Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stsci.edu/~INS/2010CalWorkshop/pancino.pdf Äàòà èçìåíåíèÿ: Mon Apr 23 22:58:47 2012 Äàòà èíäåêñèðîâàíèÿ: Sat Mar 1 14:27:29 2014 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: m 106

The 2010 STScI Calibration Workshop Space Telescope Science Institute, 2010 Susana Deustua and Cristina Oliveira, eds.

An insight into the flux calibration of Gaia G-band images and BP/RP spectrophotometry
E. Pancino INAF ­ Bologna Observatory, Via Ranzani 1, I-40127 Bologna, Italy Abstract. The Gaia mission is describ ed, focussing on those technical asp ects that are necessary to understand the details of its external (absolute) flux calibration. On board Gaia there will be two spectrophotometers, the blue one (BP) and the red one (RP) covering the range 330-1050 nm, and the white light (G-band) imager dedicated to astrometry. Given the fact that Gaia's focal plane will constitute 105 CCDs and the sources will cross the the focal plane at constant sp eed, at different p ositions in each of the foreseen passages (on average 70­80, but up to 350) in the mission lifetime, the "simple" problem of calibrating the integrated BP/RP and G-band magnitudes and the low resolution BP/RP sp ectra flux turns into a very delicate and complicated issue, including CTI effects, LSF variations across the focal plane and with time, CCD gating to avoid saturation and the like. The calibration model requires a carefully selected set of 200 Sp ectrophotometric Standard Stars (SPSS) with a nominal precision of a few %, with resp ect to Vega.

1.

The Gaia mission

Gaia is a cornerstone mission of the ESA Space Program, presently scheduled for launch in 2012. The Gaia satellite will p erform an all-sky survey to obtain parallaxes and prop er motions to µas precision for ab out 109 point-like sources and determine astrophysical parameters (Teff , log g, E (B - V ), metallicity etc.) for stars down to a limiting magnitude of V 20, plus 2-30 km/s accuracy (dep ending on sp ectral typ e), radial velocities for several millions of stars down to V < 17. Such an observational effort has b een compared to the mapping of the human genome for the amount of collected data and for the impact that it will have on all branches of astronomy and astrophysics. The exp ected end-of-mission astrometric accuracies are almost 100 times b etter than the HIPPARCOS dataset (see Perryman et al. 1997). This exquisite precision will allow a full and detailed reconstruction of the 3D spatial structure and 3D velocity field of the Milky Way galaxy within 10 kp c from the Sun. This will provide answers to long-standing questions ab out the origin and evolution of our Galaxy, from a quantitative census of its stellar p opulations, to a detailed characterization of its substructures (as, for instance, tidal streams in the Halo, see Ibata & Gibson, 2007, Sci. Am., 296, 40), to the distribution of dark matter. The accurate 3D motion of more distant Galactic satellites (as globular clusters and the Magellanic Clouds) will b e also obtained by averaging the prop er motions of many thousands of memb er stars: this will provide an unprecedented leverage to constrain the mass distribution of the Galaxy and/or non-standard theories of gravitation. Gaia will determine direct geometric distances to essentially any kind of standard candle currently used for distance determination, setting the whole cosmological distance scale on an extremely firm basis. 86


Gaia flux calibration Table 1: Exp ected numb ers of sp ecific ob jects observed by Gaia. Typ e Extragalactic sup ernovae Resolved galaxies Quasars Solar system ob jects Brown dwarfs Numb ers 20 000 106 ­107 500 000 250 000 50 000 Typ e Extra-solar planets Disk white dwarfs Astrometric microlensing events Photometric microlensing events Resolved binaries (within 250 p c) Numb ers 15 000 200 000 100 1000 107

87

As challenging as it is, the processing and analysis of the huge data-flow from Gaia is the sub ject of thorough study and preparatory work by the Data Processing and Analysis Consortium (DPAC), in charge of all asp ects of the Gaia data reduction. The consortium comprises more than 400 scientists from 25 Europ ean institutes. Gaia is usually describ ed as a self-calibrating mission, but it also needs external data to fix the zero-p oint of the magnitude system and radial velocities, and to calibrate the classification and parametrization algorithms. Al l these additional data are termed auxiliary data and have to be available, at least in part, three months before launch. While part of the auxiliary data already exist and must only b e compiled from archives, this is not true for several comp onents. To this aim a coordinated program of ground-based observations is b eing organized by a dedicated inter CU (Coordination Unit) committee (Ground Based Observation Group), that promotes synergies and avoids duplications of effort. 1.1. Science goals and capabilities

Gaia will measure the p ositions, distances, space motions, and many physical characteristics of some billion stars in our Galaxy and b eyond. For many years, the state of the art in celestial cartography has b een the Schmidt surveys of Palomar and ESO, and their digitized counterparts. The measurement precision, reaching a few millionths of a second of arc, will be unprecedented. Some millions of stars will be measured with a distance accuracy of better than 1 per cent; some 100 million or more to better than 10 p er cent. Gaia's resulting scientific harvest is of almost inconceivable extent and implication. Gaia will provide detailed information on stellar evolution and star formation in our Galaxy. It will clarify the origin and formation history of our Galaxy. The results will precisely identify relics of tidally-disrupted accretion debris, prob e the distribution of dark matter, establish the luminosity function for pre-main sequence stars, detect and categorize rapid evolutionary stellar phases, place unprecedented constraints on the age, internal structure and evolution of all stellar typ es, establish a rigorous distance scale framework throughout the Galaxy and b eyond, and classify star formation and kinematical and dynamical b ehavior within the Local Group of galaxies. Gaia will pinp oint exotic ob jects in colossal and almost unimaginable numb ers: many thousands of extra-solar planets will b e discovered (from b oth their astrometric wobble and from photometric transits) and their detailed orbits and masses determined; tens of thousands of brown dwarfs and white dwarfs will b e identified; tens of thousands of extragalactic sup ernovae will b e discovered; Solar System studies will receive a massive imp etus through the observation of hundreds of thousands of minor planets; near-Earth ob jects, inner Trojans and even new trans-Neptunian ob jects, including Plutinos, may b e discovered. Gaia will follow the b ending of star light by the Sun and ma jor planets over the entire celestial sphere, and therefore directly observe the structure of space-time ­ the accuracy of its measurement of General Relativistic light b ending may reveal the long-sought scalar correction to its tensor form. The PPN parameters and , and the solar quadrup ole


88

Pancino

Figure 1: The two Gaia telescop es, mounted on a compact torus, p oint towards two lines of sight separated by 106.5o , and converging on the same focal plane. c ESA moment J2, will b e determined with unprecedented precision. All this, and more, will b e obtained through the accurate measurement of star p ositions. We summarize some of the most interesting ob ject classes thatwill b e observed by Gaia, with estimates of the exp ected total numb er of ob jects, in Table 1. For more information on the Gaia mission: http://www.rssd.esa.int/Gaia. More information for the public on Gaia and its science capabilities are contained in the Gaia information sheets1 . An excellent review of the science p ossibilities op ened by Gaia can b e found in Perryman et al. (1997). 1.2. Launch, timeline and data releases

