The Infrared Medium-deep Survey. III. Survey of Luminous Quasars at 4.7 \leq z \leq 5.4
Yiseul Jeon, Myungshin Im, Dohyeong Kim, Yongjung Kim, Hyunsung David Jun, Soojong Pak, Yoon Chan Taak, Giseon Baek, Changsu Choi, Nahyun Choi, Jueun Hong, Minhee Hyun, Tae-Geun Ji, Marios Karouzos, Duho Kim, Jae-Woo Kim, Ji Hoon Kim, Minjin Kim, Sanghyuk Kim, Hye-In Lee, Seong-Kook Lee, Won-Kee Park, Woojin Park, Yongmin Yoon
aa r X i v : . [ a s t r o - ph . GA ] J un Draft version June 27, 2017
Preprint typeset using L A TEX style emulateapj v. 01/23/15
THE INFRARED MEDIUM-DEEP SURVEY. III.SURVEY OF LUMINOUS QUASARS AT 4.7 ≤ Z ≤ ⋆ Yiseul Jeon † , ‡ , Myungshin Im † , ‡ , Dohyeong Kim † , Yongjung Kim , Hyunsung David Jun , Soojong Pak ,Yoon Chan Taak , Giseon Baek , Changsu Choi , Nahyun Choi , Jueun Hong , Minhee Hyun , Tae-Geun Ji ,Marios Karouzos , Duho Kim , Jae-Woo Kim , Ji Hoon Kim , Minjin Kim , Sanghyuk Kim , Hye-In Lee ,Seong-Kook Lee , Won-Kee Park , Woojin Park , Yongmin Yoon Draft version June 27, 2017
ABSTRACTWe present our first results of the survey for high redshift quasars at 5 . z . .
7. The search forquasars in this redshift range has been known to be challenging due to limitations of filter sets usedin previous studies. We conducted a quasar survey for two specific redshift ranges, 4.60 ≤ z ≤ ≤ z ≤ is and iz . Using these filters and a new selection technique, we were able to reduce thefraction of interlopers. Through optical spectroscopy, we confirmed seven quasars at 4.7 ≤ z ≤ − . < M < − . ∼ Keywords: observations – quasars: emission lines – quasar: general – quasar: supermassive black holes– surveys INTRODUCTION
Observations have shown that large numbers of quasarsare found at z ∼ > ∼ M ⊙ (e.g., Jiang et al. 2007; Kurk et al. 2007; Jun et al. 2015;Wu et al. 2015) and appear to be vigorously evolving(Shen et al. 2011; Jiang et al. 2010; Im 2009; Jun et al.2015). However, there is a dearth of quasars with mea- Center for the Exploration of the Origin of the Universe(CEOU), Astronomy Program, Department of Physics & As-tronomy, Seoul National University, 1 Gwanak-ro, Gwanak-gu,Seoul 151-742 Korea LOCOOP, Inc., 311-1, 108 Gasandigital2-ro, Geumcheon-gu,Seoul, Korea Jet Propulsion Laboratory, California Institute of Technol-ogy, 4800 Oak Grove Dr., Pasadena, CA 91109, USA School of Space Research, Kyung Hee University, 1732Deogyeong-daero, Giheung-gu, Yongin-si, Gyeonggi-do 446-701,Korea SongAm Space Center, 103, 185 Gwonnyul-ro, Jangheung-myeon, Yangju-si, Gyeonggido 482-812 Korea Astronomy Program, Department of Physics & Astronomy,Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 151-742 Korea Arizona State University, School of Earth and Space Explo-ration, PO Box 871404, Tempe, AZ 85287-1404, U.S.A. Korea Astronomy and Space Science Institute, 776Daedeokdae-ro, Yuseong-gu, Daejeon, Republic of Korea Subaru Telescope, National Astronomical Observatory ofJapan, 650 North A’ohoku Place, Hilo, HI 96720, U.S.A. ⋆ Based on observations made with ESO Telescopes at the LaSilla Paranal Observatory under programme 091.A-0878. † Visiting Astronomer, Kitt Peak National Observatory, Na-tional Optical Astronomy Observatory, which is operated bythe Association of Universities for Research in Astronomy, Inc.(AURA) under cooperative agreement with the National ScienceFoundation. ‡ E-mail: [email protected], [email protected] sured black hole masses that makes it difficult to inves-tigate how they evolved at 5 < z < ∼
6; (2)construct the black hole mass function to understand thecosmic emergence of the most massive quasars; (3) inves-tigate the spectral energy distributions (SEDs) of quasarsto explore whether quasars with very massive black holeshave a lower accretion disk temperature (Laor & Davis2011; Wang et al. 2014).The redshift gap at 5 < z < < z < ∼ Jeon et al. i - z [ A B ] z=4.0 z=4.5 z=5.0z=5.4z=5.8Quasar TrackBrown DwarfsStarSDSS Point Source (a) 0 1 2 3 4i - z [AB]0123 z - J [ A B ] z=5.0 z=5.8 z=6.0z=6.4z=6.8Quasar TrackBrown DwarfsStarSDSS Point Source (b) Figure 1.
