contrast Sum Type c. lamp line lamp line lamp pointed pointed pointed pointed Frequency per year 1 6 2600 1000 1 1 1 # Calibrations 9 9 1 1 8 4 1 # Exposure (exp. time) 5 (60) 5 (60) 1 (12) 1 (300) 1 (1000) 7 (1000) 10 (60) Total time ksec/year 2.7 16.2 31.2 300.0 8.0 28.0 0.6 Eff. cost % time 0.03 0.15 0.30 2.85 0.08 0.27 0.01 3.69

level). The distortion mapping requirements describ ed in Section 4.1.4 are sufficient to meet this criterium. For 100 one can calibrate the sp ectral trace with columns of small slits separated by 300 shutters in sp ectral direction and by a few shutters in spatial direction. A few of these masks should b e sufficient to calibrate the full system. For 1000 the calibrations need more individual exp osures since only one column of slits can b e observed at the same time (otherwise sp ectra will overlap). We assume that 5 exp osures of 60s p er disp ersing element are sufficient to p erform the calibration. In order to allow for a optimal extraction we estimate that a knowledge of the trace of 1/10 (rms) of a detector pixel is necessary. We note that an instability of the grating wheel will result in bulk shifts of the trace. If the stability of the grating wheel cannot deliver a high accuracy one can use the contemp oraneous wavelength-zero-p oint calibrations (describ ed b elow) to adjust the trace mo del after each grating wheel move. The calibration of the disp ersion solution can b e p erformed in a similar way as the sp ectral trace calibration with the line-lamp switched on. In order to achieve an overall wavelength calibration accuracy of 1/10 (rms) of a resolution element, the disp ersion solution has to b e known very accurately (<1/20 FWHM, rms). This knowledge can b e achieved by building on an accurate ground based mo del of the disp ersion solution. Here measurements with a fine wavelength grid are required (>100 lines p er sp ectral band). In orbit this mo del will only b e adjusted on wavelength scales determined by the line sampling of the internal lamps ( 10 lines p er band). The stability of this solution should b e such that bi-monthly calibrations are sufficient. While we exp ect the sp ectral trace mo del and the disp ersion solution to b e very stable with time, the wavelength zero p oint calibrations will need to b e p erformed more frequently. In this do cument we take a conservative approach and assume that a calibration is necessary after each prism/grating wheel move. We estimate 2600 grating wheel moves p er year. We assume that the normal line lamps used for the disp ersion solution can also b e used for the zero p oint. By averaging the measurements of several (3) lines one can reduce the exp osure time to ab out 12 sec which corresp onds to the minimal readout time for a full frame. It is envisaged that by fitting a low order p olynomial to the p eaks of the lines one can adjust the global trace mo del in the data-reduction


­ 28 ­ pip eline. The zero-p oint wavelength calibration obtained with an internal lamp is in principle only valid for a science ob ject which fills the slit uniformly. However, typical JWST targets may b e b etter describ ed by a p oint source which will lead to wavelength shifts if the target is not in the middle of the slit. If the exact p osition within the slit is known the effect can b e mo delled or alternatively one can use a dither strategy to smo othly illuminate the slit (see also Section 4.1.3). As detailed in Section 3.4, the goal is to ensure an absolute accuracy of the wavelength calibration of 1/10 of a resolution element. The size of the wavelength shifts for p oint sources should b e minimised by the design of the sp ectrograph. In order to aid the data reduction pip eline we recommend to obtain after each successful target acquisition a "through slit" image of the science targets. In this way the actual p ositions of targets within the slits can b e determined. We estimate exp osures times of 300s. The exp osure time should b e user selectable. The use of internal line sources is mandatory for the sp ectroscopic calibrations since a completely external calibration is virtually imp ossible with an unstable grating wheel. If the line lamps are realized via a Fabry-Perot etalon or similar, the lamps themselves need to b e calibrated on a regular basis with external line sources (see Castertano & Holfeltz 2003, STScI-JWST-R-2002-0005). We assume that the lamps are stable on a yearly timescale and that eight different calibration channels need to b e calibrated. The knowledge of the line spread function is imp ortant for astrophysical diagnostics and for the deconvolution of complex sp ectral features ( 1000, 3000). The detailed diffraction prop erties of the MSA play an imp ortant role here and need to b e mo delled in the future. Extensive ground based testing is needed to confirm the mo del. We require that the line-spread function is mapp ed out with ground based testing at least to 10 -3 (TBC) of the line p eak. The in orbit calibration requires observations of astronomical narrow-line sources observed at several p ositions over the field of view (FOV) and as a function of target p osition within the slit. These calibrations are exp ected to b e carried out in detail at the b eginning of the mission and then verified on a yearly basis. Currently we estimate 5 exp osures are needed to calibrate the line-spread function for each disp ersing element as a function of wavelength. If no suitable line sources can b e found the numb er of observations may increase significantly. The contrast, i.e. the ratio b etween transmitted light with a micro-shutter on and off, is a driving factor in the design of NIRSp ec as it ultimately determines the faintness of the sources observable with the instrument and the fraction of the sp ectra "sp oiled" by the un-suppressed light of bright sources in the field of view. Contrast issues and confusion limits for the JWST NIRSp ec have b een discussed in detail by Freudling et al. (2002). The determination of the contrast is essential for observation planning and will b e extensively carried out on the ground. Orbital verification will take place at the b eginning of the mission and will continue on a yearly basis by comparing observations of bright astronomical sources with the mirror selected in the grating wheel and with the MSA elements on and off. We estimate ten exp osures are sufficient to


­ 29 ­ achieve the calibration.

4.2.

