© ESO, 2012

1. Introduction

The VISTA variables in the Vía Láctea (VVV) survey is mapping 562 square degrees in the Galactic bulge and the southern disk in the near-infrared (Minniti et al. 2010). The VVV survey gives near-infrared (near-IR) multi-colour information in five passbands: Z (0.87 μm), Y (1.02 μm), J (1.25 μm), H (1.64 μm), and K_s (2.14 μm), as well as time coverage spanning over five years, which will complement past/recent, current, and upcoming surveys such as 2MASS, DENIS, GLIMPSE-II, VPHAS+, MACHO, OGLE, EROS, MOA, and GAIA.

VVV is an ESO Public Survey, i.e., the observational raw data are made available to the astronomical community immediately, whereas the reduced data will be published once a year in a data release through ESO. This paper describes and characterises the first data release (DR1) of the VVV survey. Some first results of the VVV survey based on early science images taken in the bulge and disk fields are highlighted in Saito et al. (2010) and Catelan et al. (2011).

The preparatory phase for the VVV survey started in 2006, with the first test observations obtained in October 2009. Regular operations started with the first survey observations in February 2010. The data collected during the first year of observations until October 2010 are the subject of this public release. Our survey is planned to be carried out for five years, and we expect to produce yearly accumulated data releases.

The VVV survey is foremost a variability study of the inner regions of the Milky Way (Minniti et al. 2010), but will also complement the existing 2MASS JHK photometry (Cutri et al. 2003), extending to much fainter limits while adding two additional filters (ZY), and providing time domain information useful for variability and proper motion studies. In particular, the higher resolution of the VVV data represents a huge advantage in crowded fields compared to previous near-IR surveys such as 2MASS and DENIS (Epchtein et al. 1994), where the single-epoch photometry was confusion-limited, reaching K_s ~ 14.3 mag. The limiting magnitude of the VVV data using aperture photometry is K_s ~ 18.0 mag in most fields. Even in the innermost fields (| b| ≤ 1°) the VVV survey reaches K_s ~ 16.5 mag, at least a magnitude deeper than the IRSF/SIRIUS survey of the Galactic Centre (Nagayama et al. 2003; Nishiyama et al. 2006, 2009).

The VVV survey was designed to complement the UKIDSS-GPS (Lucas et al. 2008), VPHAS+ (see Arnaboldi et al. 2007), and the GLIMPSE-II surveys (Benjamin et al. 2003). The UKIDSS-GPS is mapping |b| < 5° in Galactic latitude in the northern plane for three epochs, while VPHAS+ also observes the Galactic plane in the optical and Hα using the ESO VLT Survey Telescope (VST).

The GLIMPSE-II survey images the central  ±10° of the plane in four bands with IRAC. VVV provides variability information for the overlap region, supporting studies of the content and distribution of stars, stellar populations, and interactions of the strong nuclear wind with the ambient of the interstellar medium (ISM) above and below the nucleus, as well as the rate and location of current star formation.

Multiband Spitzer public surveys with IRAC (mid-IR at 3.6 μm, 4.5 μm, 5.6 μm and 8.0 μm) and MIPS (far-IR, 23.7 μm, 71.4 μm weighted average wavelength), respectively, cover the mid-plane (| b| < 1°) at 65° < l < 10° and −10° < l < −65°. The southern half of these surveys overlaps with the VVV disk area, allowing the detection and characterisation of star-formation regions and probing the structure of the inner disk of the Galaxy. In optically obscured regions the IRAC data complement the VVV survey and those of VST/VPHAS+ by tracing the influence of the most massive stars on star formation.

In addition, we complement the existing bulge microlensing surveys such as OGLE (Udalski et al. 1993; Szymański et al. 2011) and MACHO (Alcock et al. 2000), which observe in optical bands, with limited or no colour information. These surveys mostly concentrate on regions of low extinction. VVV will provide useful variability information for the overlap regions.

The data are of excellent quality in general, and only a small fraction did not pass our quality controls and had to be reacquired. These DR1 data have passed all the initial quality controls as performed by the survey team in collaboration with the Cambridge Astronomical Survey Unit (CASU)¹. We checked image defects, telescope problems, seeing, zero point, magnitude limit, ellipticity, and airmass, to name a few. However, it is important to stress that the data quality and calibrations will improve with subsequent data releases.

Here we will address the general information for the community, e.g., on the survey area and strategy, data quality, progress in the observations, published source lists, as well as examples of specific applications. In addition, the data, procedures, and additional information are available through the ESO archive², the VVV survey science team homepage³, CASU, and through the VISTA science archive (VSA) webpage⁴.

