Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stecf.org/documents/isr/ISR0101.pdf Äàòà èçìåíåíèÿ: Thu May 11 17:33:36 2006 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 20:11:43 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: http astrokuban.info astrokuban

Version 3.0 June 18, 2001

Parasitic Light in NGST instruments: the accuracy of photometric redshifts and the effect of filter leaks in the visible and near IR camera.
Stefano Cristiani ST European Coordinating Facility, European Southern Observatory Karl-Schwarzschild-Strasse 2, D-85748 Garching bei Munchen ¨ Stephane Arnouts European Southern Observatory Karl-Schwarzschild-Strasse 2, D-85748 Garching bei Munchen ¨ Robert A.E. Fosbury ST European Coordinating Facility, European Southern Observatory Karl-Schwarzschild-Strasse 2, D-85748 Garching bei Munchen ¨ ABSTRACT A detailed analysis of NGST multiband photometry applied to the reference case of the study of high-redshift galaxies has been carried out with simulations based on galaxy SEDs derived from the currently available empirical and model templates and on plausible standard filter-sets. In order to correctly identify star forming galaxies in the redshift range 5 z 9 and earlytype galaxies above z 2 and avoid confusion with other SEDs it is mandatory to have photometric information in optical bands, besides a standard IR filter-set like F 110 F 160 K L M . In particular by adding the V 606 I 814 and z Gunn filters a good discrimination is obtained above z 5 for star forming galaxies and z 1 for early-types. The case for an extension of the NGST wavelength domain to the optical range is therefore strongly supported by this analysis. The effects of leaks in the filter blocking have also been investigated. In spite of rather pessimistic assumptions (a constant leak at a level of 10 4 of the peak transmission over the whole spectral range or a leak of Gaussian shape placed at 1.5 times the effective wavelength of the filter with an amplitude of 10 3 of the peak transmission and a width of 5% of the central wavelength) the effects are not dramatic: the accuracy of the determination of photometric redshifts with a standard V 606 I 814 z Gunn F 110 F 160 K L M filter set does not significantly deteriorate for a sample limited to MAB 30. In the range 5 5 z 12 5 a rather high accuracy with a typical error zsim z phot 0 2 is achieved. With the filter blocking achieved by the present standard technology, extended and continuous spectral coverage appears to be the driving factor to maximize the scientific output.
¦ ¦ ¤ ¸ ¡ ¥ ¸ ¦ ¤ ¢ ©¨ § ¸ ¡ § ¡


­2­ 1. Introduction The NGST Design Reference Mission (DRM) is a representative science program elaborated in sufficient detail to aid in the development of functional requirements for the NGST mission. One of the main results of the DRM has been the identification of a large fraction of subprograms (8 out of 12) requiring observations at optical wavelengths. Such an extension poses significant technical challenges. In particular it requires an efficient blocking of photons at undesired wavelengths over a range covering more than three octaves. In the following we examine the consequences of leaks in the transmission of filters for a typical programme of determination of photometric redshifts of high-redshift galaxies.

2.

Baseline of the Instrument

Table 1 shows the main characteristics of the Visible-NIR camera as conceived by the Visible and NIR camera Subcommittee of the "Ad Hoc Science Working Group" (ASWG). Two options are envisaged: an optimal and a bare bones version. More details can be found at the URL http://www.stecf.org/ngst/instruments.html.

3.

Band-passes of the filters
¸

We have assumed as reference bandpasses for the filters of the optical-NIR camera the standard V 606 I 814 z Gunn F 110 F 160 K L M set commonly used on HST with WFPC2 (V 606 I 814) or NICMOS (F 110 F 160) and from the ground (Fig. 1). Details about the filter properties are given in Table 2 The actual choice of the filters to be used in space with NGST may be different but the arguments developed in the following do not depend critically on this detail.

Table 1. Visible-NIR Camera Optimal
¨ ¦ ¸¦

Bare Bones
¦ ¸¦ ¨







¦

range Sampling FOV


0 6 5µm 0 03 44 10

10

5 0µm 0 03 22 10


­3­

Table 2. Name

Properties of the assumed filter set ° (A) 6031 8011 9125 11288 16096 21640 37789 47657
¡

F606 F814 z-Gunn F110 F160 Ks L M

2050 1400 1200 5239 3750 2750 5800 6400

0.096 0.417 0.527 0.698 1.313 1.841 2.917 3.380

F814 F110 F606 Zgunn 1

Ks F160 L

0.8

0.6

0.4

0.2

0

Fig. 1.-- The reference filter set used for the simulations.