Conclusions

This do cument gives an overview of the basic calibration strategy and time consumption for the NIRSp ec on b oard of JWST. The description of the calibration strategy is presented only in general terms since many details of the sp ectrograph and its detectors are not determined at this p oint. Naturally this do cument will evolve when more information b ecomes available. Typical basic calibrations, including detector, optics, sensitivity, geometric and sp ectroscopic calibrations, which are needed during one year of op eration, are summarised in Tables 1­6. The total amount of calibration time is estimated to b e 14.5%. Using parallel observations for darks one can free up 3%. Critical calibration issues which are identified in this do cument are listed b elow: · Extensive instrument mo delling, including the effects of all ma jor comp onents of the sp ectrograph, is required to design an efficient calibration strategy in orbit. The resulting sp ectrograph mo del needs to b e confirmed with detailed pre-flight observations and a thorough b eginning of mission verification. The Flat-Field (FF) determined by the detectors requires a particularly complex pre-flight calibration. We note that the determination of the full pixel-to-pixel FF as a function of wavelength is extremely difficult, if not imp ossible, to obtain in orbit. In the current strategy the FF calibration will largely rely on a ground based mo del with o ccasional up-dates from data obtained in orbit. · The provision of internal lamps (continuum and line sources) is mandatory since the full calibration with astronomical targets would take up a significant amount of additional exp osure time (>>5%) and put strict limits on the target sample construction. In the case of an unstable grating wheel, the zero p oint calibrations cannot b e carried out without internal lamps. · The scientific requirement for the sp ectrophotometric accuracy is as stringent as 5% (rms) for some observations. In order to meet this criterium with the current design of NIRSp ec, a well defined dithering technique seems necessary. The dithering needs to deliver a semi-uniform illumination of the slit in order to remove throughput variations intro duced by diffraction on the MSA supp ort structure and science target misplacement intro duced by the fixed nature of the MSA grid. The optimal dither pattern is probably dep endent on the detailed diffraction prop erties of the MSA and therefore needs to b e determined with future accurate instrument mo delling.


­ 30 ­ · In order to p erform the target acquisition a very accurate knowledge of the distortion mo del is needed (4.5 mas rms). Furthermore, the distortion map must b e very stable (drift 2 mas rms) in b etween calibrations. It is also essential that the prism/grating wheel shows a high rep eatability (< 1 pix) even if contemp oraneous offset calibrations will b e carried out. · The influence of regular telescop e mirror re-phasing on the fo cus adjustment of NIRSp ec and p otentially the overall calibration concept is unclear at this p oint. Currently it is assumed that the re-phasing (monthly, TBC) will demand re-calibrations of the fo cus, the PSF and the distortion mo del. Therefore, it is imp ortant to provide efficient means to carry out these calibrations. For this purp ose the AFP should carry a set of clear fo cus ap ertures. · Key to an efficient calibration strategy is a well b ehaved sp ectrograph where sensitivity parameters change only smo othly across the FOV and thus a rather coarse sampling of the parameter space is sufficient to calibrate the full system. · The currently available standard stars for sp ectrophometric calibrations are probably not suitable for NIRSp ec. Extensive ground based preparation work will b e needed to establish a useful list of secondary standards. · At this p oint there is little information on the exp ected temp erature changes on JWST during the observations. Future studies will have to consider p otential influences on the calibration concept in more detail (e.g., temp erature sensitivity of detector flat-field).


­ 31 ­

Table 8: Acronym description Acronym Description AFP Ap erture Fo cal Plane DRM Design Reference Mission FF Flat Field FGS Fine Guidance Sensor FOV Field of View FPA Fo cal Plane Array IFU Integral Field Unit James Webb Space Telescop e JWST Micro-Shutter Array MSA NIRCam Near-Infrared Camera NIRSp ec Near-Infrared Sp ectrograph POM Pick-off Mirror Science Instrument SI TA Target Acquisition


­ 32 ­ 5. Acknowledgements

This do cument makes free use of the information provided in the two JWST calibration do cuments provided by Casertano (2001) and Henry & Casertano (2002).

REFERENCES Adelb erger K. L., Steidel C. C., Shapley A. E., Pettini M. 2002, ApJ, accepted Arribas S., Jakobsen P., Fosbury R., Freudling w., Pixels, slit and facets for a MEMS-type spectrograph for NGST, 2002, ISR NGST 2002-01 Casertano S., NGST Calibration Overview, 2001, STScI-NGST-R-0014A Freudling W., Slit throughput and image qualiy of a MEMS-type spectrograph for NGST, 2002, ISR NGST 2002-04, http://www.stecf.org/ngst/stecf.html Freudling W., Cristiani S., Fosbury R.A.E., Jakobsen P., Pirzkal N., Contrast issues and confusion limits for the JWST NIRSpec, 2002, ISR NGST 2002-03, http://www.stecf.org/ngst/stecf.html Henry R., Casertano S., Analysis of Calibration Planning needs, 2002, STScI-NGST-TM-2002-0001 Pirzkal N., Pasquali A., Walsh J., 2002, astro-ph0212021 Regan M., Valenti J., Freudling W., Kuntschner H. & Fosbury R.A.E., JWST Near Infrared Sp ectrograph (NIRSp ec) Op erations Concept Do cument. Version 1.0, June 10, 2003 Steidel C. C., Kollmeier J.A., Shapley A. E., C.W., Dickinson M., Pettini M., 2002, ApJ 570, 526 Veilleux, S., Kim, D.-C., & Sanders, D. B. 1999a, ApJ, 522, 113

A This preprint was prepared with the AAS L TEX macros v5.0.