THE INFRARED MEDIUM-DEEP SURVEY. V. A NEW SELECTION STRATEGY FOR QUASARS AT z > 5 BASED ON MEDIUM-BAND OBSERVATIONS WITH SQUEAN

Abstract

Multiple color selection techniques are successful in identifying quasars from wide-field broadband imaging survey data. Among the quasars that have been discovered so far, however, there is a redshift gap at 5 ≲ z ≲ 5.7 due to the limitations of filter sets in previous studies. In this work, we present a new selection technique of high redshift quasars using a sequence of medium-band filters: nine filters with central wavelengths from 625 to 1025 nm and bandwidths of 50 nm. Photometry with these medium-bands traces the spectral energy distribution (SED) of a source, similar to spectroscopy with resolution R ~ 15. By conducting medium-band observations of high redshift quasars at 4.7 ≤ z ≤ 6.0 and brown dwarfs (the main contaminants in high redshift quasar selection) using the SED camera for QUasars in EArly uNiverse (SQUEAN) on the 2.1-m telescope at the McDonald Observatory, we show that these medium-band filters are superior to multi-color broad-band color section in separating high redshift quasars from brown dwarfs. In addition, we show that redshifts of high redshift quasars can be determined to an accuracy of Δz/(1 + z) = 0.002 - 0.026. The selection technique can be extended to z ~ 7, suggesting that the medium-band observation can be powerful in identifying quasars even at the re-ionization epoch.

1. INTRODUCTION

Recently, more than 0.4 million quasars have been discovered (Flesch 2015), among which 70 lie at z ≳ 6 (e.g., Fan et al. 2003, 2006; Willott et al. 2007, 2010; Jiang et al. 2008, 2009, 2015; Mortlock et al. 2009, 2011). Using these high redshift quasars, we can study the evolution of supermassive black holes (SMBHs; e.g., Jiang et al. 2007; Kurk et al. 2007; Jun et al. 2015) and the metal enrichment history of the early universe when the age of the universe was less than 1 Gyr (e.g., Jiang et al. 2007; Kurk et al. 2007; Juarez et al. 2009; De Rosa et al. 2011, 2014). In addition, they allow us to examine the line-of-sight condition of the intergalactic medium (IGM), such as the re-ionization state (e.g., Fan et al. 2006; Carilli et al. 2010; McGreer et al. 2011).

However, there exist redshift gaps in current high redshift quasar surveys, especially in the range of 4.8 ≤ z ≥ 5.7. Although some quasar surveys were performed at z ~ 5 or z ~ 6 (e.g., Zheng et al. 2000; Sharp et al. 2001; Schneider et al. 2001; Fan et al. 2003, 2006; Mahabal et al. 2005; Cool et al. 2006; Willott et al. 2007, 2010; Jiang et al. 2008, 2009, 2015; Wu & Jia 2010; Matute et al. 2013; McGreer et al. 2013), they could not detect quasars at 5 ≲ z ≲ 5.7 efficiently (Figure 16 in Jun et al. 2015). This was due to the limitations of the filter systems currently employed by these surveys as described below.

Most of the quasars at z ~ 5 were discovered by the Sloan Digital Sky Survey (SDSS; Richards et al. 2002; McGreer et al. 2013), using the r−i color for selecting r-dropouts and the i−z color for removing brown dwarfs. Also Ikeda et al. (2012) used the r − i vs. i − z color-color diagram and discovered only one type-2 quasar at z = 5.07. However, these studies were unsuccessful in finding quasars at 5 ≲ z ≲ 5.7 because the r-dropout quasars, starting from z ~ 4.6, blend with brown dwarfs on color-color diagrams for z ~ 5.1 (Jeon et al. in preparation). Wu & Jia (2010) adopted new color cuts, J−K vs. i − Y to find high redshift quasar candidates, but their method is limited to z < 5.3. On the other hand, for quasar surveys at z ~ 6, i−z vs. z−J color-color diagrams are usually used. However the i-dropouts start from z ~ 5.7; thus, most surveys for z ~ 6 are limited to z > 5.7. Therefore we need a new method and a new filter system which exploits the wavelength range between conventional filters to find quasars at redshift between 5.1 and 5.7.