¸

FWHM ° (A)

AB Vega correction

M


­4­ 4. Typical SEDs of the sources and the locus of galaxies in multicolor space To simulate a range of galaxy SEDs we have adopted the templates of Coleman et al. (1980) for a typical elliptical, Sbc, Scd and irregular galaxy plus a GISSEL (Bruzual and Charlot 2001) template for a very blue galaxy (with a constant star formation rate and an age of 0.1 Gyr). The SEDs are shown in Fig. 2 and in the following will be referred to as extended CWW templates. Fig. 3 shows the evolution of the apparent magnitude of typical galaxies of 1011 M in the F 606 and L bands and gives a rough idea of the magnitudes of interest. It can be appreciated how an elliptical fades away relatively quickly in the F 606 band, while other star-forming galaxies remain - as observed - significantly brighter. The upper panel of Fig. 4 shows the evolution of the colors of the galaxies described in the previous section in a F 606 F 814 vs. F 110 K plane. The color tracks are limited to z 3 5 for an elliptical SED and to z 5 for Sbc galaxies. With this combination of bandpasses it is straightforward to identify starforming galaxies with z 5 and ellipticals at z 1. The lower panel shows the same in a F 110 F 160 vs. F 160 K plane and demonstrates how, without optical bands, it becomes difficult to isolate star-forming galaxies at z 9.
¸ ¦ ¡ ¸ ¡ ¸ ¸

5.

Photometric redshifts and the effects of "(red) leaks"

In order to quantify the different performance of different sets of filters and the effects of "red-leaks" in a filter configuration extending from the visible to the infrared we have carried out simulations of the assignment of photometric redshifts to a sample of faint galaxies. To this end, we have tried to generate a simulated photometric catalog covering a very large range of galaxy spectral properties and to estimate photometric redshifts from an economic and independent a set of templates as possible. Galaxies were simulated in a redshift interval 0 z 20 according to GISSEL library. This spectral synthesis model is controlled by a number of free parameters. The star formation rate for a galaxy at a given age is governed by the assumed e f ol d ing star formation time-scale . Several values of and galaxy ages are necessary to reproduce the different observed spectral types. For the shape of the initial mass function (IMF) we have assumed a Salpeter law, which turns out not to be a critical choice. In addition to the GISSEL parameterization, we have added the internal reddening of each galaxy by applying the observed attenuation law for local starburst galaxies by Calzetti et al. (1997), with the EB V normalization varying between 0 and 0.2 mag. We have also included the Lyman absorption produced by the intergalactic medium as a function of redshift. More details on this type of simulations can be found in Arnouts et al. (1999). Fig. 5 shows the galaxy counts as a function of the apparent magnitude, the age and redshift distribution, the z mag plane characterizing the simulation. Magnitudes and colors have been derived from the simulated SEDs by convolution with the filter passbands described in Sect.3. Measurement errors have been added, according to the NGST mission simulator (http://www.ngst.stsci.edu/nms/main/),
¸ ¤ ¸


­5­

Fig. 2.-- The spectral energy distribution of five template galaxies.


­6­

Fig. 3.-- Evolution of the apparent magnitude of typical galaxies of 1011M .



­7­



¸





¸

Fig. 4.-- Redshift evolution of the colors of galaxies. Upper panel: F 606 panel: F 110 F 160 vs. F 160 K .
¸ ¸

F 818 vs. F 110

K . Lower


­8­

Fig. 5.-- Characteristic distributions of the simulation of galaxies derived from GISSEL models: differential counts as a function of the apparent magnitude (upper left), redshift distribution (upper and lower right), the z mag plane (lower left). In the upper left panel the continuous and dashed lines refer to the differential distribution in M and I magnitude, respectively. In the upper right panel the dashed line refers to galaxies with ages larger than 1 Gyr, the continuous line to ages between 108 and 109 yr. The colors of the dots in the lower left panel (in the electronic version) correspond to different galaxy types with the same color coding of Fig. 4.
¸


­9­ assuming an exposure time of 104s. From this set of "simulated observed colors" photometric redshifts have been derived using a standard procedure of 2 optimization, comparing the observed fluxes with errors to the extended CWW set of templates described in Sect. 4. The 2 has been computed as:


