GAIA Spectroscopy, Science and Technology ASP Conference Series, Vol. XXX, 2002 U.Munari ed. Information recovery from overlapping GAIA spectra Tomaˇz Zwitter University of Ljubljana, Dept. of Physics, Jadranska 19, 1000 Ljubljana, Slovenia, [email protected] 3 0 0 Abstract. TheRVSinstrumentaboardGAIAisaslitlessspectrograph, 2 so some spectral tracing overlap is inevitable. We show that this back- n ground can be accurately modeled and subtracted. Radial velocity accu- a racy is not degraded significantly unless the star density is higher than J the value typical for the Galactic plane. 3 1 1 v 1. Introduction 9 0 GAIA RVS is a slitless spectrograph, so some degree of spectrum overlap is 2 always present (Zwitter & Henden 2002). The overlap is proportional to reso- 1 0 lution, but even low resolutions (R ∼ 5000) do not avoid it. On the other hand 3 most astrophysical information other than radial velocity is washed away at res- 0 olutions lower than R = 10000 (e.g. Bono 2002; Thevenin 2002). Overlapped / h stars are different during each transit, so some spectra of the same object suffer p from bad overlaps, others do not. For this purpose it is essential that sidereal - spin and precession periods of the satellite are not commeasurable. o r Information recovery from overlapped spectra can use information from t s other GAIA instruments. In particular, the astrometric position measurements a tell which stars are overlapping at each of the transits of a given field over the : v focal plane. Also the distance between spectra heads (that could be treated as a i X rough “velocity” shift) is accurately known from thestar mappersin the spectro plane. r a Photometricclassification ofstellar spectraallows tochoosecorrectspectral templates for all overlapping spectra. Expected errors will besmall, of theorder of 100 K in temperature, 0.2 in logg and similar in metallicity. This information will build up during the mission, so reduction reiterations will be needed. 2. Radial velocity measurement Radial velocity supplies the sixth component in the position-velocity space, so it is the prime reason for existence of a spectrograph aboard GAIA. We ran a set of simulations to recover radial velocity. The results allowing only for zodiacal light background and read-out noise were published in Zwitter (2002), up-to-date results are discussed in Munari (2002). Zodiacal light and read-out noise are not the only source of background. Spectra of other stars (partly) overlap our spectrum, depending on star density and spectral resolution (Zwitter & Henden 2002). It is important that we know 1 2 Zwitter T. the positions, magnitudes and rough spectral types of overlapping background stars. We assume that: • Stellar positionsareaccurately known. Starmappersaswellasastrometry easily justify this assumption. • Stellarmagnitudesareaccuratelyknown. Typicalerrorswillbe∼ 0.001mag for bright stars and ≤ 0.02 mag for stars of V = 18 (ESA-SCI 2000(4)). This is well below the spectroscopic shot noise, so this assumption is jus- tified. Contemporaneous star mapper flux measurements can supply the required information in case of variable stars. • Stellar types are roughly known. A mismatch of 250 K in temperature and 0.5 in logg and [Fe/H] was assumed. This is some 3-times larger than expected typical final-mission stellar classification errors from photometry (Jordi 2002). So this assumption is justified even for mid-mission analysis. Simulated star types clustered around K1 V, i.e. a typical spectral type of background stars, and their luminosity function followed results of Zwitter & Henden (2002). • Stellar radial velocities are roughly known. The assumed errors were V−14.0 σ(vr) = vo 2.51 , vo = 0.2km/s (1) where σ(vr) is the standard deviation of the difference between assumed and true radial velocities and V is the visual magnitude of the star. This is compatible to mission averaged results for a K1 V star (Zwitter 2002). It turns out that radial velocity spread of background stars is not critical for the final results, so this assumption is justified. Figure1showsexamplesofbackgroundspectra. Verticalaxisisspectralflux inunitsofcontinuumfluxfromaK1VstarofmagnitudeV = 14. Thusthelevel of 0.01 corresponds to a contribution of a V = 19 background