To discover quasars at 5 ≲ z ≲ 5.7, we have been conducting an imaging survey as a part of the Infrared Medium-deep Survey (IMS; Im et al. in preparation). This imaging survey has used the Camera for QUasars in EArly uNiverse (CQUEAN; Park et al. 2012; Kim et al. 2011; Lim et al. 2013), with custom-designed is and iz filters that compensate for previous filter systems. Since the central wavelengths of these filters are located between r and i, and between i and z, respectively, it was expected that these filters will be effective in selecting high redshift quasars at 5 < z < 6, where SDSS or other filter systems cannot explore (Jeon et al. in preparation). The quasar candidates were selected from r-dropout objects of z < 19.5 mag over 3,400 deg2 using multi-wavelength data such as SDSS, the United Kingdom Infrared Telescope Infrared Deep Sky Survey (UKIDSS; Lawrence et al. 2007) Large Area Survey (LAS), the Two Micron All Sky Survey (2MASS), and theWide-field Infrared Survey Explorer (WISE). So far, 6 r-dropouts in 4.7 ≤ z ≤ 5.4, and 2 i-dropouts in 5.9 ≤ z ≤ 6.1 have been confirmed as quasars with optical spectroscopic observations (Jeon et al. in preparation); these are referred to as IMS quasars. The discovery of 6 r-dropout quasars at 4.7 ≤ z ≤ 5.4 is encouraging, but the same technique has not been successful at uncovering quasars at 5.4 < z < 5.7. This is mostly because the adopted color cuts cannot yet remove contaminants such as late type stars or brown dwarfs effectively. Therefore, it is highly desirable to have a new method based on a new filter system for selecting quasars in this range.

Matute et al. (2013) adopted a spectral energy distribution (SED) fitting method and discovered a faint quasar at z = 5.41 from the Advance Large Homogeneous Area Medium Band Redshift Astronomical (ALHAMBRA; Moles et al. 2008) survey over ~1 deg2. Their survey included a filter set composed of 20 optical medium-band and 3 broad-band J/H/Ks filters, and an extensive library of SEDs including those of active galaxies. Despite the fact that high redshift quasar candidates are contaminated by many low redshift galaxies and late type stars, they were able to select the quasar at z = 5.41 via SED fitting. This demonstrates that low resolution spectroscopy with medium-band filters can be an effective way to select high redshift quasars.

As an attempt to improve the selection of high red-shift quasars, we devised a set of 9 medium-band filters to better trace SEDs of various sources and modified CQUEAN to allow the usage as many as 20 filters. We now call this modified CQUEAN the SED camera for QUasars in EArly uNiverse (SQUEAN; Kim et al. in preparation; Choi et al. 2015). Like CQUEAN, SQUEAN operates on the 2.1 m Otto Struve Telescope at the McDonald Observatory. To explore the potential of selecting high redshift quasars using our mediumband filters, we observed known high redshift quasars and brown dwarfs, and conducted SED fitting to distinguish these two populations. This paper reports the result of the study. Section 2 describes the overview of our new medium-band filters and two test observations. In Sections 3 and 4, we describe the performance from the result of medium-band photometry including SED fitting results. We discuss the feasibility of using medium-band filters on small or medium-class telescopes for future high redshift quasar selection in Section 5. We summarize our results in the final section. Throughout this paper, we use a cosmology with ΩM = 0.3, Ω∧ = 0.7, and H0 = 70 km s−1Mpc−1, and the AB magnitude system.

 

2. OBSERVATIONS

2.1. SQUEAN Medium-band Filters

The SQUEAN medium-band filter set consists of nine filters, m625, m675, m725, m775, m825, m875, m925s, m975, and m1025, where ”m” stands for ”medium” and the number indicates the central wavelength in nm (Table 1). The bandwidths of these filters are fixed to 50 nm. In addition, SQUEAN has an m925n filter, whose bandwidth is 25 nm and has a central wavelength identical to that of m925s. These filters are off-the-shelf products from Edmund Optics, Inc.. The optical characteristics of the filters are described in a separated SQUEAN instrument paper (Kim et al. in preparation).