2

where Fobser ved i is the flux observed in a given filter i, i is its uncertainty, Ft em pl at e i is the flux of the template in the same filter and the sum is over the various filters. The template fluxes have been normalized to the observed ones with the factor A minimizing the 2 value : 2 i
"

A

ved i

Fig. 6 shows the comparison between photometric redshift estimates obtained with a V 606 I 814 z Gunn F 110 F 160 K L M filter set and a filter set limited to the infrared bands. As suggested in the previous section in order to correctly identify star forming galaxies in the redshift range 5 z 9 and avoid confusion with other SEDs it appears mandatory to have, besides the IR filters, photometric information in the optical bands. In the following we will adopt the set V 606 I 814 z Gunn F 110 F 160 K L M as standard. In order to explore the effects of filter leaks the following approach was adopted: 1. two types of leaks have been foreseen: a constant leak at a level of 10 4 of the peak transmission over the whole spectral range and a leak of Gaussian shape placed at 1.5 times the effective wavelength of the filter with an amplitude of 10 3 of the peak transmission and a width of 5% of the central wavelength. The effect of Gaussian leaks on the filter transmission is shown in Fig. 9 2. The leaks have been included in the system spectral response and new magnitudes and colors have been produced with the GISSEL models. 3. The new colors have been used to produce photometric redshifts with the extended CWW templates. No knowledge of the leaks was used in this step, i.e. the system responses assumed in the z-phot estimation are those of Fig. 1, without leaks. The typical errors deriving from photometric uncertainties and unaccounted leaks are illustrated in Fig. 7-8 and 10-11. The dot-dashed, long-dashed, short-dashed and continuous lines correspond to the 95, 90, 68 and 50 percentile of the absolute deviation zsim z phot , respectively. It can be seen that errors are in all cases relatively large at z 5. This is an obvious effect due to the lack of bandpasses bluer than V606, and the consequent difficulty to identify the Lyman break at relatively low redshift and distinguis h it, for ° example, from the 4000 A break at lower redshift.
§ ¸ § ¨ ¤ ¤ ¸ ¸

$

#

i

!




Fobser

Ft

em pl at e i


i

Ft

2 em pl at e i 2 i

"

i

!

¸





Fobser

ved i

A Ft i

em pl at e i

2

(1)

(2)


­ 10 ­

Fig. 6.-- Comparison between the accuracy of the photometric redshift estimates with a filter set including IR and optical bands (upper panel V 606 I 814 z Gunn F 110 J F 160 H K L M ) and a filter set limited to infrared bands (lower panel). The dot-dashed, long-dashed, short-dashed and continuous lines correspond to the 95, 90, 68 and 50 percentile of the absolute deviation zsim z phot , respectively. The galaxy sample is limited to MAB 29.
§ ¸ § ¸


­ 11 ­

Fig. 7.-- Typical errors of the photometric redshifts with a V 606 I 814 z Gunn F 110 J F 160 H K L M filter set. The upper panel shows the result obtained assuming only the standard photometric error, the lower panel includes the effects of a constant leak at a level of 10 4 of the peak transmission over the whole spectral range. The dot-dashed, long-dashed, short-dashed and continuous lines correspond to the 95, 90, 68 and 50 percentile of the absolute deviation zsim z phot , respectively. The galaxy sample is limited to MAB 28.
¸ ¤ § ¸ §


­ 12 ­

Fig. 8.-- Same as Fig. 7, but down to MAB



30.


­ 13 ­

F814 F110 F606 Zgunn 1

Ks F160 L

M

0.8

0.6

0.4

0.2

0

Fig. 9.-- The response of the V 606 I 814 z Gunn F 110 F 160 K L M filter with the addition of a leak of Gaussian shape placed at 1.5 times the effective wavelength of the filter with an amplitude of 10 3 of the peak transmission and a width of 5% of the central wavelength.
¤ ¸


­ 14 ­

Fig. 10.-- Typical errors of the photometric redshifts with a V 606 I 814 z Gunn F 110 F 160 K L M filter set. The upper panel shows the result obtained assuming only the standard photometric error, the lower panel includes the effects of a leak of Gaussian shape placed at 1.5 times the effective wavelength of the filter with an amplitude of 10 3 of the peak transmission and a width of 5% of the central wavelength. The dot-dashed, long-dashed, short-dashed and continuous lines correspond to the 95, 90, 68 and 50 percentile of the absolute deviation zsim z phot , respectively. The galaxy sample is limited to MAB 28.
¸ § ¸ ¤ §


­ 15 ­

Fig. 11.-- Same as Fig. 10 but down to MAB



30.