This paper is organized as follows: Sect. 2 describes the area coverage and the observations with VISTA as well as the data processing. Section 3 describes the photometric quality (limits, accuracy), and completeness. Section 4 discusses a comparison with 2MASS. Section 5 presents a comparison between the VVV DR1 (aperture) catalogues and point spread function (PSF) photometry. Section 6 describes the astrometric data quality. Section 7 presents density maps for the bulge and disk fields. Section 8 describes a suitability test of the DR1 data for difference image analysis.

The final section summarizes and presents our conclusions, including important caveats regarding this DR1, and future improvements to be implemented in DR2. Finally, the VVV tile coordinates are listed in the appendix.

2. Survey area and observations

2.1. Telescope and instrument

The telescope used to carry out the VVV survey is VISTA (Visible and Infrared Survey Telescope for Astronomy), a 4-m class wide-field telescope with a single instrument, VIRCAM (VISTA InfraRed CAMera; Dalton et al. 2006; Emerson & Sutherland 2010), located on its own peak at ESO’s Cerro Paranal Observatory in Chile, about 1500 m away from the VLT. Its primary mirror has a diameter of 4.1 m, providing a f /3.25 focal ratio at the Cassegrain focus where the instrument is mounted. The secondary mirror has a 1.24 m diameter. The start of survey operations of the telescope was on April 1, 2010, but most VISTA surveys, including VVV, started collecting science observations a few months earlier, in parallel with the last phases of the scientific performance verification and operations fine-tuning performed by Paranal Observatory staff. During the first year of operations the mirrors were coated with silver, which is optimized for near-IR observations.

With 1.64 deg diameter VIRCAM offers the largest unvignetted field of view in the near-IR regime on 4-m class telescopes. It is equipped with 16 Raytheon VIRGO 2048 × 2048 pixels² HgCdTe science detectors, with average pixel scale. Each individual detector therefore covers  ~694 × 694 arcsec² on the sky. The achieved image quality (including seeing) is better than  ~ on axis. The image quality distortions are up to about 10% across the wide field of view. The detectors are arranged in a 4 × 4 array, with large spacings of 90% and 42.5% of the detector size along the X and Y axes, respectively. A single pointing, called a pawprint, covers 0.59 sq. deg, and provides partial coverage of the field of view. By combining six pawprint exposures with appropriate offsets, a contiguous coverage of a field is achieved with at least two exposures per pixel except at two edges. In all VISTA observations this field is called a tile and covers a 1.64 sq. deg field of view. Throughout this paper we will use a tile as the individual exposure.

VIRCAM has four additional optical CCDs, two for guiding and two for active optics. For exposures longer than  ~40 s the active optics is run in parallel mode with the observations. Owing VVV’s very short individual exposures (see Sect. 2.3) the active optics correction is only performed every  ~30 min or after a larger offset. This, combined with the need to survey the area fast (hence minimizing the overheads for more frequent active optics corrections), and typical seeing on Paranal limits the image quality obtained for the survey data to typically  ~0.9–1.0 arcsec (see Sect. 2.4).

VIRCAM is equipped with five broad-band filters (Z, Y, J, H, and K_s) and two narrow-band filters centred at 0.98 and 1.18 μm. The VVV survey uses all five broad-band filters spanning from 0.84 to 2.5 μm. Their effective wavelengths and relative extinctions are given in Table 1, while the transmission curves are shown in Fig. 1, compared to a typical atmospheric transmission profile for airmass 1.0 and 1 mm water vapour in the atmosphere.

For more details about the telescope and instrument we refer the interested reader to the VIRCAM instrument web pages⁵, and the VISTA/VIRCAM user manual (Ivanov & Szeifert 2009).

2.2. Survey area

The VVV survey area consists of 348 tiles, 196 tiles in the bulge and 152 in the disk area. These two components were planned to cover 520 sq. deg, as follows: (i) the VVV bulge survey area covers 300 sq. deg between −10° ≤ l ≤  +10° and −10° ≤ b ≤  +5°; and (ii) the VVV disk survey area covers 220 sq. deg between 295° ≤ l ≤ 350° and −2° ≤ b ≤  +2°. However, to maximize the efficiency of the tilling process (see Sect. 2.3), the survey area definition tool (SADT; Hilker et al. 2011) produced some shifts at the edges of the survey area, and as the result an area of  ~562 sq. deg (42 sq. deg larger) was observed. Thus, the observed area is within −10.0° ≲ l ≲ +10.4° and within −10.3° ≲ b ≲ +5.1° in the bulge, and 294.7° ≲ l ≲ 350.0° and −2.25° ≲ b ≲ +2.25° in the disk. The VVV survey area and tile numbering are shown in Fig. 2, while the list of all tile centres in equatorial and Galactic coordinates is given in Table A.1. The tile names start with “b” for bulge and “d” for disk tiles, followed by the numbering shown in Fig. 2.