Table 1a Kim et al. in preparation

Table 2Color cuts and redshift ranges for quasar selection in Figure 6

Figure 1 shows filter transmission curves of the medium-band filters (thin solid lines) and other SQUEAN filters (dashed lines) with the detector quantum efficiency (QE; thick solid line) taken into account. The figure shows that the medium-band filter bandwidths are several times narrower than those of broadband filters, such as SDSS r/i/z.

Figure 1.The QE of SQUEAN (thick solid line) and the combined response function of the detector QE and filters. The thin solid lines are for our medium-band filters and the dashed lines are for the CQUEAN g/r/i/z/Y.

2.2. SQUEAN Medium-band Observations

We carried out test observations with a subset of medium-band filters installed in CQUEAN in 2014 November and with a complete set of medium-band filters in SQUEAN in 2015 February. During the two observing runs, we observed 6 IMS quasars at 4.7 ≤ z ≥ 6.0 and 5 brown dwarfs (Table 3) with m675, m725, m775, m825 and m875 filters. The observations were carried out under mostly clear sky conditions. However, both observations were executed during bright time, and seeing values varied from 1.3” to 4.0”.

Table 3Note. a Integration times of the m675,m725,m775,m825, and m875 filters respectively, for the test observations b Seeing values in the m675,m725,m775,m825, and m875 filters respectively, for the test observations c Yi et al. (2015) provide a more accurate redshift, zspec = 4.76. d Jiang et al. (2015) provide a more accurate redshift, zspec = 5.923 ± 0.003.

Preprocessing, such as bias subtraction, dark subtraction and flat fielding were conducted with usual data reduction procedures using the IRAF1 noao.imred.ccdred package. Then we combined images of each field and each filter in average. We used the ccmap task of IRAF and SCAMP (Bertin 2006) to derive astrometric solutions.

SExtractor (Bertin & Arnouts 1996) was used for source detection and photometry. We derived auto-magnitudes which are taken as the total magnitudes. We estimated the limiting magnitudes for each medium-band filter using the data taken on 2015 February 8. The 5σ detection limits of a point source with 1 hour integration time are 23.7 mag for m675 and m725, 23.3 mag for m775 and m825, and 22.7 mag for m875-bands.

2.3. Photometric Calibration

For photometric calibration, we used stars that have SDSS photometry and are in the same field as our target. Using SDSS r/i/z magnitudes of a star, we did a X2 fitting to the SED of the object to find the best spectral template among stellar spectral templates. We used the stellar templates from Gunn & Stryker (1983), that contain 175 spectra with various stellar types. From the best-fit template, we calculated model magnitudes for each medium-band filter by convolving their filter transmission curves and defined these values as their standard magnitudes (Mstd) of each filter. Figure 2 shows examples of the X2 fitting of stars, with their r/i/z magnitudes (boxes) and the best-fit templates (solid lines). Then, the Mstd of the stellar sources were compared with their instrument magnitudes (mins; mins = − 2.5 log(DN)), and the differences between the Mstd and the mins were used to estimate the zero-point (Zp). During the Zp calculation, faint objects or objects with large reduced values were rejected: we chose stars of and instrument magnitude error < 0.05 mag for each band. Figure 3 shows an example of the Zp calculations for each filter in the case of IMS J143704.82+070808.3. The crosses are for all sources in the field and the boxes are for sources that satisfy the conditions explained above. After 3σ clipping, we calculated the Zp as the average value of the magnitude differences. For the Zp uncertainty, we used the standard deviation value of the Zp values. The average zero-point error is about 0.05 mag. The standard magnitudes of our targets were calculated as Mstd = Zp + mins.

Figure 2.Examples of the X2 fitting of SDSS stars. The solid line shows the best-fit template and the boxes are from SDSS r/i/z photometry.

Figure 3.The Zp calculations of IMS J143704.82+070808.3 field for each filter. The crosses are for all sources in the field and the boxes are sources that satisfy the conditions described in Section 2.3.

