1. INTRODUCTION

Searching for extrasolar planets is one of the primary research themes that is pursued with most modern astronomical instruments. In particular, the detection of earth-like planets in the habitable zone is regarded as the primary milestone of revealing the existence of extra-terrestrial life. According to the well-known database of extrasolar planets (http://exoplanet.eu; Schneider et al. 2011), as of November 2015, a total of about 2 000 planets have been detected since the first extrasolar planet was discovered around the pulsar PSR1257+12 (Wolszczan & Frail 1992). This number includes about 1 000 transiting planets discovered with the Kepler space telescope. In addition, the Kepler mission lists about 3 700 candidates that show transit-like signatures (from http://exoplanetarchive.ipac.caltech.edu).

It was first proposed by Mao & Paczynśki (1991) that gravitational microlensing can be a fruitful way to search for extrasolar planets. More than 10 years after this theoretical prediction, the first planetary microlensing event was reported by Bond et al. (2004). The microlensing detection rate is very low because this effect occurs only when both the source and lens star are precisely aligned along the observer's line of sight. Therefore, detecting microlensing planets requires longterm monitoring of a large amount of stars by using a wide-field photometric survey system. The discovery of the current ~40 planets with microlensing is the result of two photometric survey groups, the Optical Gravitational Lensing Experiment (OGLE; Udalski et al. 1994, 2015) and the Microlensing Observations in Astrophysics (MOA; Sumi et al. 2011), which are dedicated to monitoring dense star fields toward the Galactic bulge. Most of these detections were achieved from collaborating with intensive follow-up observation groups such as the Microlensing Follow-Up Network (Micro-FUN; Gaudi et al. 2008), the Probing Lensing Anomalies NETwork (PLANET; Beaulieu et al. 2006), and RoboNet (Tsapras et al. 2009). Compared to more successful planet detection methods, i.e., the radial velocity and transit, microlensing is more sensitive to planets in the outer region of planetary systems and is less dependent on the mass ratio between a planet and its host star. These are very important advantages for detecting earth-like planets in the habitable zone around solar-type stars.

KASI initiated the KMTNet project in January 2009, which is to install 2 m-class wide-field optical telescopes and mosaic CCD cameras at three southern observatories. These telescopes and cameras were planned to be used for 24-hour continuous monitoring of the Galactic bulge with a high cadence of about 10 minutes. This system configuration and observation strategy were proposed by the microlensing community as a next generation microlensing experiment, and the community predicted that the KMTNet-like system could detect a few hundred of extrasolar planets per year, including freeoating planets (Gaudi et al. 2010; Gaudi 2012; Henderson et al. 2014).

As the project name implies, the KMTNet aims primarily to discover extrasolar planets based on the analyses of gravitational microlensing events, especially the detection of earth-mass planets in the habitable zone. In addition, we anticipate a variety of scientific cases that can take advantage of the wide-field observation system. The KMTNet has a similar size of telescope mirror and field of view (FOV) with two well-known wide-field photometric survey systems, the Panoramic Survey Telescope And Rapid Response System (Pan-STARRS; Kaiser et al. 2010) and the SkyMapper (Keller et al. 2007). This implies that the systems' targets like the earth-approaching objects and external galaxies can also be good research subjects to be pursued with the KMTNet. Given the KMTNet's focus on the Galactic bulge for the detection of extrasolar planets, other science programs are performed in the season when the Galactic bulge is not observable.

In this paper, we present the technical specifications of the observation system and test observations in Section 2. The data transfer and processing are summarized in Section 3. Finally, we describe the KMTNet science programs in Section 4.

2. WIDE-FIELD OBSERVATION SYSTEM

2.1. Optical Telescope

Wide-field optics is essential for this survey due to the very low detection rate of microlensing phenomena. The most important technical requirement was, therefore, to have a large FOV of more than 2.0 by 2.0 square degrees. Considering this requirement and some technical constraints, we designed a telescope that had a primary mirror of 1.6 m in diameter, an effective focal length of 5.16 m, and a prime-focus type optical configuration (Kim et al. 2010), as summarized in Table 1.

Table 1Summary of the KMTNet observation system

Image quality was also an important specification as reliable photometric results of stars in crowded fields are required. We specified the Delivered Image Quality (DIQ) in the I-band and V -band to be Full Width at Half Maximum (FWHM) < 1.0 arcsec within a 1.2 degree radius FOV, under an atmospheric seeing of 0.75 arcsec and up to 60 degrees zenith distance. After a reasonable allocation of the DIQ budget, the tolerance of the optical design was set to 0.376 arcsec (Kim et al. 2011). The optics consisted of a purely parabolic primary mirror and four prime-focus corrector lenses with all-spherical surfaces. This design made it easier to fabricate, align, and test the optics (Poteet et al. 2012). The fourth lens with the plano-convex surface acts as the camera's dewar window. The optical design was optimized in the I-band and there was negligible vignetting in the entire FOV.