­ 16 ­ The effect of the leaks, despite the rather pessimistic assumptions, is in general a moderate increase of 0 23 the error level. For example for z 11 and MAB 30 the 68 percentile corresponds to zsim z phot with the standard photometric errors, while it increases at zsim z phot 0 25 with the unaccounted constant 4 leak at a level of 10 of the peak transmission over the whole spectral range. No qualitative change of the accuracy is observed when the leaks are introduced. No relevant difference is also found between the continuous and the Gaussian leak cases. Again for z 11 and MAB 30 the 68 percentile corresponds to zsim z phot 0 27 when the leak of Gaussian shape is added. In the range 5 5 z 12 5 a rather good accuracy is achieved with a typical error zsim z phot 0 2 in all cases down to MAB 30. The situation deteriorates slightly around z 15, but the result is due mainly to the insufficient sampling of the SEDs between the Ks and the M filters, in particular to the "hole" at 2.9 µm. The importance of the leaks, relative to the uncertainty due to photometric errors, decreases with de0 18 0 23 0 23 for the no-leak, constant-leak and creasing flux. At MAB 30 and z 11 zsim z phot Gaussian leak case, respectively.
¦¢ ¦¢ ¦ § ¸ § ¦ § ¸ ¦ § ¥ ¦ ¦ § ¸ ¦ § %¨ § ¸ § & ¤ ¦ § ¸ §

6. Conclusions A detailed analysis of NGST multiband photometry applied to the reference case of the study of highredshift galaxies has been carried out with simulations based on galaxy SEDs derived from the currently available empirical and model templates and on plausible standard filter-sets. In order to correctly identify star forming galaxies in the redshift range 5 z 9 and early-type galaxies above z 2 and avoid confusion with other SEDs it is mandatory to have photometric information in optical bands, besides a standard IR set like F 110 F 160 K L M . In particular by adding the V 606 I 814 and z Gunn filters a good discrimination is obtained above z 5 for star forming galaxies and z 1 for earlytypes. The case for an extension of the NGST wavelength domain to the optical range is therefore strongly supported by this analysis. The effects of leaks in the filter blocking have also been investigated. In spite of rather pessimistic assumptions (a constant leak at a level of 10 4 of the peak transmission over the whole spectral range or a leak of Gaussian shape placed at 1.5 times the effective wavelength of the filter with an amplitude of 10 3 of the peak transmission and a width of 5% of the central wavelength) the effects are not dramatic: the accuracy of the determination of photometric redshifts with a standard V 606 I 814 z Gunn F 110 F 160 K L M filter set is not significantly deteriorated for a sample limited to MAB 28. In the range 5 5 z 12 5 a rather good accuracy a typical error zsim z phot 0 2 is achieved.
¤ ¢ ¡ ¡ ¡ ¸

As standard specifications for leaks are filters are produced, it does not look as though application described here. The interpretation be more seriously compromised. Extended and to maximize the scientific output of the NGST

10 4 and as any larger leaks could leaks constitute a serious problem, at of fields containing highly reddened continuous spectral coverage appears camera.

be measured once the least for the particular sources may, however, to be the driving factor

¤

¦





¦

¸



¦



¤

¨ §

¸

§


­ 17 ­ 7. Acknowledgments

We are grateful to A.Moorwood for enlightening discussions about the performances of IR filters and detectors and a critical reading of the manuscript.

REFERENCES Arnouts S., Cristiani S., Moscardini L., Matarrese S., Lucchin F., Fontana A., Giallongo E., 1999, MNRAS, 310, 540 Bruzual, G., Charlot, S., 2001 (GISSEL) in preparation Calzetti D., 1997, in Waller W.H. et al. eds., The Ultraviolet Universe at Low and High Redshift: Probing the Progress of Galaxy Evolution, AIP Conf. Proc. 408. AIP, Woodbury, 403 Coleman, G.D., Wu, C.C., Weedman, D.W. 1980, ApJS, 43, 393 (CWW)

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