This calibration technique was compared to results from standard star observations. We used data from the observation of 2015 February 5. BD+33 2642 was observed with the 5 filters at two different airmasses. To calculate the standard magnitudes of the star for each filter, we used its spectrum obtained from Oke (1990), and the filter transmission curves. The Zp values from the standard star data are found to be consistent with the values derived from SDSS stars.

 

3. PROPERTIES OF MEDIUM-BAND PHOTOMETRY

3.1. Filter Response Curves with Quasar Spectra

We plot the filter response curves of medium-band filters (thin solid lines) with quasar SED templates at different redshifts (thick solid lines) in Figure 4. For the quasar SEDs, we used the median composite spectrum of SDSS quasars from Vanden Berk et al. (2001) and adopted the IGM attenuation model from Madau et al. (1996). We also plot the model magnitudes from the SED templates in each filter with squares. We can notice that the Lyman α (Lyα) emission line is located in the m725-band at z = 5.0, in the m775-band at z = 5.5 and between the m825 and m875-bands at z = 6.0, and the model magnitudes show a sharp drop in flux shortward of these filters. We can see that the photometry with these medium-bands trace the SED of sources, similar to spectroscopy with resolution R ~ 15.

Figure 4.Filter response curves (thin solid lines) with quasar spectra (thick solid lines) redshifted to z = 5.0, 5.5, and 6.0. The model magnitudes in each filter (squares) trace the SEDs of the sources similar to spectroscopy with R ∼15.

3.2. Medium-band Spectral Energy Distributions

Table 4 lists the medium-band magnitudes of our targets with g/r/i/z/Y/J/H/K-band photometry from SDSS and UKIDSS LAS. The magnitude errors of the medium-band filters are computed by combining the standard SExtractor estimates based on Poisson statistics and the zero-point errors from Section 2.3.

Table 4Photometric information of targets

The medium-band filters trace SEDs of high redshift quasars and brown dwarfs differently. In Figure 5, we show the medium-band SEDs of quasars and brown dwarfs, where the red squares are the medium-band data and the blue squares are the SDSS r/i/z magnitudes. The gray lines in the upper two rows are quasar templates from Vanden Berk et al. (2001) redshifted appropriately, with IGM attenuation (Madau et al. 1996). The gray lines in the lower two rows are model brown dwarf spectra with a temperature of 2,200K Burrows et al. (2006). Note that IMS J012247.33+121623.9 is a broad absorption line (BAL) quasar, which shows a deep absorption feature near the wavelengths of the m775-band. High redshift quasars show a sharp drop in flux in the filter that covers the wavelength range blueward from the redshifted Lyα emission line. On the other hand, SEDs of brown dwarfs smoothly increases from short to long wavelengths. Such a difference in SEDs is much less pronounced in the r/i/z broad-band photometry. The drastic magnitude drop at the redshifted Lyα is key to distinguishing high red-shift quasars from brown dwarfs, and the medium-band filters offer sufficient spectral resolution for such a purpose. These dropouts can also be recognized in colorcolor diagrams.

Figure 5.Medium-band SEDs of quasars and brown dwarfs from SQUEAN observations (red squares). Also we plot the SDSS r/i/z data (blue squares), the redshifted quasar SEDs (gray lines in upper two rows), and the model brown dwarf spectra (gray lines in lower two rows). This figure demonstrates that high redshift quasar SEDs and brown dwarf SEDs can be distinguished easily thanks to the fine wavelength sampling of the medium-band filters.