The first idea for Gaia b egan circulating in the early 1990, culminating in a prop osal for a cornerstone mission within ESA's science program submitted in 1993, and a workshop in Cambridge in June 1995. By the time the final catalogue will b e released approximately in 2020, almost two decades of work will have elapsed b etween the original concept and mission completion. Gaia will b e launched by a Soyuz carrier (rather than the originally planned Ariane 5) in 2012 from French Guyana and will start op erating once it reaches its Lissa jous orbit around L2 (the unstable Lagrange p oint of the Sun and Earth-Moon system), a month later. Two ground stations will receive the compressed Gaia data during the 5 years2 of op eration: Cebreros (Spain) and Perth (Australia). The data will then b e transmitted to the main data centers throughout Europ e to allow for data processing. We are presently in technical development phase C/D, and the hardware is b eing built, tested and assembled. Software development started in 2006 and is presently producing and testing pip elines with

1 2

http://www.rssd.esa.int/index.php?pro ject=GAIA&page=Info sheets overview. If ­ after careful evaluation ­ the scientific output of the mission will b enefit from an extension of the op eration p eriod, the satellite should b e able to gather data for one more year, remaining within the Earth eclipse.


Gaia flux calibration

89

Figure 2: Left: the scanning law of Gaia during main op erations; Right: the average numb er of passages on the sky, in ecliptic coordinates. c ESA the aim of delivering to the astrophysical community a full catalogue and dataset ready for scientific investigation. Apart from the end-of-mission data release, foreseen around 2020, some intermediate data releases are foreseen. In particular, there should b e one first intermediate release covering either the first 6 months or the first year of op eration, followed by a second and possibly a third intermediate release, that are presently b eing discussed. The data analysis will proceed in parallel with observations, the ma jor pip elines re-processing all the data every 6 months, with secondary cycle pip elines ­ dedicated to sp ecific tasks ­ op erating on different timescales. In particular, verified science alerts, based on unexp ected variability in flux and/or radial velocity, are exp ected to b e released within 24 hours from detection, after an initial p eriod of testing and fine-tuning of the detection algorithms. 1.3. Mission concepts

During its 5-year op erational lifetime, the satellite will continuously spin around its axis, with a constant sp eed of 60 arcsecsec. As a result, over a p eriod of 6 hours, the two astrometric fields of view will scan across all ob jects located along the great circle p erp endicular to the spin axis (Figure 2, left panel). As a result of the basic angle of 106.5o separating the astrometric fields of view on the sky (Figure 1), ob jects transit the second field of view with a delay of 106.5 minutes compared to the first field. Gaia's spin axis does not p oint to a fixed direction in space, but is carefully controlled so as to precess slowly on the sky. As a result, the great circle that is mapp ed by the two fields of view every 6 hours changes slowly with time, allowing rep eated full sky coverage over the mission lifetime. The b est strategy, dictated by thermal stability and p ower requirements, is to let the spin axis precess (with a period of 63 days) around the solar direction with a fixed angle of 45o . The ab ove scanning strategy, referred to as "revolving scanning", was successfully adopted during the Hipparcos mission. Every sky region will b e scanned on average 70-80 times, with regions lying at ±45o from the Ecliptic Poles b eing scanned on average more often than other locations. Each of the Gaia targets will b e therefore scanned (within differently inclined great circles) from a minimum of approximately 10 times to a maximum of 250 times (Figure 2, right panel). Only p oint-like sources will b e observed, and in some regions of the sky, like the Baade's window, Centauri or other globular clusters, the star density of the two combined fields of view will b e of the order of 750 000 or more p er square degree, exceeding the storage capability of the onb oard processors, so Gaia will not study in detail these dense areas.


90

Pancino

Figure 3: The 105 on the Gaia focal plane. c ESA 1.4. Focal plane

Figure 3 shows the focal plane of Gaia, with its 105 CCDs, which are read in TDI (Time Delay Integration) mode: ob jects enter the focal plane from the left and cross one CCD in 4 seconds. Apart from some technical CCDs that are of little interest in this context, the first two CCD columns, the Sky Mapp ers (SM), p erform the on-b oard detection of point-like sources, each of the two columns being able to see only one of the two lines of sight. After the ob jects are identified and selected, small windows are assigned, which follow them in the astrometric field (AF) CCDs where white light (or G-band) images are obtained (Section 1.5.). Immediately following the AF, two additional columns of CCDs gather light from two slitless prism sp ectrographs, the blue sp ectrophotometer (BP) and the red one (RP), which produce disp ersed images (Section 1.6.). Finally, ob jects transit on the Radial Velocity Sp ectrometer (RVS) CCDs to produce higher resolution sp ectra around the Calcium Triplet (CaT) region (Section 1.7.). 1.5. Astrometry

The AF CCDs will provide G-band images, i.e., white light images where the passband is defined by the telescop e optics transmission and the CCDs' sensitivity, with a very broad combined passband ranging from 330 to 1050 nm and p eaking around 500­600 nm (Figure 4). The ob jective of Gaia's astrometric data reduction system is the construction of core mission products: the five standard astrometric parameters, p osition (, ), parallax ( ), and prop er motion (µ , µ ) for all observed stellar ob jects. The exp ected end-of-mission precision in the prop er motions is exp ected to b e b etter than 10 µas for G<10 mag stars, 25 µas for G=15 mag, and 300 µas for G=20 mag. For parallaxes, considering a G=12 mag star, we can exp ect to have distances at b etter than 0.1% within 250 p c, 1% within 2700 p c, and 10% within 10 kp c. To reach these end-of-mission precisions, the average 70­80 observations p er target gathered during the 5-year mission duration will have to b e combined into a single, global, and self-consistent manner. 40 Gb of telemetry data will first pass through the Initial Data Treatment (IDT) which determines the image parameters and centroids, and then perform ob ject cross-matching. The output forms the so-called One Day Astrometric Solution (ODAS), together with the satellite attitude and calibration, to sub-milliarcsecond accuracy. The data are then written to the Main Database.