Color-color diagrams adopted by Fan et al. (1999) (left) and Willott et al. (2009) (right) for high redshift quasar selection.The black solid lines with asterisks are quasar redshift tracks, the triangles are model colors of brown dwarfs, the squares are model colorsof stars, and the crosses are point-like sources from the SDSS Star Catalog. The quasar tracks from z = 5.1 to z = 5.7 coincide with latetype stars or brown dwarfs. The solid boxes indicate the quasar selection boxes. µ m)020406080100 T h r oughpu t ( % ) g r i z Yis iz R e l a ti v e F ν Figure 2.
Filter transmission curves of is and iz (black solidlines), SDSS gri , and LSST zY bands (colored dashed lines),and the QE of the CCD (gray solid line) of CQUEAN. Thegreen line represents the SDSS composite quasar spectrum fromVanden Berk et al. (2001) redshifted to z = 5, with IGM attenua-tion (Madau et al. 1996). r − i vs. i − z color-color diagram to identify quasars atz > r -dropout quasars) and Willott et al.(2009) used the i − z vs. z − J color-color diagram forquasars at z ∼ i -dropout quasars). Thesolid boxes indicate their quasar selection criteria. Wesee that r -dropout quasars at z > i -dropout quasars at z < r -dropout technique alone can-not be used for z ∼ ugriz or the Two Micron All Sky Survey (2MASS) JHK fil-ters cannot separate quasars at 5.1 < z < < z < is and iz , to supplement the previous filter sys-tems for selecting quasars at this redshift range. Sincethe central wavelengths of these filters are located be-tween r and i , and between i and z , respectively, we canselect high redshift quasars at this redshift gap, wherethe SDSS or other filter sets cannot explore. Second,we needed a special optical detector which has bettersensitivity than previous CCDs at longer wavelengths,leading to more efficient observations with the is and iz filters. Considering these requirements, we developed aCCD camera system, the Camera for QUasars in EArlyuNiverse (CQUEAN; Kim et al. 2011; Park et al. 2012;Lim et al. 2013). Equipping a deep-depletion CCD chipto provide high quantum efficiency (QE) at 0.7 – 1 µ m,we conducted follow-up imaging observations of quasarcandidates with the is and iz filters and narrowed downthe quasar candidates. CQUEAN was installed on the2.1-m Otto Struve Telescope at McDonald Observatoryin 2010 August, and it has since been used to obtain pho-tometric data for many scientific programs, including ourhigh redshift quasar survey. In Figure 2, we plot the filtertransmission curves of is and iz (black solid lines), andthe SDSS gri and the Large Synoptic Survey Telescope zY bands (colored dashed lines) installed on CQUEAN,with the QE of the CCD taken into consideration (graysolid line). The green line represents the SDSS compositequasar template redshifted to z ∼ ∼ is and iz filters.Section 2 describes our quasar selection algorithmincluding color cuts, multi-wavelength data used, andimaging and spectroscopic follow-up