While the whole area was observed in the JHK_s, 95% of these tiles satisfy the stringent photometric and image quality parameters and are classified as completed. In the ZY bands the completion is somewhat lower, with 59% completed tiles. The tiles completed in the first season (until October 26, 2010) can be seen in Figs. 3–5 for ZY, JH, and K_s, respectively. Figure 5 also includes the tiles with at least one epoch observed during the variability campaign in the K_s band.

In the first season of the variability campaign 22 tiles in the disk area had five K_s epochs taken, while the majority had one or more additional K_s epochs completed. The completion rates individually for the bulge and disk areas are given in Table 2. Table A.1 states for each tile whether the observations in a given filter are completed, and how many additional K_s epochs were obtained until October 26, 2010.

Figure 6 shows cumulative distributions of image quality and airmass for the observed tiles in the 2010 season. The median image quality in the J, H and K_s filters is better than 09 as measured on combined tile images, while it is close to 10 for the Z and Y filters.

2.3. Observing strategy

The first observations collected for the VVV survey were taken during the VISTA science verification period in October 2009 when one field in the Galactic bulge at α = 18:02:58.872, δ = −28:36:59.04 (J2000) was observed in the ZYJHK_s filters (called SV field). In addition to the nearly-simultaneous multi-band images of the SV field, 11 K_s band exposures were taken to test the variability observing strategy. To establish the necessary number of exposures for proper sky subtraction, depending on crowding and number of resolved objects in the field, three additional tiles were observed in the K_s band, two bordering directly on the SV field and one at an offset position  ≈ 1° south of the science target. Based on these early observations, we adjusted the observing strategy, which differs slightly between the bulge and disk fields (i.e., with respect to exposure time, number of co-added exposures, and combination of tiles for sky subtraction).

Like all VISTA observations, the VVV survey is carried out in service mode. The basic observational unit is the so-called OB (observation block). The multi-filter, single-epoch OBs have been split into JHK_s and ZY OBs. The variability monitoring OBs have only single filter: K_s. Observation blocks for two, three or four tiles were executed back-to-back to ensure that a sufficient number of offset images were taken for each filter to create a high-quality background sky frame.

The survey area was defined with the help of the SADT, which provides the tile centres as well as guide and active optics stars that are necessary for the efficient execution of the survey OBs. Apart from the edges of the survey area, the input to SADT is also the tile pattern, which defines the large offsets that fill in the inter-detector gaps, and the size of the smaller (jitter) offsets that are executed at each of the six pawprint positions that together make a tile. All VVV OBs used the “Tile6n” pattern. In addition, at each of the six pawprint positions two smaller offsets are executed using the “Jitter2u” pattern. This means that in total there are 12 exposures per filter, but given the large offsets, each pixel is covered by at least four exposures, except for the pixels along the y-edge of the tile (two jitter positions and at least two pawprints). These, however, have overlaps with adjacent tiles. Therefore the complete survey area is covered by at least four exposures in each filter. Each image is a co-addition of NDIT (number of detector integration time) exposures lasting DIT (detector integration time) seconds each. The total exposure times for bulge and disk tiles are given in Table 3.

2.4. Data processing

The VVV observations were pipeline-processed within the VISTA data flow system (VDFS) pipeline at the CASU (Lewis et al. 2010). The processing was performed on a night-by-night basis, and consisted of the following data reduction steps executed in the order described.

The mean dark current exposure, taken with the same DIT and NDIT values, is subtracted from each image.

A linearity correction is applied for individual detectors using information on the readout time, exposure time, and the reset image time. A “reset” exposure of 1.0 s for every exposure is subtracted from each exposure within the data acquisition system, prior to writing the image to the disk.

The flat-field correction is made by dividing by a mean twilight flat-field image to remove small-scale quantum efficiency variations and the large-scale vignetting profile of the camera, as well as to normalize the gain of each detector to a common median value.

The sky background correction removes the large-scale spatial background emission. Tests made with the science verification observations showed that 12 exposures (six pawprint  ×  two jitters) taken for each VVV tile do not yield good sky subtraction because of severely crowded fields, leaving “holes” at the positions of bright stars or very crowded regions with many overlapping stellar PSFs. Therefore a sky background map is produced for VVV by combining all exposures for a given filter taken within several concatenated OBs (tiles). Owing to variability of the near-IR sky, the need to take at least 24 images with the same filter within  ~30 min poses limits to the exposure times for individual tiles and the number of different filters that can be included in each OB.

A “destriping correction” is performed by subtracting the low-level horizontal stripe pattern introduced by the readout electronics of the VIRCAM detectors.

Jitter stacking is performed to align two slightly shifted images taken at a given pawprint position, combining them into a single image for each pawprint. The shifts are computed using the positions of many hundreds of stars detected in all images.