Gaia flux calibration

91

Figure 4: Left: the passbands of the G-band, BP and RP; Right: a simulated RP disp ersed image, with a red rectangle marking the window assigned for compression and ground telemetry. c ESA The next step is the Astrometric Global Iterative Solution (AGIS) processing. AGIS processes together the attitude and calibration parameters with the source parameters, refining them in an iterative procedure that stops when the adjustments b ecome sufficiently small. As soon as new data come in, on the basis of 6 months cycles, all the data in hand are reprocessed together from scratch. This is the only scheme that allows for the quoted precisions, and it is also the philosophy that justifies Gaia as a self-calibrating mission. The primary AGIS cycle will treat only stars that are flagged as single and non-variable (exp ected to b e around 500 millions), while other kinds of ob jects will be computed in secondary AGIS cycles that utilize the main AGIS solution. Dedicated pip elines for sp ecific kinds of ob jects (asteroids, slightly extended ob jects, variable ob jects and so on) are b eing put in place to extract the b est p ossible precision. Owing to the large data volume (100 Tb) that Gaia will produce, and to the iterative nature of the processing, the computing challenges are formidable: AGIS processing alone requires some 1021 FLOPs which translates to runtimes of months on the ESAC computers in Madrid. 1.6. Sp ectrophotometry

The primary aim of the photometric instrument is mission critical in two resp ects: (i) to correct the measured centroids p osition in the AF for systematic chromatic effects, and (ii) to classify and determine astrophysical characteristics of all ob jects, such as temp erature, gravity, mass, age and chemical comp osition (in the case of stars). The BP and RP sp ectrophotometers are based on a disp ersive-prism approach such that the incoming light is not focussed in a PSF-like sp ot, but disp ersed along the scan direction in a low-resolution sp ectrum. The BP op erates b etween 330­680 nm while the RP b etween 640-1000 nm (Figure 4). Both prisms have appropriate broad-band filters to block unwanted light. The two dedicated CCD strip es cover the full height of the AF and, therefore, all ob jects that are imaged in the AF are also imaged in the BP and RP. The resolution is a function of wavelength, ranging from 4 to 32 nm/pix for BP and 7 to 15 nm/pix for RP. The sp ectral resolution, R=/ ranges from 20 to 100 approximately. The disp ersers have b een designed in such a way that BP and RP spectra are of similar sizes (45 pixels). Window extensions meant to measure the sky background are implemented. To compress the amount of data transmitted to the ground, all the BP and RP sp ectra ­ except for the brightest stars ­ are binned on chip in the across-scan direction, and are transmitted to the ground as one-dimensional sp ectra. Figure 4 shows a simulated RP sp ectrum, unbinned, b efore windowing, compression, and telemetry.


92

Pancino

Figure 5: Simulated RVS end-of-mission sp ectra for the extreme cases of 1 single transit (b ottom sp ectrum) and of 350 transits (top sp ectrum). c ESA The final data products will b e the end-of-mission (or intermediate releases) of global, combined BP and RP sp ectra and integrated magnitudes MBP and MRP . Ep och sp ectra will b e released only for sp ecific classes of ob jects, such as variable stars and quasars, for example. The internal flux calibration of integrated magnitudes, including the MG magnitudes as well, is exp ected at a precision of 0.003 mag for G=13 stars, and for G=20 stars goes down to 0.07 mag in MG , 0.3 mag in MBP and MRP . The external calibration should b e p erformed with a precision of the order of a few p ercent (with resp ect to Vega). 1.7. High-resolution sp ectroscopy

The primary ob jective of the RVS is the acquisition of radial velocities, which combined with positions, proper motions, and parallaxes will provide the means to decipher the kinematical state and dynamical history of our Galaxy. The RVS will provide the radial velocities of ab out 100­150 million stars up to 17th magnitude with precisions ranging from 15 km s-1 at the faint end, to 1 km s -1 or b etter at the bright end. The sp ectral resolution, R=/ will b e 11 500. Radial velocities will b e obtained by cross-correlating observed sp ectra with either a template or a mask. An initial estimate of the source atmospheric parameters will b e used to select the most appropriate template or mask. On average, 40 transits will b e collected for each ob ject during the 5-year lifetime of Gaia, since the RVS does not cover the whole width of the Gaia AF (Figure 3). In total, we exp ect to obtain some 5 billion sp ectra (single transit) for the brightest stars. The analysis of this huge dataset will b e complicated, not only b ecause of the sheer data volume, but also b ecause the sp ectroscopic data analysis relies on the multi-ep och astrometric and photometric data. The covered wavelength range (847-874 nm) (Figure 5) is a rich domain, centered on the infrared calcium triplet: it will not only provide radial velocities, but also many stellar and interstellar diagnostics. It has b een selected to coincide with the energy distribution peaks of G and K type stars, which are the most abundant targets. In early typ e stars, RVS sp ectra may contain also weak Helium lines and N, although they will b e dominated by the Paschen lines. The RVS data will effectively complement the astrometric and photometric observations, improving ob ject classification. For stellar ob jects, it will provide atmospheric parameters such as effective temp erature, surface gravity, and individual abundances of key


Gaia flux calibration

93

elements such as Fe, Ca, Mg, Si for millions of stars down to G 12. Also, Diffuse Interstellar Bands (DIB) around 862 nm will enable the derivation of a 3D map of interstellar reddening. 1.8. The DPAC

ESA will take care of the satellite design, build and testing phases, of launch and op eration, and of the data telemetry to the ground, managing the ESAC data center in Madrid, Spain. The data treatment and analysis is the resp onsibility of the Europ ean scientific community. In 2006, the announcement of opp ortunity op ened by ESA was successfully answered by the Data Processing and Analysis Consortium (DPAC), a consortium that presently consists of more than 400 scientists in Europ e (and outside) and more than 25 scientific institutions. The DPAC executive committee (DPACE) oversees the DPAC activities; work has b een organized among Coordination Units (CU) in charge of different asp ects of data treatment: · CU1. System Architecture (manager: O' Mullane), dealing with all asp ects of hardware and software, and coordinating the framework for software development and data management. · CU2. Data Simulations (manager: Luri), in charge of the simulators of various stages of data products, necessary for software development and testing. · CU3. Core pro cessing (manager: Bastian), developing the main pip elines such as IDT, AGIS and astrometry processing in general. · CU4. Ob ject Pro cessing (managers: Pourbaix & Tanga), for the processing of ob jects that require sp ecial treatment such as minor b odies of the Solar system, for example. · CU5. Photometric pro cessing (manager: van Leeuwen), dedicated to the BP, RP, and MG processing and calibration, including image reconstruction, background treatment, and crowding treatment, among others. · CU6. Sp ectroscopic Pro cessing (managers: Katz & Cropp er), dedicated to RVS processing and radial velocity determination. · CU7. Variability Pro cessing (managers: Eyer & Evans & Dubath), dedicated to processing, classification and parametrization of variable ob jects. · CU8. Astrophysical Parameters (managers: Bailer-Jones & Thevenin), developing ob ject classification software and, for each ob ject class, software for the determination of astrophysical parameters. · CU9. Catalogue Pro duction and Access (to b e activated in the near future), resp onsible for the production of astrophysical catalogues and for the publication of Gaia data to the scientific community. These are flanked by a few working groups (WG) that deal with aspects that are common among the various CUs, such as the GBOG (Ground Based Observerations Group), which coordinates the ground based observations for the external calibration of Gaia) or of general interest (such as the Radiation task force, serving as the interface b etween DPAC and industry in all matters related to CCD radiation tests).


94 2.