observations. Thephotometric and spectroscopic analysis of our discoveredquasars are shown in Section 3. We discuss our quasar se- MS III (cid:43)(cid:80)(cid:82)(cid:87)(cid:86)(cid:3)(cid:53)(cid:38)(cid:53)(cid:53)(cid:3)(cid:38)(cid:52)(cid:26)(cid:3)(cid:67)(cid:80)(cid:70)(cid:3)(cid:55)(cid:45)(cid:43)(cid:38)(cid:53)(cid:53)(cid:3)(cid:46)(cid:35)(cid:53)(cid:3)(cid:38)(cid:52)(cid:19)(cid:18)(cid:3)(cid:69)(cid:67)(cid:86)(cid:67)(cid:78)(cid:81)(cid:73)(cid:3) (cid:43)(cid:47)(cid:53)(cid:3)(cid:51)(cid:87)(cid:67)(cid:85)(cid:67)(cid:84)(cid:3)(cid:53)(cid:87)(cid:84)(cid:88)(cid:71)(cid:91)(cid:3)(cid:67)(cid:86)(cid:3)(cid:22)(cid:16)(cid:25)(cid:3)(cid:151)(cid:3)(cid:92)(cid:3)(cid:151)(cid:3)(cid:23)(cid:16)(cid:22) (cid:3) (cid:53)(cid:82)(cid:71)(cid:69)(cid:86)(cid:84)(cid:81)(cid:85)(cid:69)(cid:81)(cid:82)(cid:91)(cid:3) (cid:49)(cid:82)(cid:3)(cid:60)(cid:72)(cid:86)(cid:3) (cid:84)(cid:15)(cid:75)(cid:15) (cid:92) (cid:15)(cid:44)(cid:15)(cid:45)(cid:3)(cid:53)(cid:71)(cid:78)(cid:71)(cid:69)(cid:86)(cid:75)(cid:81)(cid:80)(cid:3) (cid:54)(cid:56)(cid:66)(cid:38)(cid:88)(cid:87)(cid:20)(cid:3) (cid:560)(cid:3) (cid:54)(cid:56)(cid:66)(cid:38)(cid:88)(cid:87)(cid:21)(cid:3) (cid:560)(cid:3) (cid:54)(cid:56)(cid:66)(cid:38)(cid:88)(cid:87)(cid:22)(cid:3) (cid:560)(cid:3) (cid:54)(cid:56)(cid:66)(cid:38)(cid:88)(cid:87)(cid:23)(cid:3) (cid:47)(cid:69)(cid:41)(cid:84)(cid:71)(cid:71)(cid:84)(cid:13)(cid:19)(cid:21)(cid:3)(cid:89)(cid:75)(cid:86)(cid:74)(cid:3)(cid:57)(cid:43)(cid:65)(cid:37)(cid:87)(cid:86)(cid:3) (cid:48)(cid:70)(cid:42)(cid:85)(cid:72)(cid:72)(cid:85)(cid:14)(cid:20)(cid:22)(cid:3) (cid:70)(cid:88)(cid:87)(cid:3)(cid:560)(cid:3) (cid:58)(cid:44)(cid:66)(cid:38)(cid:88)(cid:87)(cid:3) (cid:53)(cid:86)(cid:84)(cid:81)(cid:80)(cid:73)(cid:3)(cid:57)(cid:43)(cid:53)(cid:39)(cid:3)(cid:38)(cid:71)(cid:86)(cid:71)(cid:69)(cid:86)(cid:75)(cid:81)(cid:80)(cid:3)(cid:89)(cid:75)(cid:86)(cid:74)(cid:3)(cid:57)(cid:43)(cid:65)(cid:37)(cid:87)(cid:86)(cid:3) (cid:37)(cid:51)(cid:55)(cid:39)(cid:35)(cid:48)(cid:3)(cid:75)(cid:85)(cid:17) (cid:75)(cid:92) (cid:3)(cid:43)(cid:79)(cid:67)(cid:73)(cid:75)(cid:80)(cid:73)(cid:3) (cid:38)(cid:52)(cid:66)(cid:38)(cid:88)(cid:87)(cid:20)(cid:3) (cid:561)(cid:3) (cid:38)(cid:52)(cid:66)(cid:38)(cid:88)(cid:87)(cid:21)(cid:3) (cid:49)(cid:82)(cid:3) (cid:49)(cid:82)(cid:3)(cid:49)(cid:82)(cid:3)(cid:60)(cid:72)(cid:86)(cid:3) (cid:60)(cid:72)(cid:86)(cid:3) (cid:60)(cid:72)(cid:86)(cid:3) (cid:53)(cid:72)(cid:77)(cid:72)(cid:70)(cid:87)(cid:3)(cid:53)(cid:72)(cid:77)(cid:72)(cid:70)(cid:87)(cid:3) Figure 3.
Schematic flow diagram of the main quasar candidate selection algorithm. lection efficiency and expected number of quasars in Sec-tion 4. Section 5 presents physical properties of the newlydiscovered quasars from the spectroscopy. We summarizethis survey in the final section (Section 6). Throughoutthis paper, we use a cosmology with Ω M = 0 .