Object detection is performed for each stacked pawprint image. Positions, fluxes measured in several apertures of different sizes, and some shape measurements are written in the source catalogue. A flag indicates the most probable morphological classification, and in particular we note that “−1” is used to denote stellar objects, “−2” borderline stellar, “0” is noise, and “+1’ is used for non-stellar objects. There are also objects with flag “−7”, denoting sources containing bad pixels, and the flag “−9” is used for saturated stars. These flags are derived mainly based on curve-of-growth analysis of the flux (Irwin et al. 2004). Figure 7 shows colour-magnitude diagrams (CMDs) for a moderately crowded bulge field (b264), comparing the distribution of the high-quality sources with all other flags. Relative numbers for each flag are also indicated in the figure. In addition to a catalogue with extracted sources, a confidence map is also computed (see Fig. 8).

The six stacked pawprint images are combined into a single deep tile and the catalogue extraction step is repeated. Tile images contiguously sample about 1.5 × 1.1 sq. deg on the sky. Confidence maps are computed for each tile, relaying information on the different exposure times for pixels across the tile (Fig. 8). The exposure times depend on the size of the jitter offsets and the number of exposures that are combined into a tile. The confidence maps clearly show the areas of detectors affected by bad pixels, such as the large patch in detector 1 (lower left corner in the right panel of Fig. 8), several rows in detector 4 (lower right corner) and the upper third of detector 16 (upper right corner). After stacking several dithered exposures in the final tile, most of these bad pixels are not noticeable (left panel of Fig. 8), but the large bad area on detector 16 has larger errors in the illumination correction map. The sensitivity of that upper third of the detector is much worse at shorter wavelengths, and some offsets with respect to 2MASS calibration for the whole tile have been found in Z, Y and J bands. Offsets at longer wavelengths in the H and K_s bands are within the calibration errors.

The first-year data release of the tile images, catalogues, and confidence maps described here is version 1.1. Detailed information about this version of VISTA data products as well as a brief description of all problems encountered during data processing is available on the CASU web page⁶.

2.4.1. Generation of multi-band catalogues

Currently the version 1.1 single-band tile catalogues from CASU are matched by the VVV team members using the STILTS package (Taylor 2006) and a KD-Tree based algorithm (Gurovich et al., in prep.) that uses the Cross et al. (in prep.) implemented source matching method for the VSA data. They are matched using astrometrically corrected tiles and catalogues, allowing a 10 offset between point sources to be considered a match. Nevertheless, tests have shown that over 90% of the stellar sources are matched within less than , as expected from the astrometric accuracy of the catalogues (see Sect. 6).

After matching the single-band catalogues, most of the spurious detections around bright stars are rejected. Unfortunately, the ellipticity sometimes varies from image to image; which then results in rejection of some sources which are only classified as “bona fide” stars in selected filters, but appear slightly more elongated, and are therefore rejected as non-stellar in other filters. A final multi-band catalogue contains  ~75–85% of the sources found in the catalogue with the least number of sources used for the matching, in general the K_s band, for low- to intermediate-extinction regions. Close to the Galactic plane, high extinction affects the source detection at short wavelengths more severely, hence the source density is limited by the detection rate in the Y and Z bands. The comparison between the different tiles is hampered not only by the different median extinction values, but also by the strong extinction variation within any given tile. As an example, we divided two tiles (b305 and d003) into 0.25°  ×  0.25° sub-sections. Median extinction values A_V and their standard deviations are given in Table 4.

Multi-band catalogues corresponding to the DR1 data have also been generated by the VISTA Science Archive (VSA). While the single-band DR1 data were already delivered through the ESO archive, the multi-band catalogues will be publicly available in due time through the ESO archive and VSA web page⁷. The VSA will also provide multi-epoch catalogues for variable sources. Part of the data reduction and storage are performed using the Geryon cluster at the Center for Astro-Engineering at Universidad Católica (AIUC)⁸.

3. Photometry

Photometric calibration of stacked pawprint images and tile images was performed using numerous detected 2MASS stars. The calibration procedure closely follows that of WFCAM (Hodgkin et al. 2009). Internal photometric accuracy is of the order of  ±2%, and for the J, H and K_s bands a similar accuracy as for 2MASS is achieved for most of the survey area. In particularly high-extinction regions and for the Z and Y filters the photometric calibration errors are somewhat larger.

Figure 9 shows the photometric errors as a function of the magnitude computed by the CASU pipeline for the five passbands in a typical disk field (d003) and in an extreme case of a crowded bulge field close to the Galactic Centre (b305). Different extinction (E(B − V)_d003 = 1.77, E(B − V)_b305 = 1.37 mag; Schlegel et al. 1998) and crowding levels as well as distinct observing strategies between disk and bulge areas (see Sect. 2.3) contribute to the shallower curves seen for the bulge data in all five passbands. The accuracy in the photometry can also be checked using the overlapping regions between the tiles. Figure 10 shows in the top panels the K_s photometry for stellar sources (also computed by the VDFS pipeline at CASU) in the overlapping area between tiles b305 and b291, and between d003 and d041. The 5σ magnitude limits reached for each passband for both bulge and disk tiles as computed by the pipeline are shown in Figs. 3–5.