Pancino The flux calibration of Gaia data

Calibrating (sp ectro)photometry obtained from the usual typ e of ground based observations (broadband imaging, sp ectroscopy) is not a trivial task, but the procedures are well known (see e.g., Bessell, 1999) and several scientists have develop ed sets of standard stars appropriated for the more than 200 photometric systems known, and for sp ectroscopic observations. Generally, magnitudes are calibrated to a standard system with equations in the form M = m + ZP + (color )+ (air mass) where M is the calibrated magnitude in a chosen photometric band, m, magnitude in the same (or very similar) band, is the color term and coefficient due to the Earth's atmospheric extinction. For the sp ectra, effect on the observed sp ectral energy distribution (SED) is parametrized S
obs

the instrumental , the extinction the instrumental as

() = R() S ()

where the observed SED, Sobs (), is the result of the convolution of the "true" SED, S(), with all the instrumental (transmissivity, quantum efficiencies) effects, which are empirically determined in the form of a resp onse curve, R(,) through the use of sp ectrophotometric standard stars (SPSS). In the case of Gaia, several instrumental effects ­ much more complex than those usually encountered ­ redistribute light along the SED of the observed ob jects. In particular these are: the TDI integration mode, the large focal plane, radiation damage and resulting CTI (charge transfer inefficiencies), and that the whole instrumental model is well known only b efore launch. 2.1. Challenges

The most difficult Gaia data to calibrate are the BP and RP sp ectra, requiring a new approach to the derivation of the calibration model (Section 2.3.) and to the SPSS needed to p erform the actual calibration (Section 3.). The large focal plane with its large numb er of CCDs makes it so that different observations of the same star will b e generally on different CCDs, with different quantum efficiencies. Also, each CCD is in a different p osition, with different optical distortions, optics transmissivity and so on. Therefore, each wavelength and each p osition across the focal plane has its (sometimes very different) PSF (p oint spread function). The TDI and continuous reading mode, combined with the need of compressing the data b efore on-ground transmission, make it necessary to translate the full PSF into a linear (compressed into 1D) LSF (line spread function), which of course add complications. In-flight instrument monitoring is foreseen, but never comparable to the full characterization that will b e p erformed b efore launch, so the real instrument ­ at a certain observation time ­ will b e different from the theoretical one assumed initially, and this difference will change with time. Radiation damage deserved sp ecial mention as it is one of the most imp ortant factors in the time variation of the instrument model. It has particular impact on the BP and RP disp ersed images since the ob jects travel along the BP and RP CCD strips in a direction that is parallel to the sp ectral disp ersion (wavelength coordinate). Radiation damage causes traps that subtract photons from each passing ob ject at a p osition corresp onding to a certain wavelength. Slow traps release the trapp ed charges once the ob ject is already passed, while fast traps can release the charges within the same ob ject, but at a different wavelength. Given the low resolution, one pixel can cover as much as 15­20 nm (dep ending on the wavelength) and therefore the net effect of radiation damage can b e to alter significantly the SED of some sp ectra. Possible solutions under testing are the equivalent of CCD preflashing, the statistical modeling of the traps b ehavior and the fact that different transits for the same ob ject will b e affected differently by CTI effects, allowing for a certain degree of correction through average or median sp ectra. Finally the PSF/LSF itself is generally


Gaia flux calibration

95

larger than one Gaia pixel in the BP/RP sp ectra, introducing a large LSF smearing effect, i.e., the spread of photons with one particular wavelength into a large range of wavelengths. In this pap er, we will adopt the current Gaia calibration philosophy, where most of these instrumental effects are taken into account during the so-called internal flux calibration. A large numb er of well b ehaved stars (internal standards) observed by Gaia will b e used to rep ort all observations to a reference instrument, on the same instrumental flux and wavelength scales. All transits for each ob ject observed by Gaia will b e then averaged to produce one single BP and RP sp ectrum for each ob ject, with its integrated instrumental magnitudes: MG , MBP , and MRP . Only for sp ecific classes of ob jects will ep och sp ectra and magnitudes will b e released, with variable stars as an obvious example. The mean and ep och sp ectra will b e mostly free from many of the problems examined just ab ove, but they will still contain residuals due to the imp erfect knowledge of the real instrument at each precise moment of time, and the most significant effects are exp ected to b e the LSF smearing and the CTI effects. In this paradigm, the internal and external flux and wavelength calibrations are treated as two entirely separate and consecutive pieces of the CU5 photometric pip eline, with different calibration models. We always start from internally calibrated BP/RP sp ectra and MG , MBP , and MRP magnitudes, without giving imp ortance to the exact way they are produced. Presently, two alternative approaches are b eing considered to maximize the precision of the global calibration procedure: the first one is a hybrid model that partially combines internal and external models (Montegriffo et al. 2010), while the second is the so-called ful l forwarding model (Carrasco et al. 2010, in preparation), using the same calibration model for b oth the internal and external calibration. 2.2. The external calibration teams

Two development units (DU) within CU5 (Photometric processing), are dedicated to the external calibration of Gaia photometry. They are DU13: Instrument absolute response characterization: ground-based preparation, coordinated by E. Pancino and DU14, Instrument absolute response characterization: definition and application', coordinated by C. Cacciari. They are b oth based in Bologna, Italy, in collab oration with the Bologna, Barcelona, and Groningen Universities. The actual team memb ers at the time of writing are: G. Altavilla, M. Bellazzini, A. Bragaglia, C. Cacciari, J. M. Carrasco, G. Cocozza, L. Federici, F. Figueras, F. Fusi Pecci, C. Jordi, S. Marinoni, P. Montegriffo, E. Pancino, S. Ragaini, E. Rossetti, S. Trager. 2.3. Disp ersion Matrix basic definition

If we concentrate now on the mean, internally calibrated BP/RP sp ectra calibration, we can write:
N

S

obs

(I ) =
i =0

T (i ) Li (I - P (i )) S

tr ue

(i )

where Sobs and Strue are the observed and "true" SEDs resp ectively, expressed the first in Gaia pixels I and the second in wavelength intervals i corresp onding to the actual sampling of the SPSS used in the flux calibration process. T(i ) is a combination of all the instrument and telescop e transmissivity functions and ap erture, while L is the LSF at a certain i , centered at the appropriate I pixel, but of course calculated over the whole wavelength interval from i =0 to N (the total numb er of samples in the tabulated SPSS sp ectrum). Such an equation can b e written in its much simpler matrix form: S
obs

=DâS

tr ue


96

Pancino

Figure 6: Graphical example of a disp ersion matrix D, derived with a simulated SPSS set for the BP (left) and RP (right) instruments, on an arbitrary color scale. c ESA where D is called a "Disp ersion Matrix", an ob ject that can b e determined if Sobs and Strue are known, i.e., using a well defined set of SPSS observed by Gaia, that also have well known SED (see b elow). Once D is prop erly determined, it can b e inverted to convert each mean, internally calibrated BP/RP sp ectrum3 (Sobs ) into a flux calibrated sp ectrum Strue S
tr ue