3, Ω Λ = 0 . H = 70 km s − Mpc − . Weuse the AB magnitude system. QUASAR SELECTION AND OBSERVATION
Quasar Candidate Selection
To select quasars at 5 < z < ∼ . The r, i , z , J ,and K magnitudes are used. Since the contaminationrate using these filters is still high, we adopted is and iz -band photometry to discriminate brown dwarfs from r -dropout objects. Then we set additional criteria toassign priorities for follow-up observations. No stellaritycut is made to avoid missing quasars that are classified tobe extended objects (e.g., due to host galaxy or noise instellarity calculation), although we used the stellarity asa way to set priorities for follow-up observation. Figure3 shows a main quasar candidate selection algorithm. r − i − z − J − K , is and iz-band Selections To select quasar candidates from broadband photom-etry, we used the dropout feature at the Lyman α (Ly α )emission line that are common in high redshift objects.The Ly α dropouts can be identified using the r − i colorfor quasars at z > r − i > > r − i , z − J ,and J − K : r − i to select dropout objects, z − J toremove brown dwarfs, and J − K to eliminate otherstars. Figure 4 shows two color-color diagrams with model brown dwarfs from Burrows et al. (2006) (greentriangles), observed brown dwarfs from Patten et al.(2006) and Zhang et al. (2009) (green squares), and stel-lar sources from the SDSS catalog (gray circles, ∼ r − i > .
5. To verify theposition of quasars at 4 < z <
6, we plotted previouslydiscovered quasars from the SDSS DR7 quasar catalogand Leipski et al. (2014) (crosses; the color indicates itsredshift, as shown on the color bar in Figure 4b). Thequasar redshift track at 4 < z < SU Cut1 ) r − i > . SU Cut2 ) [0 < J − K < ∩ [ − < z − J < . ∩ [( z − J ) < ( J − K ) + 0 . SU Cut1 is for selecting the r -dropout objects and SU Cut2 is for weeding out late type stars and browndwarfs. Since
SU Cut1 does not adopt the i − z cut, un-like Fan et al. (1999), quasar candidates at z ∼ SU Cut2 is close to the stellar locus (gray circles)and part of the stellar sources selected from
SU Cut1 isstill located inside
SU Cut2 (black circles inside
SU Cut2 ),the selected sample is still significantly contaminated bystars (more than 99% of the selected objects are expectedto be stars; see Section 4.2).To reduce stellar contamination in our sample, we im-pose magnitude cuts in the shorter wavelength data, aswell as in the z -band. We set the magnitude cuts asbelow: SU Cut3 ) u , g fainter than the 3 σ detection Jeon et al. -1 0 1 2 3 4r - i [AB]-101234 i - z [ A B ] z=4.0 z=4.5 z=5.0 z=5.5z=6.0 IMS QuasarsQuasars in LiteratureQuasar Redshift TrackModel Brown DwarfsBrown Dwarfs from Patten+06 and Zhang+09SDSS Stellar SourcesSDSS Stellar Sources with r-i > 1.5 (a)-1.0 -0.5 0.0 0.5 1.0 1.5J - K [AB]-0.50.00.51.01.52.0 z - J [ A B ] Redshift of Quasars in Literature (b)
Figure 4.
Two color-color diagrams we adopted for quasar se-lection at 5 . z . .
7. We plot the model brown dwarfs (greentriangles), observed brown dwarfs from Patten et al. (2006) andZhang et al. (2009) (green squares), stellar sources from SDSS(gray circles), previously discovered quasars (crosses), and the red-shift tracks of quasars at 4 < z < r − i > .
5, showing a highcontamination rate even after the z − J − K cut. We plotted our6 new quasars with red circles (this work) and most of them arewithin the selection boxes. One exception is IMS J0324+0426 inthe r − i − z color-color diagram, which was selected using the colorcuts of McGreer et al. (2013). limits ( u > .
85 and g > .
55 mag)
SU Cut4 ) z < . > < z < < z < ∼ ∼ is ∼
30 (Section4.2), showing that about 99% of these sources will be in-terlopers. This is because the selected candidates fromthese two color-color diagrams are still contaminated bystellar sources, which are shown as the black circles inside
SU Cut2 in Figure 4b. To eliminate these contaminants,
Selection method A -1 0 1 2 3r - is [AB]-0.50.00.51.01.52.0 i s - i z [ A B ] z=4.0 z=4.6 z=5.0z=5.4z=5.5 (a) IMS QuasarsCandidatesSDSS QuasarsQuasar z TrackStarsModel Brown DwarfsStar Forming GalaxyPassive GalaxyIMS QuasarsCandidatesSDSS QuasarsQuasar z TrackStarsModel Brown DwarfsStar Forming GalaxyPassive Galaxy
Selection method B i z - J [ A B ] z=5.2 z=5.5 z=6.0z=6.3z=6.5 (b) Figure 5.