3.1. Saturation

VIRCAM detectors saturate at different levels, mostly around 33   000−35   000 ADU. Detector #5 has the shallowest well depth (saturation at  ~24 000 ADU), but detector #13 has the highest non-linearity with of  ~10% at 10 000 ADU. This, combined with the fairly bright near-IR sky, results in a restricted dynamic range in the photometry.

Prior to reaching saturation, the VISTA detectors have a nonlinear regime. Although linearity correction is performed in the pipeline data reduction based on observations of an illuminated dome screen, for the stars close to the saturation limit there are still quite significant deviations with respect to 2MASS because of the residual non-linearity. This can be seen clearly in Fig. 2 of Gonzalez et al. (2011a) and also in the VMC survey paper by Cioni et al. (2011).

The overlapping regions also help us to check the linearity of the photometry at the saturation limit. In the top panels of Fig. 10 we see that for both bulge and disk data, several of the brightest stars slightly scatter from the linear distribution close to the saturation level, even taking into account that these are “−1” (stellar) sources in all VDFS single band catalogues from CASU. Some saturated stars are also present in the CASU catalogues (flagged as “−9”, see Fig. 7), but these are not present in Fig. 10.

3.2. Photometric completeness

Although the main goal of this paper is to describe the content of the first data release for the VVV survey, we now demonstrate briefly how strongly the completeness of the source catalogues provided by CASU depends on the location of the selected area within the survey.

To do this, we carried out artificial star experiments, comparing the detection rate for artificial stars (AS) added to the images using the CASU source detection package imcore (Irwin et al. 2004). We selected two different tiles from the bulge area, namely b204 and b314, which represent different levels of crowding. From these tiles we cut a small 2000  ×  2000 pixel² area () for the completeness tests. The two fields were centred on α₂₀₀₀ = 18:12:13.173, δ₂₀₀₀ = −38:07:49.220 for b204 (l,b = 354.72°,−9.37°), and α₂₀₀₀ = 17:29:28.920, δ₂₀₀₀ = −36:00:09.800 for b314 (l,b = 352.21°,−0.92°), respectively. The field b204 represents a less crowded section of the survey area, whereas b314, close to the Galactic Centre, is one of the most crowded fields. The K_s band source catalogues for the complete tile and the respective sub-sections contain 482,004/11,584 (b204) and 1,137,615/23,691 (b314) sources. We point out that in a preliminary test we applied the source detection package to the complete b204 tile and confirmed the number of sources obtained by the CASU pipeline, which validates the source detection in the AS experiments.

In our completeness tests we added 5000 AS to the original images for 16 individual magnitude intervals, each 0.2 mag wide. The positions of these stars were chosen randomly, but the same positions and K_s magnitudes were used for both fields. To create the AS, we derived a single point spread function (PSF), using isolated stars within the corresponding image. We leave the completeness test, which was conducted with a variable PSF for later. The AS images were constructed using the IRAF task addstar⁹.

Using imcore, we created source catalogues for each of the AS images, which we call “output” and compared them to the original source list “input”. The ratio of recovered AS to the original 5000 stars for a magnitude range between 12.0 ≤ K_s ≤ 18.2 mag is shown in Fig. 11. In the first test (solid lines) we considered an AS detection to be positive if a source was found within 1 pixel of the inserted position. However, in addition to the simple detection we also need to know with which accuracy the “input” magnitude is recovered. Based on the AS experiments of the “ACS globular cluster survey” (Sarajedini et al. 2007; Anderson et al. 2008) we first allowed a 0.75 mag offset between input and output magnitude, which we finally reduced to 0.5 mag. This is based on the assumption that the stellar densities in both our fields are significantly lower than that within the central region of a globular cluster that is targeted by the ACS globular cluster survey. The completeness curve for those tests is shown in Fig. 11 by dashed lines.

Based on the different completeness pattern we find that the source detection efficiency reaches 50% for stars with 17.8 ≲ K_s ≲ 18.1 mag in field b204, and 16.4 ≲ K_s ≲ 16.9 mag in field b314, respectively, depending on the restrictions applied. We also note that the completeness is somewhat lower for the brightest stars (12.0 to 12.2 mag) in both fields. A possible explanation may be that those stars are close to the saturation limit and the photometric accuracy is therefore reduced.