The primary mirror was made of Zerodur ultra-low expansion glass. The mirror surface was coated with protected silver. It showed an excellent surface reflectivity of about 97% in the I-band with an expected lifetime of more than 4 years (Vucina et al. 2008). During the telescope operation period, the mirror will be recoated with silver or aluminum when necessary, depending on the site environment. The primary mirror is supported both axially and laterally inside the mirror cell, with a whiffle tree structure of 27 points and 12 elastomer support balls used around the outer edge of the mirror.

The telescope mechanical structure was designed on the basis of the successful 1.3 m telescope for the Two Micron All Sky Survey (2MASS; Skrutskie et al. 1997) project (Poteet et al. 2012). The telescope mount is a fork-type equatorial, as shown in Figure 1. If we use the Difference Image Analysis (DIA) technique for stellar photometry (see Section 3), the relative positions and intensity profiles of stars in a CCD image should remain the same during the observation run. Our experience showed that the equatorial type is more reliable in achieving this requirement than the Alt-Azi mount.

Figure 1.The KMTNet 1.6 m telescope (top) and enclosure building (bottom) installed at SAAO in South Africa. The Southern African Large Telescope (SALT) enclosure is shown at the left backside.

The telescope focus typically changes with variations in temperature. We designed a focus control system that moves the telescope head ring assembly, which is mounted on a flexural stage (Kim et al. 2011; Poteet et al. 2012). Three high precision actuators, 120 degrees apart from each other, push or pull the three spots of the exure, which moves the assembly and results in a focus change. Both the prime-focus optics and detector camera are attached to the head ring and move together as a single unit. This arrangement to control the focus has the important advantage of enabling us the tilt alignment of the telescope optics by controlling each actuator separately.

A commercial dome of 9.2 m in diameter from ASHDOME was installed at the telescope enclosure (Kappler et al. 2012), as shown in Figure 1. The windscreen moves up and down with the dome shutter. It blocks the strong wind and also significantly reduces the effect of moonlight. The enclosure has a ventilation system that cools down the telescope in day time. This minimizes the mirror seeing caused by air turbulence around the primary mirror in nighttime. We used the well-known telescope control program PC-TCSTM from COMSOFT, which contains built-in routines for dome rotation and dome shutter open/close. The dome has a position sensor and can be rotated synchronously with the telescope by using the PC-TCS program. Figure 2 shows the control architecture of the KMTNet subsystems.

Figure 2.Control architecture of the KMTNet subsystems.

A drawer-style filter changer can mount four spectral filters simultaneously (Figure 3). The filter is a square of 310 mm by 310 mm and has a thickness of 10 mm. Johnson-Cousins BV RI filters were installed at the three host sites. A camera shutter was attached to the filter changer as a single unit between the third corrector lens and the fourth one (i.e., camera window). The sliding shutter with rectangular blades results in uniform illumination over the entire FOV, even with a short exposure time of 1.0 second. As shown in Figure 2, the shutter open/close is driven by a camera controller.

Figure 3.Filter changer and camera shutter assembly.

2.2. Mosaic CCD Camera

Figure 4 shows the first KMTNet camera installed at the prime focus of the telescope at CTIO. A focal plane assembly of the camera consists of four 9k by 9k science CCDs (e2v CCD290-99) and four 1k by 1k guide CCDs (e2v CCD47-20). The science CCDs have a physical pixel dimension of 10 μm by 10 μm, which corresponds to a pixel angular scale of 0.40 arcsec. A total of about 340 million pixels cover 2.0 by 2.0 square degrees on the sky. The CCDs were coated with a deep depletion multi-4 design to achieve maximum quantum effciency (QE) across the 4 000-9 000 Å region. The resulting QE is about 85 % in B & V -bands, about 90 % in R-band, and about 80 % in I-band. The readout noise is about 10 electrons at a readout rate of 500 kHz and the full well depth is higher than 90 000 electrons, slightly different between readout channels. Four frame-transfer CCDs are mounted at the north-south-west-east sides of the science CCDs and used for the telescope autoguiding.

Figure 4.(Top) The camera's focal plane assembly mounted with four mosaic science CCDs and four small guide CCDs. (Bottom) The camera attached to the KMTNet telescope at CTIO. The head electronic box is mounted on the telescope top ring, shown at the left.

In order to decrease the dark noise, the science CCDs were cooled down to -110 ℃ by using three Polycold Compact Cooler (PCC) mechanical refrigerators. The noise was determined to be negligible at less than 1 electron per pixel per minute. The refrigerators have no moving parts in the cold heads and they provide a very low vibration performance. The three PCC compressors are located outside the telescope in a separate equipment room (Atwood et al. 2012). The inside of the camera's dewar is maintained at a high vacuum level of less than 5 × 10-5 Torr.