=D

-1

The main advantage of this approach is that D contains (and therefore corrects empirically for) the actual effects of LSF smearing ­ even if the real shap e of the LSF is not perfectly known a priori. More than that, the effective LSF ­ as determined with the chosen SPSS set ­ at each wavelength can b e extracted from each column of the matrix. The matrix rows represent instead the effective passbands corresp onding to each Gaia pixel, including the full effect of LSF smearing. This p eculiar prop erty of the disp ersion matrix makes it the b est (and p ossibly only) solution to the external calibration of Gaia BP/RP sp ectra. By definition, the disp ersion matrix D contains also the actual disp ersion function, which can b e seen in Figure 6 as the curved structure close to the diagonal of the matrix. Finally, an imp ortant by-product of the describ ed calibration model is the absolute wavelength calibration of the BP/RP sp ectra to a precision of at least a few tenths of a Gaia pixel4 (Montegriffo & Bellazzini 2009b), which is automatically p erformed together with the absolute flux calibration. There are a few problems in the use of the disp ersion matrix as prop osed. We will discuss in the three following Sections the two most imp ortant ones and their prop osed solutions: (1) the matrix is rectangular and its inversion is not so straightforward; (2) the matrix needs a set of indep endent vectors to b e determined in a non-degenerate way (which also implies that the set of SPSS must b e carefully chosen). 2.4. Smoothing the input SPSS sp ectra

âS

obs

Clearly, the disp ersion matrix is a rectangular matrix: the Gaia observations have a smaller numb er of samples (pixels) than the SPSS sp ectra used to build their calibration model (wavelength sampling). Inverting a rectangular matrix is a non-trivial task so if we want
3

Incidentally, for some ob ject classes that need it, such as variable stars, single transits ­ the so called epoch spectra ­ will b e published. The describ ed calibration model can b e applied also to single ep och sp ectra once they have b een internally calibrated, i.e., rep orted to a common, instrumental scale of flux and wavelength. This first estimate of the wavelength zero p oint and scale precision is based on a slightly outdated calibration model formulation, an therefore has to b e considered just as an upp er limit to a more realistic uncertainty (Montegriffo, 2010, private communication.

4


Gaia flux calibration

97

Figure 7: On b oth panels, grey dots are simulated sp ectra of different metallicity and reddening (this explains the parallel sequences). Green dots are white dwarfs and hot sub dwarfs, while red symb ols are different typ es of red stars with significant absorption features, i.e., molecular bands. Abscissae represent the BP­RP color, ordinates are the difference b etween the "known" magnitude of the used SEDs and the "calibrated" ones obtained with a disp ersion matrix. Left: all p oints are calibrated with a matrix built only with hot sp ectral typ es (green dots): the reddest stars are calibrated with an error of 0.15 mag and more. Right: all p oints are calibrated with a matrix determined using also 10 red SPSS with absorption features (red symb ols): all stars with SEDs similar to the 10 red stars (2 mag obs

= De â S

smooth

However, De can still b e non-diagonal or degenerate and b efore proceeding we must find the criteria to build the b est p ossible De with the data in hand. 2.5. The ideal SPSS set

One reason why the disp ersion matrix is non-diagonal, is that the SPSS adopted set can never be an orthogonal set of independent calibrators: stars are all similar to each other, they have all a black-b ody like continuum with some features (absorption and emission lines or bands). As a result, if a disp ersion matrix is built with a particular set of SPSS, such as white dwarfs and hot sub dwarfs (the ideal calibrators in the classic sp ectroscopic observations), it will b e able to calibrate prop erly only ob jects with similar sp ectra, i.e., relatively smooth, with some absorption lines in the blue part of the sp ectrum. An example of the ab ove case is shown in Figure 7, where two different disp ersion matrices are used to calibrate the same set of simulated Gaia observations. In the first


98

Pancino

Figure 8: Left: example of a degenerate disp ersion matrix, De (which is square, see text), where the diagonal is drowned into noise-like patterns due to the degenerate (non indep endent) set of SPSS used to construct it. Right: an example of a diagonal disp ersion matrix, obtained with an appropriate SPSS set and with the use of a nominal disp ersion matrix to further reduce degeneracy (see text). c ESA case (left panel of Figure 7), a disp ersion matrix is built using only white dwarfs and hot sub dwarfs, with a minority of solar typ e stars, and it can clearly b e seen that red stars with deep absorption bands are calibrated with an error of 0.15 mag at least, failing the sp ecified requirements. In the second case (right panel of Figure 7), a small numb er (10) of red stars with deep absorption bands are included in the SPSS set used to build a second disp ersion matrix. The second disp ersion matrix is able to calibrate all red stars with absorption features to b etter than 1%, exceeding the requirements. This example shows the imp ortance of sp ectral features in the SPSS set used in construction of the disp ersion matrix. Hot stars have prominent absorption lines in the blue, but no features in the red. The addition of a few red stars with absorption bands "trains" the matrix in the calibration of stars with features in the red (effectively reducing degeneracy). Similarly, problems are encountered in the calibration of emission line ob jects (p eculiar hot stars and quasars, for example). But it is quite difficult to include these ob jects into the SPSS set since they are often variable. Even if several typ es of ob jects are included when determining the disp ersion matrix, other effects can have a large impact on the degeneracy, such as edge effects. For these reasons, the accurate choice of the SPSS set is crucial, but does not solve the problem of degeneracy once and for all. 2.6. Nominal disp ersion matrix

To further reduce degeneracy of the effective disp ersion matrix De , we can use other constraints such as the fact that we know most asp ects of the instrument from pre-launch characterization. These include the quantum efficiency of the CCDs, the optical layout and transmissivity, the nominal LSF at various p ositions along the focal plane and at different wavelengths, the nominal disp ersion function and its variation along the focal plane. The slow change of these with time can also b e monitored to a certain extent, and included in the models. We can therefore separate the disp ersion matrix in a part that is theoretically modeled based on pre-launch instrument description and on its (partially reconstructed) variation with time, which we call Dn or nominal dispersion matrix, and in a part that is completely


Gaia flux calibration

99

unknown, which can b e considered as a correction matrix K, made of the residual corrections after the nominal model is taken into account (Montegriffo et al. 2010) De = K â Dn The nominal matrix will b e clean: diagonal and non-degenerate (see Figure 8). The correction matrix will b e partially degenerate, but all signal that lies far away from the diagonal can b e safely considered spurious (the system varies in a continuous way, the corrections must b e "small" compared to the nominal system), and the part of the correction matrix close to its diagonal can b e easily modeled. To summarize all the previously defined steps, once an appropriate SPSS set is chosen, the calibration model b ecomes S
obs

= De â S

smooth

= K â Dn â S

smooth

and the matrices involved can b e easily inverted to calibrate all Gaia observations since they are all square and (almost completely) diagonal. 2.7. Integrated magnitudes

A classical approach can b e adopted for the absolute flux calibration of integrated MG , MBP , and MRP magnitudes (Ragaini et al., 2009a,b) in the form M = m + ZP where M is the calibrated magnitude, m the internally calibrated one observed by Gaia, and ZP is the required zero-p oint. No significant color term app ears necessary. However, if we consider that an integrated magnitude M is the convolution of the sp ectral distribution Strue and the effective passband B, we can calibrate integrated magnitudes with the same approach adopted for BP/RP sp ectra, with a much more homogeneous procedure from the p oint of view of pip eline code writing M =S
tr ue