4. Comparison with 2MASS photometry

The VVV observations are approximately four magnitudes deeper than 2MASS. In addition, a very important factor is the excellent image quality (with seeing –) in the entire VVV survey area for the multi-band, single epoch observations. This allows us to reach the red clump magnitude across the entire bulge (Gonzalez et al. 2011a; Saito et al. 2011), and therefore to study the stellar populations and the structure of the inner Galaxy to an unprecedented level of detail, such as the identification of RR Lyrae and derivation of accurate distances. Figure 12 shows a CMD for 10′ × 10′ regions in the inner bulge (b305) and the outer bulge (b235) of VVV stellar sources compared to 2MASS CMDs of the same regions. Because the VVV photometry is much deeper, it allows us not only to trace the red clump even in the most extincted regions, but also to study the stellar populations behind the bulge. A detailed study of the colour transformations between VISTA and 2MASS photometric systems for VVV disk fields will be presented by Soto et al. (in prep.).

5. VVV DR1 catalogues vs. PSF photometry

The photometric catalogues published in this, as well as any following VVV data release, are based on aperture photometry only, computed by the CASU pipeline on individual tiles (see Sect. 2.4). However, for very crowded fields (e.g., the innermost Galactic Centre or the central region of stellar clusters), more complete and deeper photometry can be obtained with PSF photometry. In this section we present tests performed by VVV team members using PSF photometry on the VVV images. The results show that the PSF fitting can reach up to 1.5 mag deeper than aperture photometry, detecting up to twice more sources for highly crowded fields, where aperture photometry is known to be inefficient, particularly for faint sources. We emphasize that the PSF data are not part of the VVV DR1.

The tests were performed using the apermag3 aperture fluxes, which are used in the CASU catalogues as well as in the VSA database as the default values to represent the flux for all images. However, the CASU pipeline measures positions and fluxes for different concentric apertures designed to adequately sample the curve-of-growth of the majority of images. The apermag1 has 1 diameter and each successive aperture increases by a factor of in diameter. For highly crowded fields the aperture photometry using the apermag3 can simultaneously fit multiple overlapping sources, but it does not perform any subtraction in order to check for fainter sources hidden underneath. In this case the apermag1 and apermag2 can be more suitable than apermag3, even taking into account that narrow apertures have more uncertain aperture corrections, especially in poor seeing conditions.

Figure 13 shows in the top-left panel the K_s vs. (J − K_s) CMD for a field with relatively low crowding and extinction (a section of tile b204), comparing aperture photometry performed by CASU with PSF photometry obtained with DoPHOT (Schechter et al. 1993) for the same field, using 5σ magnitude limits in both cases. All structures are correctly seen in both cases, and both aperture and PSF photometries reach K_s ~ 18.0 mag.

The bottom-left panel of Fig. 13 shows the magnitude distribution of all sources found in the CASU catalogue (dashed line) as well as that of stellar sources only (solid line). Even taking into account that aperture photometry contains some brighter stars that are not present in the PSF photometry, the total number of sources found by PSF photometry in the K_s-band image of b204 is 589   187, in comparison to 482   004 (all sources) and 273   550 (stellar flag) sources present in the CASU catalogue.

For comparison, the same analysis is performed in the right panel of Fig. 13 for the inner bulge field b305, where the effects of crowding and extinction are significant. The PSF photometry allows us not only to resolve the high-density areas, but it also reaches  ~1.5 mag deeper, i.e., K_s ~ 17.5 mag, while the aperture photometry is limited to K_s ~ 16.0 mag. The bottom-right panel of Fig. 13 shows by how much the PSF photometry exceeds the aperture photometry for faint sources. The total number of sources found by PSF photometry for field b305 in the K_s-band is 4   601   529 in comparison to 1   201   557, (all flags) and 818   706 (stellar only) sources present in the CASU catalogue.

Tests performed by VVV team members show that the PSF photometry can be even deeper using a DAOPHOT-ALLFRAME suite of routines (Stetson 1994) customized for the VVV data (Mauro et al., in prep.).

6. Astrometry

The native VISTA WCS distortion model for pawprints is based on zenith-pole-north (ZPN) projection and is available in image headers. The distortions are radial and are well described by (1)with k₁ = 0.3413 arcsec/pix being the plate scale at the centre, and k₃/k₁ = 44, k₅/k₁ = 10   300 are distortion coefficients in angular units of radians. Higher order terms are negligible.

The median WCS rms is  ~70 mas and is dominated by the 2MASS astrometric errors. We note that in the coadded pawprints as well as in the complete tiles the pixels were resampled to a common spatial scale and the astrometric distortions were removed according to Eq. (1). The resampling was also required in the pawprints because the jitter offsets are sufficiently large (i.e., 150) to affect the coadding (see also Table 3) owing to the size of the distortions. Another way to evaluate the internal astrometric accuracy is to use overlaps between different tiles. Hundreds of stars are detected independently in two adjacent tiles. Figure 10 shows in the top panels the photometry in K_s for overlapping regions between tiles in the inner bulge (b291 and b305) and disk (d003 and d041) areas. Different observing strategies, in particular exposure times (see Table 3), for the disk and bulge areas, as well as the high background brightness owing to unresolved stars in the inner bulge, make the bulge photometry shallower than that in the disk region. The astrometric accuracy is shown in the bottom panels of Fig. 10 in terms of the distribution Δδ vs. Δα × cosδ. Typical values for the astrometric accuracy are  ~25 mas for a K_s = 15.0 mag source and  ~175 mas for K_s = 18.0 mag. Despite the shorter time baseline of the VVV survey, the typical proper motion measurements should reach the accuracy between  ~7 mas yr^-1 (K_s = 15.0 mag) to  ~15 mas yr^-1 (K_s = 18.0 mag) after the five years of the VVV campaign.

