Test area of the SAGE survey

Stellar atmospheric parameters can be obtained through well-defined photometric systems, e.g., the Strömgren-Crawford (SC) photometric system. But until now, there have been a few sky surveys or catalogs, e.g., the Geneva-Copenhagen Survey (GCS, Nordstöm 2004) and Hauck & Mermilliod Catalog (HM, Hauck & Mermilliod 1998), but they have shallow limiting magnitudes, which are only around V = 8. Another recent southern-sky survey, SkyMapper which is lead by Australian National University (ANU), is also based on their own photometric system. In the northern sky, we still lack deeper sky surveys (V ∼ 15), dedicated to deriving the stellar atmospheric parameters of a large sample.

Therefore, we are performing the deep Stellar Abundance and Galactic Evolution(SAGE) photometric survey (V = 15 with signal to noise ratio (S/N) of 100σ) of the northern sky. This paper introduces a test area chosen from the survey on which we merged single-epoch catalogs produced by the science data pipeline (SDP) into a master catalog, and analyzed the data. The procedures that are part of the SDP include image correction, astrometric calibration, photometry and flux calibration.

In order to obtain the stellar atmospheric parameters more accurately and more efficiently, we designed a new photometric system: the SAGE system, by combining our new self-designed filters with some existing photometric bands. This system consists of eight filters: Strömgren-u, SAGE-v, SDSS g, r, i, Hα_wide, Hα_narrow and DDO-51 (The brief names for the eight filters are: u_SC, v_SAGE, g, r, i, Hα_w, Hα_n and DDO51 respectively) The SAGE system will not only help to provide stellar atmospheric parameters, e.g., effective temperature, surface gravity and metal abundance, but also to study Galactic structure and evolution. Fan et al. (2018) introduces the SAGE photometric system, the survey strategy and the related scientific goals in detail.

The SAGE survey is a northern sky survey with the SAGE photometric system, with its 5σ depths expected to be ∼21.5mag in u_SC-band, ∼21.0,mag in v_SAGE-band, and ∼19.5,mag in g-, r- and i-bands. In the SAGE survey, the observations started in 2015 and we plan to finish the whole project, including observations, photometry, flux calibrations and astrometric calibrations in four or five years. A photometric catalog with uniform depth will be produced by the SAGE survey, which will help in scientific research on the Milky Way and even external galaxies.

The SAGE survey will cover an area of ∼12,000 deg² in the northern sky with Decl. δ > −5°, excluding the bright Galactic disk (|b|<10°), and the sky area of 12 h< R.A. <18 h. Details about coverage can be found in Zheng et al. (2018).

Three telescopes are used in the SAGE survey. The first is the 90-inch (2.3-meter) Bok telescope, operated by Steward Observatory of and administered by the University of Arizona (hereafter Bok), which is located at Kitt Peak National Observatory; the second is the Nanshan One-meter Wide-field Telescope at Nanshan Station of Xinjiang Astronomical Observatory, Chinese Academy of Sciences (NOWT); the third is the Zeiss-1000 Telescope at Maidanak Astronomical Observatory (MAO), Ulugh Beg Astronomical Institute, Uzbek Academy of Sciences (UBAI). Fan et al. (2018) and Zheng et al. (2018) introduced the telescopes and relevant observations that are part of the SAGE survey.

The SAGE survey started observations in the autumn of 2015. By the end of Jan. 2018, observations of g, r and i bands with NOWT were completed. Meanwhile, the observations of u_SC- and v_SAGE-bands on Bok are ∼ 2/3 completed and observations of the two bands are expected to be completed at the end of 2019. Observations with the MAO 1-m telescope are scheduled to start in the autumn of 2018. The up-to-date status of the observation progress can be checked on our website¹

In this paper, we discuss a test area that is used as a sample: NGC 6791, an open cluster. We performed a series of dithered observations on it with Bok in the u_SC- and v_SAGE-bands. The pointing and coverage of multiple exposures are shown in Figure 1. The blue box indicates the outline of this test area; the 16 black crosses are the centers of each exposure while the black dotted box is the coverage of one exposure corresponding to 1.08° × 1.03°. The red circle at the center is the approximate position and size of NGC 6791.