âB

Since generally the passband B is sampled differently than the SPSS flux table Strue , we must smooth to one or b oth Strue and B. Similarly to the case of BP/RP sp ectra, we can split B into two comp onents B = K â Bn where K' is a correction vector, made of the actual residuals to a theoretically known ­ or nominal ­ effective passband Bn , known b efore launch and slowly varying with time due to several causes, the most imp ortant one b eing the decrease in CCD quantum efficiency due to radiation damage. With this kind of treatment, the problem b ecomes a simple least square fitting problem to derive the unknown K' vector (Ragaini et al., 2010, in preparation). 2.8. RVS calibration

The p ossibility of flux calibrating the RVS sp ectra has b een so far considered a secondary problem, since b oth radial velocities and astrophysical parameters can b e derived without the need of an absolute flux scale attached to the sp ectra. A preliminary set of considerations (Trager, 2010) shows that in principle the SPSS grid for the calibration of Gaia G-band and BP/RP data, that is presently under construction, should b e sufficiently sampled to ensure a flux calibration of RVS sp ectra as well. We exp ect the topic to b e further explored by CU6 (Sp ectroscopic processing) in the near future, but we will not consider RVS sp ectra flux calibration in this pap er.


100

Pancino

Figure 9: The RA/Dec distribution (right). The three Pillars are marked the ecliptic p oles are marked in green prop ortional to the SPSS magnitude. are marked in b oth panels. c ESA 3.

of Primary (left) and Secondary SPSS candidates in red in the left panel, while the targets close to in the right panel. The size of symb ols is inversely The two strip es at ±45 deg from the Ecliptic Poles

The Gaia grid of sp ectrophotometric standard stars

From the ab ove discussion, it is clear that the Gaia SPSS grid has to b e chosen with great care. The Gaia SPSS, or b etter their reference flux tables (corresp onding to Strue in the previous Sections) should conform to the following general requirements (van Leaven et al. 2010): · Resolution R=/ 1000, i.e., they should over sample the Gaia BP/RP resolution by a factor of 4­5 at least; · Wavelength coverage: 330­1050 nm; · Typical uncertainty in the absolute flux scale, with resp ect to the assumed calibration of Vega, of a few p ercent, excluding small troubled areas in the sp ectral range (telluric bands residuals, extreme red and blue edges), where it can b e somewhat worse. The total numb er of SPSS in the Gaia grid should b e of the order of 200­300 stars, including a variety of sp ectral typ es. Clearly, no such large and homogeneous dataset exists in the literature yet5 . It is therefore necessary to build the Gaia SPSS grid with new, dedicated observations. We describ e the characteristics of the Gaia SPSS and of the dedicated observing campaigns in the following sections. 3.1. SPSS Candidates

We have followed a two step approach (Bellazzini et al. 2007) that first creates a set of Primary SPSS, i.e., well known SPSS that are calibrated on the three Pil lars of the CALSPEC6 set, describ ed in Bohlin et al. (1995,2007), and tied to the Vega flux calibration
5

The CALSPEC database (Bohlin, 2007) is not large enough for our purp ose, esp ecially considering the strict criteria describ ed b elow. Its extension to more than 100 SPSS is eagerly awaited, but still not available to the public. http://www.stsci.edu/hst/observatory/cdbs/calsp ec.html

6


Gaia flux calibration

101

Figure 10: The preliminary reductions (no telluric correction, only TNG (Telescopio Nazionale Galileo) observations, library extinction curve, and so on) of star GD 71 (top panel) are compared with the CALSPEC sp ectrum (b ottom panel). Prominent telluric features are marked in b oth panels. Except for the sp ectral edges ­ which will need to b e reconstructed with the use of models ­ the main b ody of the sp ectrum is always close to the CALSPEC one within 1% or b etter. c ESA by Bohlin & Gilliland (2004) and Bohlin et al. (2007). An example of the kind of sp ectra obtained for Pillar GD 71 (with DoLoRes at the TNG is shown on Figure 10. The Primary SPSS will constitute the ground-based calibrators of the actual Gaia grid, and need to conform to the following requirements (van Leeuwen 2010): · Primary SPSS have sp ectra as featureless as p ossible; · Primary SPSS shall b e validated against variability; · Primary SPSS have already well known SEDs; · The magnitude of each Primary SPS results in S/N 100 p er pixel over most of the wavelength range when observed from the ground with 2m class telescop es; · The location of Primary SPSS is in non crowded areas of the sky; · Primary SPSS cover a range of RA and Dec to ensure all-year-round ground based observations from b oth hemispheres. The Primary SPSS candidates set is describ ed in more detail in Altavilla et al. (2008), and some of the most imp ortant sources for Primary candidates are the CALSPEC grid, Oke (1990), Hamuy et al. (1992,1994), Stritzinger et al. (2005) and others. The actual Gaia SPSS grid, or Secondary SPSS, conforms to a different set of requirements (van Leeuwen 2010): · Secondary SPSS have sp ectra as featureless as p ossible (but see b elow for exceptions); · Secondary SPSS shall b e validated against variability;


102

Pancino

Figure 11: Left: example of a short-term variability curve for a constant SPSS candidate. Right: example of a short-term variability curve for a variable SPSS candidate. c ESA · The magnitude and sky location (i.e., numb er of useful, clean transits, see Carrasco et al. 2006, 2007) of Secondary SPSS grants a resulting S/N 100 p er sample over most of the wavelength range when observed by Gaia (end of mission); · Secondary SPSS cover a range of sp ectral typ es and sp ectral shap es, as needed to ensure the b est p ossible calibration of all kinds of ob jects observed by Gaia. As already mentioned, Secondary SPSS will b e mostly hot and featureless stars but will include a small numb er of selected sp ectral typ es, to ensure that the calibration model can work on all ob ject typ es. More details on Secondary SPSS can b e found in Altavilla et al. (2010), including a long list of literature catalogs and online databases from which the candidates are extracted. Clearly, all the Primary SPSS that, at the end of the data reductions, will satisfy also the criteria for Secondary SPSS, will b e included in the Gaia SPSS grid. Additionally, sp ecial memb ers of the Secondary SPSS candidates are: (1) a few selected SPSS around the Ecliptic Poles, two regions of the sky that will b e rep eatedly observed by Gaia, in the first two weeks after reaching its orbit in L2, for calibration purp oses; (2) a few M stars with deep absorption features in the red; (3) a few SDSS stars that have been observed in the SEGUE sample (Yanny et al., 2009), since the SEGUE sample has the p otential of b eing extremely useful in the Gaia flux calibration (Bellazzini et al. 2010), both internal (relative) and external (absolute); (4) a few well known SPSS that are among the targets of the ACCESS mission (Kaiser et al., 2010), dedicated to the absolute flux measurement of a few stars b esides Vega. 3.2. Observation strategy and campaigns