star. Randomly chosen examples of spectraat different resolutions (R) and star densities (n) are plotted. Star density n = 1200 (V < 17) stars per square degree is common at o high Galactic latitudes (|b| > 20 ), the value of 6100 is typical for the Galactic 2 plane, while n = 40000 (V < 17) stars/deg is representative of rather high o stellar densities, encountered in ≤ 10% of cases at |b| < 20 (Zwitter & Henden 2002). Number of overlapping background spectra and their flux level generally increases with resolution. One should not conclude that these rather bright and jumpy background spectra jeopardize derivation of radial velocities. The background can be very well modeled from information that is available (positions, magnitudes, rough spectral types and velocities of overlapping stars - see above). Note that this info yields spectral tracings (thin lines in Fig. 1) that are very similar to those of real stars (thick lines). Figure 2 reports final results for different star densities. Note that errors on derived radial velocity do not increase significantly, except for the faintest targets (V > 17) and the highest star densities (n = 40000 (V < 17) stars per square degree). Even those errors could be reduced by filtering spectra so that only regions of spectral lines are cross-correlated. Curves are a bit noisy due to Information recovery from overlapping GAIA spectra 3 6 4 2 0 1 0.8 0.6 0.4 0.2 0 0.2 0.15 0.1 0.05 0 8500 8550 8600 8650 8700 8500 8550 8600 8650 8700 8500 8550 8600 8650 8700 Figure 1. Examples of background spectra of different resolutions (R) and star densities (n). Thick lines: actual spectra. Thin lines: their reconstruction from available information. A small vertical shift was applied to the latter for better legibility. a limited Monte Carlo computing time, but the main result is clear: Spectral lines of background stars do not spoil radial velocity analysis. The background stars merely increase the level of the background, thus increasing its shot noise. 3. Recovery of other astrophysically relevant information GAIA spectrograph will beable to measure other information than radial veloc- ities: abundances of individual elements, rotational velocity, temperature, logg, and metallicity (see Thevenin 2002, Gomboc 2002). All these will be recovered from detailed spectral profiles of individual lines. For this purpose: (a) spec- tral line profiles should be measured with high-enough accuracy; (b) the profile should not be spoiled by notable spectral lines from overlapping background stars. The first condition effectively limits the analysis to bright targets, V < 15 and the second to environments of moderate star density, n < 6000 (V < 17) stars per square degree. 4. Conclusions Radial velocities can be recovered even for faint targets. Because the spectral shape of the background can be modeled accurately, overlapping stars degrade the RV accuracy mainly by increasing the background shot noise. It was found 4 Zwitter T. Figure 2. Radial velocity errors for a mission averaged spectrum of a K1 V star at different resolutions R. (a) no spectral overlaps, (b) observations at star densities characteristic for high Galactic latitudes, (c) for the Galactic plane, and (d) for a high density environment. that overlapping spectra generally do not reduce the radial velocity accuracy substantially, even if observing at high resolution (R ∼ 10000) and close to the Galactic plane. Degrading becomes substantial if the S/N per 1 ˚A bin of the non-overlapped spectrumis < 5. ThishappensatV ∼ 17.5 formission averaged spectra and at V ∼ 15 for single transits. Recovery of other astrophysical informationismoredifficultandmostlylimitedtobrighttargetsinnottoodense environments. For accurate modeling of the background the quality of stellar models as well as an accurate knowledge of flux throughput of the instrument (and filters used to bracket the spectral range) is of utmost importance. References Bono G. 2002, this volume deZeeuwP.T.,GilmoreG.,PerrymanM.A.C.,etal. 2000, GAIA-Composition, Formation and Evolution of the Galaxy, ESA-SCI(2000)4 Gomboc A. 2002, this volume Jordi C. 2002, this volume Munari U. 2002, this volume Thevenin F. 2002, this volume Zwitter T. 2002, A&A386, 748 Zwitter T., Henden A. 2002, this volume