A test for these measurements can already be achieved using 2MASS and VVV datasets, which provide a time baseline of 11 years. Cross-matching of sources between VVV and 2MASS was performed for a 10 × 10 arcmin² field in tile b249. The astrometric differences between stellar sources in these two catalogues can be used to derive proper motions of stars in terms of μ_l and μ_b. Clarkson et al. (2008), based on two epochs of HST imaging, showed that foreground disk stars observed in the colour-magnitude diagrams towards the bulge can be separated based on Gaussian fits to the distribution of their proper motions with mean differences of (Δμ_l,Δμ_b) = (3.22 ± 0.15,0.81 ± 0.13) mas yr^-1.

Figure 14 shows the selection of stars in the CMD for matched stars between VVV and 2MASS that belong to disk and bulge populations together with the distribution of μ_l and μ_b for each selection. Although the distributions are wider than those presented in the HST analysis of Clarkson et al. (2008), the mean differences between proper motions of disk and bulge are still evident based on these data showing differences of (Δμ_l,Δμ_b) = (3.39 ± 0.25,1.25 ± 0.24) mas yr^-1, in good agreement with those of Clarkson et al. (2008).

7. VVV Source maps

Figures 15 and 16 show all objects with stellar flag detected in each tile in the DR1 data. Only stars matched in the J, H and K_s filters have been plotted. The total number of sources found in the bulge region is 7.06 ×  10⁷, while the disk has 9.29 ×  10⁷ sources. The stellar density (corrected for the total field size) is higher in the disk because of the deeper photometry in all five filters compared to the bulge observations (see Table 3, and Figs. 3–5).

Note that these density maps already provide a wealth of information on the extinction and the structure of the inner Galaxy, which will be investigated in detail in subsequent papers (e.g., Gonzalez et al. 2011a,b; Saito et al., in prep.).

The regular grid pattern seen in both the bulge and disk areas is due to the overlap of adjacent tiles. The individual catalogues used in Figs. 15 and 16 contain  ~10% of the total number of sources twice, contributed by two independent tiles, and leading to the much higher source density shown in the two figures. The detection of a significant percentage of sources in two independent tiles not only helps to test and/or confirm the photometric and astrometric calibration, but also to carry out quality control. These density maps are also used to identify problematic tiles, i.e., incomplete data readouts, missing tiles, and observations made under strongly varying seeing conditions (e.g., tiles b216 and b343).

Moreover, those overlap regions will also be beneficial for one of the main scientific goals of the VVV survey. Because the different tiles are observed independently (with the exception of concatenated tiles), the variable stars included on those tiles will obtain twice as many K_s epochs, and hence much better sampled light curves (see also Catelan et al. 2011).

8. Difference image testing

Whilst DR1 does not include photometry based on difference image analysis (DIA), we have tested the suitability of the data for DIA analysis based on a modified double-pass version of the ISIS package Alard & Lupton (1998) originally developed for the Angstrom M31 microlensing survey (Kerins et al. 2010).

Currently, the VVV science verification (SV) field comprises the largest number of epochs and therefore most DIA testing has been performed on this dataset. One problem with difference imaging of VVV data is that often the seeing is so good that the point spread function (PSF) is poorly sampled by the pixel size of around . For good DIA kernel convolution we usually require more than 2.5 pixels/FWHM. To minimize undersampling, instead of convolving a good seeing reference image to a poorer seeing target, we selected a poor seeing reference frame and convolved the target image to it. The DIA image is therefore constructed by minimizing (2)where the sum is over the pixel coordinates (i,j), R and T are the reference and target images, B describes the differential background between R and T, K is the convolution kernel and D is the resulting difference image (see Alard & Lupton 1998, for details of how K is constructed). Before difference imaging we added the sky images back onto the sky-subtracted image stacks.

Figure 17 shows a K_S band pawprint from one epoch of the VVV SV bulge field. The lower row shows the resulting difference images. Whilst the difference image quality is reasonably clean, there are many black residuals occurring around the many saturated stars in the field. However, variable objects are as easy to pick out as in the circled example in the zoomed DIA panel in Fig. 17.