A basic consideration when starting the observations of such a large campaign, is that the traditional sp ectrophotometry techniques require too much observing time: each SPSS should b e sp ectroscopically observed in p erfect photometric conditions, ideally more than once. No TAC (Time Allocation Committee) would grant such a large amount of observing time to a prop osal that does not contain any cutting edge science in it. We therefore chose (Bellazzini et al. 2007) to split the problem into two parts: (1) sp ectra are taken in good sky conditions, but not necessarily p erfectly photometric; they


Gaia flux calibration

103

are calibrated with the help of a Pil lar or Primary SPSS thus recovering the correct sp ectral shape; (2) absolute photometry in the B, V, and R (sometimes I) Johnson-Cousins bands is taken in photometric sky conditions and used to fix the sp ectral zero-point by means of synthetic photometry. This is motivated by the fact that absolute photometric night p oints are faster to obtain than sp ectra. A subset of SPSS candidates will b e sp ectroscopically observed in photometric sky conditions, to check the whole procedure. Besides the Main Campaign just describ ed, it is necessary to monitor candidate SPSS for constancy (Auxiliary Campaign), since very few of them have systematically b een monitored in the literature, and there are illustrious examples of stars that showed unexp ected variability (Landolt & Uomoto, 2007). An example of a different kind of problem, that could greatly b enefit from good quality dedicated imaging, is star HZ 43. It was initially chosen by Bohlin et al. (1995) to b e one of the Pillars, and later rejected b ecause of an optical companion lying 3" away, only visible in the V band, that made it useless as an SPSS from the ground. Most of our SPSS candidates are WDs close to the instability strip, and, which, sometimes have p oorly known magnitudes, so it is necessary to monitor them for short-term variability on scales of 1­2 hours (Figure 11). Binary systems are frequent and can b e found at all sp ectral typ es, so we also monitor all our candidates for long-term variability (3 years) collecting approximately 4 night p oints p er year. These two monitoring campaigns rely on relative photometry (using stars in the field of view) to derive variability curves. An SPSS candidate is considered constant if it does not vary with an amplitude larger than a few milli-mags. The facilities that are b eing used for the two observing campaigns are (Federici et al., 2007, Altavilla et al. 2010): · EFOSC@NTT (New Technology Telescop e) La Silla, Chile (primarily Main campaign); · ROSS@REM (Rapid Eye Munt Telescop e) , La Silla, Chile (primarily Auxiliary campaign); · LARUCA@1.5m, San Pedro Martir, Mexico (primarily Auxiliary campaign and absolute photometry); · BFOSC@Cassini, Loiano, Italy (primarily Auxiliary campaign); · CAFOS@2.2m, Calar Alto, Spain (primarily Main campaign); · DoLoRes@TNG (Telescopio Nazionale Galileo), La Palma, Spain (primarily Main campaign); Observations started in the second half of 2006, comprising more than 35 accepted prop osals. We have b een awarded a total of 230 observing nights approximately, at the rate of 33 p er semester. More than 50% of this time resulted in at least partially useful data. Given the large numb er of facilities involved, and of different observers, it has b ecome necessary to establish strict observing protocols (Pancino et al. 2008,2009). The campaigns are now more than 50% complete, with the sp ectroscopy observations 75% complete, and we exp ect to complete all our observing campaigns around 2013. 3.3. Data reduction and analysis

The large amount of data collected needs to b e reduced and analyzed with the maximum p ossible precision and homogeneity. An initial set of data is collected for each CCD/instrument/telescop e combination and an Instrument Familiarization Plan (IFP) is


104

Pancino

Figure 12: Example of the 2nd order contamination in the DoLoRes grism LR-R, and of its correction for stars HZ 44 and G 146­76. The solid black curves are the corrected sp ectra while the red curves are the contaminating light coming from the 2nd order disp ersed blue light. c ESA

conducted, to derive shutter times, linearity, calibration frames and lamps stability, photometric distortions, 2nd order contamination of sp ectra (Figure 12), and so on. This plan is now almost complete, and the protocols are presently b eing finalized and written. The data reduction is regulated by strict data reduction protocols, that are presently being finalized. While the data reduction methods are fairly standard, care must b e taken in considering the characteristics of each instrument, as determined during the IFP, to extract the highest p ossible quality from each instrument. Semi-automatic quality check (QC) criteria are defined for each kind of observation (minimum & maximum S/N, seeing and roundness requirements of images, presence of bad columns, companions, and so on). Only frames that pass the QC are reduced. For imaging, we term "data reduction" the removal of the instrument characteristics (dark, bias, flat-field, fringing), QC, and the measurement of ap erture photometry with SExtractor (Bertin & Arnouts, 1996). The data products are 2D reduced images and ap erture magnitude catalogues. For sp ectroscopy, we term "data reduction" the removal of instrumental features (dark, bias, flat-field, illumination correction, wavelength calibration, 2nd order contamination correction, relative flux calibration, telluric features removal), followed by QC and sp ectra extraction. The data products are 2D reduced frames, 1D extracted and wavelength calibrated spectra, 1D flux calibrated sp ectra, 1D telluric absorption corrected sp ectra. The data reduction procedures are well advanced for photometry (almost half of the data reduced) and have just started for sp ectra (10% of the data reduced) at the moment of writing. The data analysis is presently in the design and testing phases. The study of shortterm variability curves is proceeding (10% of the data analyzed, see Figure 11). Absolute photometry and relative sp ectroscopy procedures are b eing refined: for example, preliminary end-to-end reduction of photometric imaging nights have b een p erformed for TNG and NTT observations, to allow us to identify those nights that were actually photometric and did not need to b e rep eated. Preliminary extinction curves have been determined for TNG and CAHA (Centro Astrono Hispano Aleman) (sp ectroscopic observations, allowing us to see ´ ´


Gaia flux calibration

105

Figure 13: The simple browsing interface of the Wiki-Bo local SPSS archive. This snapshot refers to the raw data archive, a similar web page exists for data products. c ESA that extinction varies in a grey manner (within a few p ercent) even in the case of some Calima (desert dust) in the sky over La Palma. The final data products for the Auxiliary campaign will b e relative magnitudes and lightcurves for all the monitored candidate SPSS; for the Main campaign, absolute magnitudes and errors will b e released together with their uncertainties and flux tables (Figure 10) in the form ((nm),F (photons s-1 m-2 nm-1 )). Possibly, also other intermediate data products will b e released (see ab ove). 3.4. Data availability

All the data SPSS ground based observations, along with the collected literature information and measurements, are stored in the CU5-DU13 local Wiki pages in Bologna (Wiki-Bo)7 . Wiki-Bo also contains all our technical documentation, internal rep orts, observation status and data products, along with literature references and sources, observing prop osals and the like. The raw and reduced data products are stored in a local archive8 for internal purp oses (Figure 13). In the future, when CU9 b egins (Catalogue production and access), it is foreseen that all the ground-based data used for the calibration of Gaia data (radial velocity standards, SPSS, sp ectral libraries, Ecliptic p ole observations, observations of Gaia itself from the ground, and so on) will b e published as well, although no decision on the format and typ e of data products has b een made yet. 4. Conclusions

The Gaia mission and its data reduction is a challenging enterprise, carried out by ESA and the Europ ean scientific community. As an example of the DPAC (Data Processing and Analysis Consortium) tasks, I have briefly summarized the problem of the external (absolute) flux calibration of sp ectrophotometric Gaia data, and more sp ecifically of the BP/RP low resolution sp ectra and the integrated G-band and BP/RP magnitudes. An innovative calibration model is presently under study and testing, and a large ( 200 - 300)
7 8

http://yoda.b o.astro.it/wiki, guest username and password can b e obtained from E. Pancino. http://spss.b o.astro.it, guest username and password can be obtained from E. Pancino.