Two color-color diagrams using is and iz -bands.Quasar candidates (gray crosses), SDSS quasars (blue squares),quasar redshift tracks (black lines with asterisks), model browndwarfs (green triangles), stars (green squares), star forming galaxyredshift tracks (blue lines), passive galaxy redshift tracks (redlines), and the two selection boxes are plotted. We plotted ourtwo new quasars with the is and iz photometry using red circlesin (a). we employed an additional selection method: photome-try from is / iz -bands.We now apply selection cuts using the is and iz -bandsof CQUEAN. The color cuts were defined using quasarredshift tracks. We optimized our quasar selection using CQ Cut1 ( r − is − iz : selection method A) or CQ Cut2 ( is − iz − J : selection method B) on color-color diagrams,which explore the redshift ranges of 4.60 ≤ z ≤ ≤ z ≤ CQ Cut1 ( r − is − iz for 4.60 ≤ z ≤ r − is > . ∩ [ is − iz < . ∩ [ is − iz < . × ( r − is ) − . CQ Cut2 ( is − iz − J for 5.50 ≤ z ≤ s − iz > . ∩ [ iz − J < . MS III IMS J153541.19+034725.9 µ m]20406080100120140 F ν [ µ J y ] u g r i z Y J H K isiz (a) IMS J102201.90+080122.2 µ m]20406080100120140 F ν [ µ J y ] u g r i z Y J H K isiz (b) Figure 6.
Examples of SEDs of u, g, r, is, i, iz, z, Y, J, H, and K -bands. The filter names are marked at each wavelength. The is and iz filters are plotted with green points. (a): A candidate withblue H − K color ( H − K = − . H − K = 0 . tracks (blue line; model colors from M51), passivegalaxy redshift tracks (red line; model colors from theBruzual & Charlot (2003) model of a passively evolving5 Gyr-old galaxy with spontaneous burst, metallicity ofZ = 0.02, and the Salpeter initial mass function), andSDSS quasars with is and iz observations for comparison(blue square). The two color cuts are denoted. About1,400 among ∼ CQ Cut1 still show a high contaminationrate because the stellar locus is found near the quasarredshift track. After considering the spectral shape ofquasars, we selected about 60 targets as promising can-didates via visual inspection of SEDs, because quasars at5 < z < H − K colors redder than thoseof dwarf stars ( H − K &
0) due to the power-law con-tinuum of quasars. During the visual inspection, SEDsthat show a turn down in flux toward longer wavelengths(Figure 6a) are rejected in comparison to those that areretained as candidates (Figure 6b).
Ancillary Selection
We set additional selection criteria for assigning priori-ties for imaging and spectroscopic follow-up observations.
WISE Selection:
The
W ISE catalog provides 3.4,4.6, and 12 micron data ( W W
2, and W ∼ − . < K − W < . W − W > − . W − W > − . W ISE bands and assigned high priorities to thesesources for follow-up observations. Figure 7 shows our ∼ W ISE detections (gray crosses),9 previously discovered quasars with
W ISE detections(blue squares), and model brown dwarfs (green trian-gles). Since the model brown dwarf templates fromBurrows et al. (2006) do not extend to the W WI Cut (purple boxes). We adopt the following selec- -1.0 -0.5 0.0 0.5 1.0W1 - W2 [AB]-1012 K - W [ A B ] (a) IMS QuasarsModel Brown DwarfsQuasars in LiteratureSelection Box (WI_Cut)IMS QuasarsModel Brown DwarfsQuasars in LiteratureSelection Box (WI_Cut) -1 0 1 2 3W1 - W3 [AB]-1012 K - W [ A B ] (b) Figure 7.