The dewar window, being the fourth lens of the wide field correctors, is used for transmitting light to the CCDs and for blocking the cool vacuum inside the dewar from the ambient atmosphere outside. The central area of the window can be cooler than the ambient dew point temperature due to the radiation cooling, which makes the water condense on the window's outside surface. Therefore, we continuously purge the outside space above the dewar window with dry gas.

The camera electronics have a few functions to produce an image such as photon capture, readout, and analog to digital conversion with 16 bits. The electronic controller also regulates the CCD temperature, opens or closes the camera shutter, and communicates with a control computer via an optical fiber link. A box in which the electronic components were assembled is mounted on the telescope top ring (Figure 4). Since any heat source along the telescope optical path degrades the image quality, the box cover is fully insulated to minimize the temperature difference between the box's outside surface and the ambient air. There is an air-to-glycol heat exchanger inside the box. The heat from the electronics is transferred into the glycol which circulates from the box to an air cooler installed in the equipment room (Atwood et al. 2012).

The readout time of an 18k by 18k full image is about 30 seconds with 32 readout channels, i.e., 8 channels per CCD. The overhead time to take one image was estimated to be about 70 seconds in total. This includes the pre-flush, the delay for closing the camera shutter, the readout, transferring the image to the computer memory, and saving it to the hard disk.

2.3. Test Observations

The first KMTNet telescope was installed successfully at CTIO in Chile on May 2014, the second one at SAAO in South Africa on August 2014, and the third one at SSO in Australia on November 2014. The test observations of the telescopes were performed by using a 4k CCD camera (SBIG STX-16803) for a few months at each site before the 18k mosaic CCD camera was installed. We carried out photometric monitoring of six globular clusters to search for variable objects and to investigate their physical properties. The data are under analysis and the results will be presented elsewhere in detail (Lee D.-J. et al. in preparation).

The first mosaic CCD camera was attached to the KMTNet telescope at CTIO in September 2014, the second one at SAAO in December 2014, and finally the third system at SSO in May 2015. Figure 5 shows a sample image taken at SSO on July 7, 2015. During the test run, we examined the radial profile of stars across the full FOV. Under an atmospheric seeing of about 0.7-0.8 arcsec, which was measured from the Differential Image Motion Monitor (DIMM) instrument, the FWHM of stellar profiles in I- and V -band images was estimated to be about 1.0 arcsec in average; as expected from the optical design, the stars at the outer four corners of the mosaic CCDs appeared to have a little larger FWHM, which were therefore excluded. The measured image quality hence satisfies the scientific requirements described above. More detailed technical performance of the telescope and camera will be presented in a separate paper (Lee et al. 2016).

Figure 5.(Left) A sample 18k mosaic CCD image of the Galactic bulge field, taken with the KMTNet-SSO system. Exposure time was 60 seconds in I-band. The gap between mosaic CCDs was measured to be about 184 arcsec for the east-west direction and about 373 arcsec for the north-south. (Right) A zoomed image with a FOV of about 3 by 6 square arcmin. The blooming features from saturated stars are visible in this image.

We obtained time-series CCD images of the Galactic bulge field during the test runs from February to September 2015. Figure 6 displays a sample light curve of the previously known RR Lyr-type star OGLE-BLG-RRLYR-7412 (RA2000 = 17:55:22.88, DEC2000 = -31:21:28.4, V = 17.629, I = 16.047). The figure shows fiux variations of the pulsating star for longer than 24 hours, measured continuously at the three KMTNet sites with a high cadence of about 10 minutes. The exposure time was 60 seconds in I-band. The folded phase diagram in the corner represents a well-defined RRab-type pulsating light curve; we used the period of 0.56497235 days and the epoch of H.J.D. 2 455 000.49654, derived by Soszyński et al. (2011).

Figure 6.Sample light curves of the RR Lyr-type pulsating star OGLE-BLG-RRLYR-7412 in the Galactic bulge field, which was observed continuously at the three KMTNet sites on June 20, 2015. The flux is in arbitrary units. The upper right corner shows the star's pulsation phase diagram.

3. DATA HANDLING

All of the raw CCD images are transferred from the three southern sites to the KMTNet data center at Daejeon in Korea via network communication. We installed a Data Transfer System (DTS) in a computer room at each site (Kim et al. 2015). The DTS continuously checks the hard disk of the observation computer, detects a new image file at the disk in real-time, converts the four images (obtained from the four mosaic CCD chips) to a single Multiple Extension FITS (MEF) format file, and sends it to the data center. The typical network bandwidth from each site to the data center is higher than 40 Mbps, so that one image file with a size of about 700 Mbyte can be transferred within 140 seconds. Most of the KMTNet observation runs have a cadence of longer than 150 seconds and therefore nearly all images can be uploaded to the data center during each night of observations.