3.3. Color-Color Diagrams

Figure 6 shows 7 color-color diagrams from combinations of all nine medium-band filters (black lines are quasar redshift tracks within specific redshift ranges, light green triangles are model brown dwarfs from Burrows et al. (2006), dark green boxes with error bars or arrows are IMS brown dwarfs, and red circles with error bars or arrows are IMS quasars). Figure 6 (a) is for m625−m675, vs. m675−m725, (b) is for m675−m725, vs. m725 − m775, (c) is for m725 − m775 vs. m775 − m825, (d) is for m775−m825 vs. m825−m875, (e) is for m825−m875 vs. m875−m925, (f) is for m875−m925 vs. m925 − m975, and (g) is for m925 − m975 vs. m975 − m1025. Figure 6 (b), (c), and (d) contain the medium-band colors from our observations. From Figure 6 (b), we can deduce that the m675-dropout is found for IMS J012247.33+121623.9 (z=4.86) and IMS J143704.82+070808.3 (z=4.97). On the other hand, it is difficult to select IMS J032407.70+042613.3 (z=4.71) as a high redshift quasar, because its Lyα drop is located at the m675-band and its colors using these 3 filters are similar to those of brown dwarfs. Note that the BAL quasar at z=4.86, IMS J012247.33+121623.9, is off from the redshift tracks due to broad absorption lines in its continuum. For similar reasons, in Figure 6 (c), IMS J222514.39+033012.6 (z=5.35) and IMS J015533.28+041506.8 (z=5.39) also show red m725 − m775 colors relative to IMS J032407.70+042613.3 (z=4.71), IMS J012247.33+121623.9 (z=4.86), and IMS J143704.82+070808.3 (z=4.97) due to the m725-dropout. We can see that Figure 6 (b), (c), and (d) can be used for selecting quasars at 4.8 < z < 5.8 because the positions of the quasars are significantly different from those of the brown dwarfs, distanced by ≳ 1 mag from the quasar tracks. For the quasar selection at z > 5.8, Figure 6 (e), (f), and (g) show that the quasars at z > 5.8 can be separated from brown dwarfs clearly. Therefore we can use these color-color diagrams from medium-band photometry for quasar candidate selections at 4.7 < z < 7. Table 2 lists the color cuts (blue solid lines in Figure 6) and the redshift ranges for quasar selection. Since Lyman break galaxies (LBGs) at high redshift have ultraviolet SEDs similar to those of quasars, these selection criteria can also be used for selecting high redshift star-forming galaxies. On the other hand, we expect that the objects satisfying the color cuts at bright magnitude (< 21 AB mag; e.g., Shim et al. 2007) are predominantly quasars.

Figure 6.Color-color diagrams of successive combinations of nine medium-band filters. The black lines are quasar redshift tracks with diamond symbols indicating redshifts at a Δz = 0.2 interval, except the last redshift point is chosen to be the redshift where Lyα is located at 12.5 nm above the central wavelength of the second filter. The light green triangles are brown dwarfs, the dark green boxes are IMS brown dwarfs, and the red circles are IMS quasars.

For comparison, we plot various color-color diagrams using SDSS and UKIDSS LAS filters in Figure 7. Since the r-band has wavelengths similar to those of m675, we plot the r −m725 vs. m725−m775 color-color diagram in Figure 7 (a). In Figure 7 (a), the colors of the model quasar redshifted to 4.7 < z < 5.1 are not different from those of brown dwarfs, contrary to Figure 6 (b). This is due to the difference in wavelength coverage between the r and m675-bands. Since the m675 filter has a narrower bandwidth compared to the r filter, the m675 − m725 color can measure the Lyα dropout feature clearly, without being affected by other spectral features. For example, IMS J143704.82+070808.3 at z = 4.97 shows a higher flux in the r-band than in the m675-band because the r filter also covers the Lyman β emission line at this redshift. Therefore, the dropout effect between r and m725 is not strong enough to distinguish quasars from brown dwarfs.

Figure 7.Similar to Figure 6, but including SDSS and UKIDSS LAS filters.

To check the possibility of selecting quasars at z ~ 6 with our medium-bands and near-infrared bands, we adopted the Y or J-bands ((b) and (c) of Figure 7). The z−Y or z−J colors of z ~ 6 quasars are different from those of brown dwarfs, which is used in a traditional selection method to select z ~ 6 quasars, such as the i − z vs. z − J color-color diagram (e.g., Fan et al. 2001, 2003, 2004, 2006; Jiang et al. 2008, 2009; Willott et al. 2007, 2009, 2010). Therefore, using these three diagrams, we are able to select quasars at z ~ 6 that are separated from the brown dwarfs.