Two color-color diagrams with
W ISE photometry andour selection boxes. We plot our ∼ W ISE detections (gray crosses), previously known z ∼ tions: WI Cut : [ W − W > − . ∩ [ − . < K − W < W − W > − . ∩ [ − . McGreer et al. (2013) discovered a number of quasars at4.7 < z < z < . ∼ W ISE selection, but not included inthe r − i − z − J − K color cuts, are also added to ourcandidate list. Candidates from Polsterer et al. (2013): Polsterer et al. (2013) provide a quasar candidate cat- Jeon et al. Table 1 Priorities for CQUEAN imaging follow-up observationsPriority Stellarity WISE McGreer+13 or Polsterer+13 Number0 yes yes yes 81 yes no yes 242 yes yes no 453 yes no no 1,0394 yes or yes or yes 1,1425 no yes 12310 others 1,105 alog containing 121,909 sources with their photometricredshifts at 2.558 ≤ z ≤ Stellarity: We use mergedClass for UKIDSS LASand type for SDSS to distinguish point sources fromextended sources. We defined that a source with mergedClass = − − 2, or type = 6, is a point source,and gave higher priorities to these sources. We did notexclude the extended sources because 17% of the dis-covered quasars from McGreer et al. (2013) are classifiedas extended sources in their i -band, meaning that somequasars may be classified as extended sources. Selection Summary The selection method used in this paper can be sum-marized as the following. We begin with an adjoint sam-ple of SDSS DR8 and UKIDSS LAS DR10. We selectobjects showing Ly α drops between r and i , and removebrown dwarfs and stars using the r − i − z − J − K color-color diagrams ( SU Cut1,2 ). To decrease the num-ber of stellar contaminants, we adopt magnitude cuts inthe u , g , and z bands ( SU Cut3,4 ). These four criteriadecrease the sample to ∼ WI Cut ) are listed as promising candidates. Objects notincluded in the r − i − z − J − K selection, but selectedfrom the McGreer et al. (2013) cuts with WISE selec-tion ( WI Cut ), are added to the candidate list. Amongthe ∼ r − is − iz and is − iz − J ,employing our new filter system and selected quasar can-didates at two redshift ranges ( CQ Cut1,2 ). For theCQUEAN imaging follow-up observations, we set priori-ties of our candidates considering the stellarity, the WISEdetection, the color cuts from McGreer et al. (2013), andcandidates from Polsterer et al. (2013). Objects showingpoint-like shapes with WISE detections as well as satis-fying the color cuts from McGreer et al. (2013) or candi-date list from Polsterer et al. (2013) were classified as theimportant candidates. Table 1 lists the priority for eachcase, with smaller numbers indicating higher priorities.We have been conducting the is and iz imaging for thehigh priority objects and about half of the sample wasimaged in these two filters. Finally, via visual inspectionof the SEDs, ∼ 60 targets were selected to be our mainsamples for spectroscopy. Optical Imaging Follow-up Observations withCQUEAN Follow-up observations of our high redshift quasar can-didates using CQUEAN began in 2010 August and are still on-going. About 1,400 among ∼ iz and 60 sec for is filters, respectively. Number of framesvaried depending on the sky conditions, such as seeingconditions and extinction. If the peak value of a targetwas greater than 80 ADU after a 30 sec exposure with iz , 2.5 (30 sec × 5) and 5 (60 sec × 5) minutes wereused as the integration times for the iz and is filters,respectively. If the signal was lower than the criterion,we exposed 5 (30 sec × 10) and 10 (60 sec × 10) minuteswith iz and is , respectively, or more.Preprocessing including bias subtraction, dark sub-traction and flat fielding, were preformed usingthe usual data reduction procedures in the IRAF noao.imred.ccdred package. Since the bias values maychange with time (Park et al. 2012), we used bias im-ages that were taken closest to the object frames, time-wise. We combined images of each field and filter inaverage. We used the ccmap task of IRAF and SCAMP(Bertin 2006) to derive astrometric solutions. SExtractor(Bertin & Arnouts 1996) was used for the source detec-tion and photometry. We derived auto-magnitudes whichare taken as the total magnitudes.For the photometric calibration, we used SDSS pho-tometry of stellar objects inside each target field. We per-formed χ fitting to the SDSS r , i , z magnitudes of stellarsources, to determine best-fit