106

Pancino

grid of SPSS with 1­3% flux calibration with resp ect to Vega is being built from multi-site ground-based observations. But once the Gaia data will b ecome available, a greater challenge will have to b e faced: the impact in almost all fields of astrophysics require that the scientific community (and not only the Europ ean one) b e adequately prepared to extract the most scientific output from the data. The training of a new generation of scientists, and the collection of complementary data, necessary to answer key questions when combined with Gaia data, should start now. The challenge requires that large groups of scientists get efficiently organized and ready to collab orate on large and comprehensive datasets.

References Altavilla, G., Bellazzini, M., Pancino, E., Bragaglia, A., Cacciari, C, Diolaiti, E., Federici, L., Montegriffo, P., Rossetti, E., 2008, "Primary standards for the establishment of the Gaia grid of SPSS. Selection criteria and a list of candidates", Gaia technical rep ort GAIA-C5-TN-OABO-GA-001 Altavilla, G., Bragaglia, Pancino, E., Bellazzini, M., A., Cacciari, C, Federici, L., Ragaini, S., 2010, "Secondary standards for the establishment of the Gaia grid of SPSS. Selection criteria and a list of candidates", Gaia technical rep ort GAIA-C5-TN-OABOGA-003 Bellazzini, M., Bragaglia, A., Federici, L., Diolaiti, E., Cacciari, C., Pancino, E., 2006, "Absolute calibration of Gaia photometric data. I. General considerations and requirements", Gaia technical rep ort GAIA-C5-TN-OABO-MBZ-001 Bellazzini, M., Altavilla, G., Cacciari, C., 2010, "Notes on the possible use of SEGUE spectrophotometry for the absolute photometric calibration of Gaia", Gaia technical rep ort GAIA-C5-TN-OABO-MBZ-002 Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 Bessell, M. S. 1999, PASP, 111, 1426 Bohlin, R. C., Colina, L., & Finley, D. S. 1995, AJ, 110, 1316 Bohlin, R. C. 2007, PASP, The Future of Photometric, Sp ectrophotometric and Polarimetric Standardization, 364, 315 Carrasco, J. M., Jordi, C., Figueras, F., Anglada-Escud´, G., 2006, "Towards a selection e of standard stars for absolute flux calibration. Signal-To-Noise ratios for BP/RP spectra and crowding due to FoV overlapping", Gaia technical rep ort GAIA-C5-TNUB-JMC-002 Carrasco, J. M., Jordi, C., Lop ez-Marti, B., Figueras, F., Anglada-Escud´, G., Amores, e E. B., 2007, "Revolving phase effect to FoV overlapping and its application to Primary SPSS", Gaia technical rep ort GAIA-C5-TN-UB-JMC-002 Carrasco, J. M. et al., 2010, in preparation, "Design of the experiment to test BP/RP ful l forwarding model", Gaia technical rep ort GAIA-C5-TN-UB-JMC-011 Hamuy, M., Suntzeff, N. B., Heathcote, S. R., Walker, A. R., Gigoux, P., & Phillips, M. M. 1994, PASP, 106, 566 Hamuy, M., Walker, A. R., Suntzeff, N. B., Gigoux, P., Heathcote, S. R., & Phillips, M. M. 1992, PASP, 104, 533 Kaiser, M. E., et al. 2010, arXiv:1001.3925 Landolt, A. U., & Uomoto, A. K. 2007, AJ, 133, 768 Montegriffo, P., & Bellazzini, M. 2009a, "A model for the absolute photometric calibration of Gaia BP and RP spectra. III. A ful l in-flight calibration of the model parameters", Gaia technical rep ort GAIA-C5-TN-OABO-PMN-003


Gaia flux calibration

107

Montegriffo, P., & Bellazzini, M. 2009b, "Quantitative estimate of the uncertainty on the wavelength calibration as derived from the absolute calibration process", Gaia technical rep ort GAIA-C5-TN-OABO-PMN-004 Montegriffo, P., et al., 2010, in preparation "Planning an experiment on source and instrument update XP processing", Gaia technical rep ort GAIA-C5-TN-OABO-PMN-005 Oke, J. B. 1990, AJ, 99, 1621 Perryman, M. A. C., Lindegren, L., & Turon, C. 1997, Hipparcos - Venice '97, 402, 743 Pancino, E., Altavilla, G., Bellazzini, M., Marinoni, S., Bragaglia, A., Federici, L., Cacciari, C., 2008, "Protocol for ground based observations of SPSS. I. Instrument familiarization tests", Gaia technical rep ort GAIA-C5-TN-OABO-EP-001 Pancino, E., Altavilla, G., Carrasco, J. M., M, Monguio, Marinoni, S., Rossetti, E., Bellazzini, M., Bragaglia, A., Federici, L., Schuster, W.., 2009, "Protocol for ground based observations of SPSS. II. Variability searches and absolute photometry campaigns", Gaia technical rep ort GAIA-C5-TN-OABO-EP-003 Ragaini, S., Bellazzini, M., Montegriffo, P., Cacciari, C., 2009a, "Absolute calibration of G and integrated BP and RP fluxes", Gaia technical rep ort GAIA-C5-TN-OABO-SR001 Ragaini, S., Montegriffo, P., Bellazzini, M., Cacciari, C., 2009b, "Absolute calibration of G and integrated BP and RP fluxes: test and limits of a simple model", Gaia technical rep ort GAIA-C5-TN-OABO-SR-002 Ragaini, S., et al., 2010, in preparation "Absolute calibration of G and integrated BP and RP fluxes: a new model", Gaia technical rep ort GAIA-C5-TN-OABO-SR-003 Stritzinger, M., Suntzeff, N. B., Hamuy, M., Challis, P., Demarco, R., Germany, L., & Soderb erg, A. M. 2005, PASP, 117, 810 Trager, S., 2010, "Spectrophotometric calibration of RVS using CU5-DU13 flux calibration tables", Gaia technical rep ort GAIA-C5-TN-UG-ST-002 van Leeuwen, F. & the CU5 DU managers, 2010, "CU5 software requirements specifications", Gaia technical rep ort GAIA-C5-SP-IOA-FVL-014 Yanny, B., et al. 2009, AJ, 137, 4377