Zoom In Zoom Out Reset image size

The science images are processed through our semi-automatic SDP. We stacked the bias images and flat-fielding images for each night, and use them to correct the science images. Then we extract objects and calibrate their flux.

The Software for Calibrating AstroMetry and Photometry (SCAMP, Bertin 2006) is used for deriving the astrometric solution, then we compute the distortion solution which is expressed by the Simple Image Polynomial (SIP, Shupe et al. 2005). The astronomical reference catalog adopted in our work is the Position and Proper Motion Extended (PPMX, Röser et al. 2008) catalog and the more accurate version, the Catalog of Positions and Proper Motions on the ICRS (PPMXL, Roeser et al. 2010) will be adopted in the future work.

We apply the Source Extractor (SExtractor, Bertin & Arnouts 1996) to extract all the sources from images and evaluate their flux, i.e. photometry. We use the Kron-like elliptical aperture photometry (MAG_AUTO) and the corresponding uncertainties (MAGERR_AUTO) as our main output photometry. We also perform aperture-correction for photometry of all the detected sources.

For the flux calibration, we convolve the flux calibrated spectrum library Hubble Space Telescope CALSPEC Flux Standards (CALSPEC, Bohlin 2014) and the Next Generation Spectral Library (NGSL, Heap & Lindler 2010) with the filter transmissions of the SAGE system to derive the u_SC and v_SAGE magnitudes of standard stars. We have chosen 21 standard stars from the libraries with all the spectral types with suitable brightness for our observations. In a photometric night, we observe the standard stars in a large airmass range dozens of times to fit the atmospheric extinction curve and the extinction coefficients, which are then applied to calibrate the flux of secondary stars that we observed. The typical uncertainty of flux calibration is ∼0.01mag.

For the u_SC and v_SAGE bands, the flux calibrations are more complicated. For the previous version, we predicted the magnitudes from the AAVSO Photometric All-sky Survey, Data Release 9 (APASS DS9, Henden et al. 2015), which provides the g, r and i band photometry, by using a polynomial color-color relationship. The colors are derived with photometry from the MILES stellar library (FalcónBarroso et al. 2011). For the current version, we applied the more accurate catalog (PS1) from the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS, Chambers et al. 2016) to replace the APASS catalog.

A detailed description of the data reduction steps and results from single exposure images are discussed in Zheng et al. (2018). The current SDP produces a final catalog from the single exposure image, based on which we have done the following corrections and calibrations.

In order to test the SDP, we perform a detailed and thorough reduction process on a test area as shown in Figure 1. Besides the data reduction for single-epoch images, we also merge the catalogs into a master catalog. In this catalog, objects from different exposures with the same position (R.A. and Decl.) will be regarded as one object, and their photometric data will be merged.

5.1. Balance the Flux of Images

For a single-epoch image, the flux calibration is simple: just match the detected objects with the reference catalog and calculate the zero point to calibrate the flux of the single image.

However, for the test fields, there are 16 frames to be considered. For the first step, we need to do the internal calibration for all of the images. By matching the common stars in the overlapping part of any two images, we compute the zero-point offset and then neutralize to calibrate the two frames into the same flux level. The balancing operation includes the following steps. First, we identify all overlapped pairs of images in the test area, including fields overlapping by sides or corners, and multiple observations of the same fields. For the test observations of NGC 6791, any two exposures are overlapped. Then for each pair, we match their catalogs by coordinates and compute the offset and standard deviation of the calibrated magnitudes. Now we can start the neutralization. For each image, we adjust its zero point with the weighted mean of its relevant offsets. After adjustment of one image, we need to update the offsets related to it before adjusting other images. To prevent the transmission effect, we adjust images in a random order, without a fixed reference image. We perform multiple rounds of adjustment until the adjustments are low enough.