In Figure 7 (d), the z−m975 vs. m975−J color-color diagram shows a clear separation between the quasar redshift track and brown dwarfs, similar to Figure 6 (g), which is also suitable for selecting z ~ 7 quasar candidates.

 

4. SED FITTING FROM MEDIUM-BAND PHOTOMETRY

To estimate the efficiency of the medium-band filters, we performed SED fitting using templates of the SDSS quasar composite spectrum from Vanden Berk et al. (2001) with IGM attenuation (Madau et al. 1996). SEDs from the medium-bands or other bands are compared to the quasar templates redshifted to various redshifts and values between the SEDs and the quasar templates were calculated. We found the best-fit template of a source using the minimization of values and estimated a photometric redshift (zphot) from the template.

We used several filter sets for the SED fitting, combining our medium-bands and SDSS filters, to compare fitting performances while using different filter sets. We define the filter sets as follows:

Filterset1={m675,m725,m775,m825,m875}

Filterset2={r,m725,m775,m825,m875}

Filterset3={m675,m725,m775,m825, z}

Filterset4={r,m725,m775,m825, z}

Filterset5={r, i, z}

4.1. Best-fit Values of Quasars and Brown Dwarfs

After finding the best-fit templates for each target, we compared the best-fit values for quasars and brown dwarfs for a specific filter set. Figure 8 shows the average values of the best-fit values for quasars and brown dwarfs for each filter set. We exclude IMS J012247.33+121623.9, since it is a BAL quasar, whose spectra are known to be clearly different from normal quasar spectra. The ranges of the best-fit values are denoted with error bars. We can see that the best-fit values from the quasar and the brown dwarfs are quite different, although some brown dwarfs show small values comparable to those of quasars. On average, for Filterset1 – 4, which include 3 medium-bands that substitute the i-band wavelengths (m725,m775,m825), the best-fit values for the quasars are 2 – 13 times smaller than those for the brown dwarfs. Therefore the SED fitting from photometry using SDSS filters and these 3 medium-band filters can be used for quasar candidate selection at 5 < z < 6 efficiently: we can choose robust quasar candidates among large samples selected from broad-band photometry, rejecting contaminants that have large values. On the other hand, in the case of Filterset5, which contains only r/i/z-bands without medium-bands, the best-fit values of the brown dwarfs are instead smaller than those of the quasars. Therefore we might misclassify brown dwarfs as actual quasars via SED fitting if we use photometry from the r/i/z-bands only. We tested with another filter set {r,m725,m775,z} and found that this filter set shows poor ability to separate quasars from brown dwarfs using their best-fit values.

Figure 8.The average values of the best-fit values for quasars and brown dwarfs from each filter set. We exclude IMS J012247.33+121623.9, since it is a BAL quasar. The ranges of the best-fit values are denoted with error bars.

4.2. Photometric Redshifts of Quasars

Using the 5 medium-band filters from m675 to m875, we estimated zphot via X2 fitting. Figure 10 shows the distributions as a function of redshift for each quasar. The zphot values are estimated from redshifts where c has a minimum value. The red dashed lines indicate zspec and the blue solid lines are for zphot. The errors of the zphot (blue) and zspec (red) are denoted in the gray boxes on each plot.

The uncertainties of zphot (Δzphot) are estimated from the 1σ deviations (confidence level of 68.3%) of the distribution, where the takes on a value one greater than its minimum. The zspec uncertainties are estimated as discussed in Jeon et al. (in preparation). We can see that zphot agree with zspec within their redshift uncertainties. The Δzphot values are less than 0.07, which shows the power of the medium-band photometry to estimate accurate zphot of high red-shift quasars via X2 fitting of SEDs. The zphot values and their errors are provided in Table 3. Figure 9 shows zphot versus zspec, with the solid line indicating zspec=zphot, and this figure also demonstrates the excellent agreement between the two values. The average absolute deviation of zphot − zspec (Δz) from this comparison is only 0.012 in Δz/(1+z) (dotted lines), which is better than the value we get from the formal error of the fit.