The background noise level of the DIA images is consistent across all arrays and performs well with respect to imcore sky noise estimates (Irwin et al. 2004). In Fig. 18 we plot the pixel histograms of DIA flux for the same SV pawprint shown in Fig. 17 for each of the 16 arrays. The fluxes are normalised to a level that is just 40% of the sky noise estimate reported by imcore. In all cases the histograms show that difference imaging noise is significantly below that predicted by imcore, as shown by the close resemblance of the normalised DIA flux distributions to a unit Gaussian curve.

The photometric performance of VVV DIA will be calibrated in greater detail in a future paper.

9. Summary

We have presented the VVV Data Release 1, describing the design, observations, data processing, data quality, and limitations. The data are of very high quality, and represent a vast improvement over existing near-IR photometry. Therefore they are useful for a wide variety of studies such as open/embedded clusters (Borissova et al. 2011; Baume et al. 2011), globular clusters (Minniti et al. 2011a; Moni Bidin et al. 2011), distance scale (Majaess et al. 2011), YSO censuses (Faimali et al., in prep.), brown dwarfs (Folkes et al., in prep.), proper motions (Minniti et al., in prep.), disk stellar populations and variable stars (Pietrukowicz et al. 2012), Galactic structure of the disk mapping the edge of the stellar disk (Minniti et al. 2011b), and the inner structure of the bulge (Gonzalez et al. 2011b), bulge stellar populations including metallicity, extinction and dust maps (Gonzalez et al. 2011a; Catelan et al., in prep.)¹⁰ high-energy sources (Greiss et al. 2011a,b; Masetti et al., in prep.), background galaxies (Amôres et al., in prep.), colour transformations between VISTA and 2MASS systems (Soto et al., in prep.), as well as enabling and complementing other Galactic structure and stellar population studies, variable star studies in clusters and in the field (pulsating variables, eclipsing binaries, and planetary transits), gravitational microlensing studies, Galactic Centre studies, ultra-high-velocity star searches, PNe searches, SN light echo searches, QSO searches, searches for faint solar system objects (e.g., NEOs, MBAs, LJ5s, TNOs), etc.

---

Acknowledgments

We gratefully acknowledge use of data from the ESO Public Survey programme ID 179.B-2002 taken with the VISTA telescope, data products from the Cambridge Astronomical Survey Unit, and funding from the FONDAP Center for Astrophysics 15010003, the BASAL CATA Center for Astrophysics and Associated Technologies PFB-06, the FONDECYT from CONICYT, and the Ministry for the Economy, Development, and Tourism’s Programa Iniciativa Científica Milenio through grant P07-021-F, awarded to The Milky Way Millennium Nucleus. R.K.S. and D.M. acknowledge financial support from CONICYT through Gemini Project No. 32080016. D.M. acknowledge support by Proyecto FONDECYT Regular No. 1090213. M.Z. and O.A.G. acknowledge support by Proyecto FONDECYT Regular No. 1110393. J.B. and F.P. are supported by FONDECYT Regular No. 1080086. J.R.A.C. is supported by CONICYT through Gemini Project No.Ãľ32090002. A.N.C. received support from Comitee Mixto ESO-Gobierno de Chile. M.S. acknowledges support by Proyecto FONDECYT Regular No. 3110188 and Comite Mixto ESO-Gobierno de Chile. Support for R.A. is provided by Proyecto FONDECYT Regular No. 3100029. R.d.G. acknowledges partial research support through grant 11073001 from the National Natural Science Foundation of China. M.M.H. is supported for this work by the US National Science Foundation under Grant Nos. 0607497 and 1009550. G.P. acknowledge support from the Millennium Center for Supernova Science through grant P06-045-F funded by Programa Bicentenario de Ciencia y Tecnologa de CONICYT. M.C., J.A.G. and I.D. acknowledge support by Proyecto FONDECYT Regular No. 1110326. M.C. is also supported in part by Proyecto Anillo ACT-86. S.L.F. acknowledges funding support from the ESO-Government of Chile Mixed Committee 2009 and from the Gemini-CONICYT grant No. 32090014/2009. B.B., B.D. and L.S.J. acknowledge support from FAPESP and CNPq. E.B.A. thanks Fundação para a Ciência e Tecnologia (FCT) under the grant SFRH/BPD/42239/2007. R.K. acknowledges support from the Centro de Astrofisica de Valparaiso and Proyecto DIUV23/2009.

References

Appendix A: VVV tile coordinates

Here we list the tile centre coordinates for all VVV pointings. There are 196 bulge tiles, with names starting with “b”, and 152 tiles in the disk area, whose names start with “d”. For each tile we provide tile centre coordinates in Equatorial and Galactic coordinates. All tiles have been observed using an identical offsetting strategy, combining six pawprints to contiguously fill 1.5 × 1.1 sq. deg area. The second-last column lists the filters for which the tile has been completed (i.e., observed within constraints), and the last column lists the number of epochs taken in the K_s band within the first observing season (DR1).

All Tables

All Figures