After the balancing, the offsets have been neutralized. As shown in Figure 2, the root mean square (RMS) of offsets reduces from 0.0525mag to 0.0055mag in the u_SC-band and from 0.0244mag to 0.0089mag in the v_SAGE-band.

Zoom In Zoom Out Reset image size

5.2. Columns of the Catalog

5.3. The Complete Magnitudes in the Two Bands

The test area is only ∼3 deg², which is only a tiny part compared to our survey. We show the magnitude distribution of all the detected sources in both the u_SC-band and v_SAGE-band from our master catalog in Figure 3.These inflexion points are ∼20.0mag in the u_SC-band and ∼21.0mag in the v_SAGE-band, which are in fact the complete magnitudes in the two bands.

Zoom In Zoom Out Reset image size

5.4. Color-Color Diagram of Stars

In order to check the zero points of the magnitude and colors in the flux calibration, we explore the distribution of point sources in our catalog as well as that from the stellar library in several dereddened color-color diagrams in Figure 4. We choose stars with S/N better than 0.02 and reliable photometry. As a comparison, we predict colors of stars from MILES by convolving the filter transmission curves with the stellar spectrum. Since we use PS1 as the reference catalog in flux calibration, we use g, r and i bands from the PS1 catalog. The colors from MILES are overplotted as black crosses while the calibrated stars are marked with green dots.

Zoom In Zoom Out Reset image size

Table 2. Columns in the Complete Catalog

Parameter	Description
OBS_DATE	Observation date, MJD
FILE_NUMBER	File number of the source
OBJ_NUMBER	Object running number from original catalog
AMP	Amplifier number where the source is located
X	Object position along x inside the amplifier
Y	Object position along y inside the amplifier
RA	Right ascension of the object center (J2000)
DEC	Declination of the object center (J2000)
FLAGS	Extraction flags by SExtractor
FINAL_ID	Id number in the final catalog

We are planning to complete the observation of u_SC- and v_SAGE-bands in 2018, and start observations at the MAO 1-m telescope from the autumn of 2018. We still need to test the entire control and optical system of the MAO 1-m telescope.

For the data reduction, we also need to improve our pipeline. We will do photometry and calibration with more precision, and we will stack overlapped images to detect deeper objects. Furthermore, we will analyze the relationship between stellar abundance and real observed colors, and try to find a batch of extremely metal-poor star candidates, which could be checked by follow-up observations with other telescopes, e.g., LAMOST.

We began the SAGE survey in 2015. Up to now, we have finished about 2/3 of the u_SC- and v_SAGE-bands, and all of the gri-bands from the observations. In this work we merge the catalogs from single-epoch images into a master catalog. We cross-match objects from different images by coordinates and then neutralize their offsets to do the internal calibration. After that, we perform flux calibration with the PS1 catalog. Furthermore, we discuss the depths and colors of the test area. We are trying to analyze the zero points of the magnitudes and colors, which are related to stellar properties, through these color-color diagrams. Tan et al. are working on the relationship between colors and stellar atmospheric parameters, while Wang et al. are trying to use deep learning methods to derive the stellar atmospheric parameters with the colors by a series of training data.

The SAGE photometric system is a self-designed photometric system with high sensitivity to stellar atmospheric parameters. We expect to obtain eight colors of about 500 million stars and derive their stellar atmosphere parameters. Meanwhile, we can also retrieve a reliable extinction map with Hα_w and Hα_nband photometry. The SAGE survey will become an important observational resource for stellar physics, and studies of the structure and evolution of the Milky Way. At the same time, this will also provide data for research on extragalactic objects.

Prof. R. Green, Prof. X. Fan and other astronomers at Steward Observatory of the University of Arizona, provided great help in observation and data reduction, and we thank them. We also received advice from the BATC group of NAOC. We appreciate the help from all of them. This work is supported by the National Natural Science Foundation of China (Grant Nos. 11373003, 11673030 and U1631102), the National Key Basic Research Program of China (2015CB857002), and the National Program on Key Research and Development Project (2016YFA0400804).