stellar spectral types. Forthis, we used the SED templates from Gunn & Stryker(1983), containing 175 spectra of various stellar types.The model is and iz magnitudes were calculated fromthe best-fit templates and these are used to define thezero-points ( Zp ) of each filter image of each field. The Zp values were calculated for each star, and we took theaverage of these values as Zp and the standard deviationof the scatters as its Zp error. The average Zp erroris about 0.05 mag. During the calculation, objects withlarge reduced χ values ( χ ν > 5) were rejected for the es-timation. Note that this photometric calibration methodis described in more detail in Jeon et al. (2016). Optical Spectroscopic Follow-up Observations We observed 47 candidates using the Kitt Peak Na-tional Observatory (KPNO) 4-m Mayall telescope andthe European Southern Observatory (ESO) New Tech-nology Telescope (NTT). The KPNO 4-m observationswere performed over three runs for 10 nights from 2013 IRAF is distributed by the National Optical Astronomy Ob-servatory, which is operated by the Association of Universities forResearch in Astronomy, Inc., under cooperative agreement withthe National Science Foundation. MS III Table 2 Spectroscopic observation summary of IMS quasarsSpectroscopy Date Telescope Target Integration Time (min) Slit Width ( ′′ )Optical 2013 Jan. 16 KPNO 4-m IMS J1022+0801 80 3.02013 May 6 NTT IMS J1437+0708 40 1.22013 May 6 NTT IMS J2225+0330 90 1.02013 May 7 NTT IMS J1437+0708 60 1.02013 Sep. 27 KPNO 4-m IMS J0122+1216 45 1.52013 Sep. 28 KPNO 4-m IMS J0155+0415 60 1.52013 Sep. 28 KPNO 4-m IMS J0324+0426 45 1.52013 Sep. 29 KPNO 4-m IMS J2225+0330 60 1.52013 Sep. 29 KPNO 4-m IMS J0122+1216 45 1.5NIR 2014 Oct. 6 Magellan IMS J0122+1216 60 1.02014 Oct. 7 Magellan IMS J0155+0415 30 1.02014 Oct. 6 Magellan IMS J0324+0426 60 1.02015 Aug. 30 Gemini-N IMS J2225+0330 53 0.675 January to September, and the NTT observation wasdone for 3 nights in 2013 May.For the observations at KPNO, we used the Ritchey-Chr´etien Focus Spectrograph in a longslit mode (RC-SPL ) with a LB1A CCD, the BL400 grating of R ∼ ′′ slit, and OG400 filter. LB1A uses a thick CCDchip, therefore it does not suffer much from fringing. Thewavelength coverage is 5,000˚ A – 10,000˚ A . For the obser-vation at the ESO NTT, we used the ESO Faint ObjectSpectrograph and Camera v.2 (EFOSC2; Buzzoni et al.1984). The EFOSC2 was used with Gr A – 11,000˚ A and R ∼ ′′ slit. We took calibration frames including bias,dark, flat, and arc. Standard stars such as G191B2B,GD153, CD-32d9927, LTT7379, LTT3864, Feige110, andHR7596 were observed for the flux calibration. The slitwidths varied from 1 . ′′ . ′′ 0, depending on the see-ing conditions. Table 2 shows the summary of the opti-cal spectroscopic observations of the discovered quasars,namely the total integration time and the slit width foreach target.We followed the typical steps for preprocess-ing, including bias subtraction, dark subtrac-tion, and flat fielding, for each science image,standard star image and arc image, using the noao.imred.ccdred package in IRAF. The spectrawere extracted using the noao.imred.kpnoslit or the noao.twodspec.apextract packages in IRAF for eachsingle image. We used an optimal aperture size for eachimage where the S/N is highest. After this, wavelengthand flux calibrations were conducted. The spectra wereflux-calibrated using spectra of the standard stars. Con-sidering the light loss due to variable seeing conditions,we scaled the spectra using broadband photometry.We chose i -band for this calibration, since we get thehighest S/N in this band for the observed spectra.The flux-calibrated spectra were combined in medianusing the scombine task of IRAF and were correctedfor Galactic extinction using values from Cardelli et al.