The raw images are preprocessed by the KMTNet pipeline in the data center. The preprocessing includes some basic corrections such as crosstalk, overscan, trimming, bias, and flat fielding. World Coordinate System (WCS) information is also added to the image header to support accurate astrometric data of each target. Researchers can access the preprocessed images within ~1 day after the observation is performed.

The time-series CCD images of the Galactic bulge are further processed with the DIA package (Woźniak 2000) to detect any variable objects such as microlensing events, transiting extrasolar planets, eclipsing binaries, pulsating stars, and stellar flares. The DIA technique has been shown to produce very accurate photometric results for observations of dense star fields like the Galactic bulge and the central region of globular clusters. We use the DIAPL version developed by Wojtek Pych (http://users.camk.edu.pl/pych/DIAPL).

4. SCIENCE PROGRAMS

The observation time for the KMTNet primary science, i.e., the Galactic bulge monitoring, has already been allocated (Park et al. 2012). Figure 7 shows the bulge observing time (solid line) between February 20 and October 22, which is about 45% of the total observing time, as well as the non-bulge time (dashed line) at CTIO in Chile. This pattern is almost identical to the other two sites, SAAO and SSO, because of their similar latitudes.

Figure 7.Observation time chart of KMTNet. The solid line denotes the Galactic bulge season, the dashed line denotes the non-bulge time.

During the bulge season, some dedicated fields near the Galactic center are continuously monitored for 24 hours a day by using the three identical systems installed in the three southern countries. The monitoring data are used for the following science cases:

1) Detection of extrasolar planets by using the microlensing method

2) Detection of transiting extrasolar planets

3) Detection of variable objects

4) Data mining in search for asteroids and comets

The non-bulge season was open to the Korean astronomical community and the following seven programs have been selected:

1) KMTNet Supernova Project (KSP)

2) The KMTNet DEep Ecliptic Patrol of the Southern sky: The DEEP-South-survey and physical characterization of asteroids and comets

3) Deep wide-field imaging of nearby galaxies

4) Detection and physical characterization of ultrafaint Milky-Way satellite galaxies in the Southern Hemisphere

5) BV I & Hα photometric survey of the Magellanic Clouds

6) KMTNet Intensive Nearby southern Galaxy group Survey (KINGS)

7) Probing the vicinity of supermassive black holes with AGN variability

The research goals and observation strategies for these science programs were summarized by Park et al. (2012). Some observation time is offered to the host countries (i.e., Chile, South Africa, and Australia) and allocated for the director/engineering time.

5. CONCLUSION

We have successfully installed three identical photometric observation systems, i.e., KMTNet, at the southern sites of CTIO in Chile, SAAO in South Africa, and SSO in Australia. Each system consists of a wide-field 1.6 m telescope and an 18k by 18k mosaic CCD camera that covers a 2.0 by 2.0 square degree in the sky. After the installation, we carried out test observation runs for several months. These runs showed that all three systems have very good performances and satisfy the scientific requirements. After the test was completed, the science run started offcially on October 1, 2015 and the observations for selected programs are underway as scheduled.

The most important specification of the KMTNet system is its wide-field of view. The parameter AΩ, called etendue, is in general use to measure the power of a wide-field observation system, where A is the effective light-gathering area of the telescope and Ω is the FOV with a single exposure of the camera. The etendue of the KMTNet is about 6.0 m2 degree2, after considering the obscuration of about 0.5 m2 by the prime-focus instruments such as the field correctors and filter/shutter assembly. The value is comparable with those of recent world-leading survey telescopes, for example, 5.2 m2 degree2 for the SkyMapper 1.35 m telescope, 6.8 m2 degree2 for the VISTA 4.0 m telescope, and 13 m2 degree2 for the Pan-STARRS 1.8 m telescope (Tyson 2010). Therefore, the KMTNet system is very competitive for wide-field photometric surveys that study asteroids and external galaxies, for instance.

Furthermore, the KMTNet system has the unique capability to monitor a target continuously for 24 hours due to the identical systems installed at three southern observatories with different timezones of 6-10 hours and a similar latitude of about -30 degrees. There are only a handful of similar kinds of telescope networks. The best-known project, Las Cumbres Observatory Global Telescope Network (LCOGTN; Brown et al. 2013), operates two 2.0 m, nine 1.0 m, and three 0.4 m telescopes installed at four northern sites and three southern observatories. All of the LCOGT telescopes have a small FOV of less than 0.19 square degrees and cannot therefore be considered as wide-field survey systems. At the present time, the KMTNet is the only wide-field telescope network with an aperture larger than 0.5 m. We believe that the KMTNet can play an important role in time domain astronomy with a special focus on microlensing events, variable stars, and supernovae.