Figure 9.Photometric redshifts from medium-band photometry versus spectoroscopic redshifts of high redshift quasars. The dotted lines indicate Δz/(1+z) = 0.012. The solid line indicates the case where zspec = zphot. The comparison shows that medium-band data can provide photometric redshifts with an accuracy of ∼1%.

Figure 10.The values as a function of redshift for each quasar when we use Filterset1. The red dashed lines indicate the zspec values and the blue solid lines indicate for the zphot values. The errors of the zphot (blue) and zspec (red) are denoted in the gray boxes.

Richards et al. (2009, 2015) provide zphot values for 3 of the 6 quasars, IMS J012247.33+121623.9, IMS J143704.82+070808.3, and IMS J222514.39+033012.6, estimated from broad-band photometry, u/g/r/i/z/J/H/K. Their estimate for IMS J012247.33+121623.9 ( from Richards et al. (2015)) does not agree with our spectroscopic redshift, although IMS J143704.82+070808.3 ( from Richards et al. (2009) or from Richards et al. (2015)) and IMS J222514.39+033012.6 ( from Richards et al. (2015)) are in agreement with our estimates. Also these estimates show large Δzphot values compared to those from our medium-band photometry.

 

5. OBSERVATION STRATEGY

In this section, we discuss whether medium-band filters can be used on small to medium-sized telescopes for high redshift quasar selection. We showed that three medium-bands (m725,m775, and m825 filters) are sufficient to remove brown dwarfs through the SED fitting when SDSS photometric information is provided. Considering the depths of the medium-bands and the average magnitude of our candidates (i = 19.9 mag and z = 19.3 mag), we need 6, 6, and 3 minutes for m725, m775, and m825-band observations with the McDonald 2.1-m telescope and SQUEAN, respectively, to reach a S/N ~ 25 (corresponding photometric error of 0.04 mag) for detection or a 5σ upper limit at 6μJy for dropouts in m725 and m775. In this case we need 15 minutes per target for photometry in these three medium-bands. If a similar filter set is used on a 0.5-m class telescope (e.g., LSGT; Im et al. 2015), several hours of observing time is necessary for each target.

It is possible to extend a similar selection method to higher redshifts (z = 6.0, 6.5 and 7.0; see Figures 6 (e), (f), and (g)). A particularly interesting point is of the redshift range at z > 6.4 where the lack of wide-field near-infrared imaging survey data and severe brown dwarf contamination make quasar selection difficult. A luminous quasar at z = 7.085 (Mortlock et al. 2011) has Y = 20.3 mag and Lyα at 985 nm. Therefore, a combination of m925s, Y, and J filters or m925s, m975, and m1025 filters can determine whether such an object is a promising quasar candidate. To securely identify the break, we need to reach m925s ~ 22.3 mag at 5σ, which is possible with a few hours of exposure time on a 2-m class telescope under nominal observing conditions.

 

6. SUMMARY

We show the results from the medium-band observations of 6 high redshift quasars at 4.7 ≤ z ≤ 6.0 and 5 brown dwarfs. The SEDs of the quasars and brown dwarfs from the medium-band photometry show discrepancies that are much more significant than when using broad-band photometry, and their distant loci on various color-color diagrams can effectively distinguish the two populations. The best-fit values of quasars and brown dwarfs are quite different even with information from only 3 medium-bands and SDSS photometry. These results from the medium-bands cannot be reproduced solely with SDSS photometry. Also, the X2 fitting of the SEDs can estimate accurate zphot within an accuracy of Δz/(1+z) ~ 0.012 if we use the 5 mediumband filters discussed in this paper.

Therefore, although previous quasar surveys are incapable of covering the redshift range between 5 and 6 due to the limitations of classic filter sets, our new technique using the medium-bands will be a powerful method for selecting high redshift quasars in this red-shift gap. Two meter class telescopes, such as the Mc-Donald 2.1-m telescope with SQUEAN, can conduct imaging follow-up observations of high redshift quasar candidates of z < 19.3 mag in these 3 medium-band filters within 15 minutes per target. For higher red-shift quasars, larger telescopes are capable of carrying out the quasar survey at z > 6 using longer bandpass medium-band filters with a high efficiency.