(1989) and Schlegel et al. (1998).We observed 47 candidates and 6 of them turned outto be high redshift quasars at 4.7 ≤ z ≤ quasars is IMS JHHMMSS.SS ± DDMMSS.S in J2000.0coordinates (IMS JHHMM ± DDMM for brevity). NIR Spectroscopic Observation To measure their black hole masses and Edding-ton ratios, we observed four of the six newly discov-ered quasars using the Folded-port InfraRed Echellette(FIRE ) spectrograph on the Magellan telescope (IMSJ0324+0426, IMS J0122+1216, and IMS J0155+0415)and using the Gemini Near Infra-Red Spectrograph(GNIRS) on the Gemini North (Gemini-N) telescope(IMS J2225+0330; program GN-2015B-Q-77). Table 2shows the summary of the Magellan and Gemini-N ob-servations.In the Magellan/FIRE observation, we used a slitwidth of 1 . ′′ 00 with the Echelle mode (R = 3,600). TheABBA pointing method was used for the sky subtrac-tion between exposures. We observed standard stars foreach target. Data for the flat fielding and the wavelengthcalibration were also taken. The data reduction was con-ducted using the IDL suite, FIREHOSE . This pipeline con-ducts the preprocessing, object extraction, telluric cor-rection, flux calibration, and spectra combining.In the Gemini-N/GNIRS observation, we used thecross-dispersed (XD) mode with the 32 line mm − grat-ing, the short blue camera, and its SXD prism. Adoptingthe slit of 0 . ′′ 675 width, we obtained R ∼ . The steps include pat-tern noise cleaning using the clearnir script, reducingthe science data using flatfield images, combining images,wavelength calibration, extracting spectra, and flux cal-ibration using standard stars.We scaled the flux of the combined spectra usingbroadband photometry. After that, the spectra were cor-rected for Galactic extinction using Cardelli et al. (1989)and Schlegel et al. (1998). HIGH REDSHIFT QUASARS Photometric Properties We list the photometric information from SDSS,UKIDSS LAS, W ISE , and CQUEAN of our newly dis- http://web.mit.edu/ ∼ Jeon et al. Table 3 General information of IMS quasarsName R.A. and Dec. (J2000.0) Redshift M IMS J032407.70+042613.3 03:24:07.70+04:26:13.3 4.70(Ly α ) a , 4.68(C iv ), 4.73(Mg ii ) − ± α ) b , 4.81(C iv ) − ± α ) c − ± α ) d , 5.26(Mg ii ) − ± α ) − ± α ) e , 5.27(C iv ) − ± Note . — z spec from other papers are all derived from Ly α a z spec =4.72 from Wang et al. (2016) b z spec =4.76 from Yi et al. (2015) and z spec =4.79 from Wang et al. (2016) c z spec =4.93 from Wang et al. (2016) d z spec =5.24 from Wang et al. (2016) e z spec =5.37 from Wang et al. (2016) Table 4 Optical photometric information of IMS quasarsName g r i z is iz IMS J0324+0426 23.95 ± ± ± ± · · · · · · IMS J0122+1216 24.29 ± ± ± ± · · · · · · IMS J1437+0708 25.02 ± ± ± ± ± ± ± ± ± ± · · · · · · IMS J1022+0801 25.23 ± ± ± ± ± ± ± ± ± ± · · · · · · Table 5 NIR photometric information of IMS quasarsName W W W W Y J H K IMS J0324+0426 18.47 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Note . — We used a dummy value of 99.99 for non-detections. Table 6 Selection methods of IMS quasarsName WISE a WISE b McGreer+13 c Polsterer+13 d r − is − iz e is − iz − J f ( K − W − W 2) ( K − W − W · · · · · · IMS J0122+1216 yes yes yes no · · · · · · IMS J1437+0708 yes yes yes yes yes · · · IMS J2225+0330 yes yes no no · · · · · · IMS J1022+0801 yes yes no no yes · · · IMS J0155+0415 yes yes no no · · · · · · a Does it satisfy the color cut of K − W − W b Does it satisfy the color cut of K − W − W c Does it satisfy the color cuts from McGreer et al. (2013)? d Is it contained in the candidate list from Polsterer et al. (2013)? e Does it satisfy the color cut of r − is − iz ? f Does it satisfy the color cut of is − iz − J ? MS III W ISE detec-tions and are located inside the W ISE color cuts (Figure7; K − W − W K − W − W < ± α emission line (Section 3.2). For the two IMSquasars with is and iz photometry, Figure 5a shows theircolors in the r − is − iz color-color diagram.Only two quasars among ∼ is and iz photometry were newly identified as high redshiftquasars in the r − is − iz color-color diagram, and noneof our candidates were discovered in the is − iz − J color-color diagram. The other quasars were selected as can-didates using the W ISE photometry or the color cutsfrom McGreer et al. (2013). The expected numbers ofquasars for each selection method from 3,400 deg are24 . +67 . − . for 4.60 ≤ z ≤ . +15 . − . for 5.50 ≤ z ≤ ≤ z ≤ ≤ z ≤ Spectroscopic Properties First, we present the optical spectra of the 6 quasars at4.7 ≤ z ≤ α shapes for the six quasars. IMSJ0324+0426, IMS J0122+1216, and IMS J1437+0708show strong Ly α emission, while the other three showsmoother shapes. These weak Ly α lines are fairlycommon at high redshift. Jiang et al. (2009) andBa˜nados et al. (2014) show that a significant fraction ofquasars at high redshift have weak Ly α (e.g., 25% of z ∼ α are dif-ficult to verify due to the imperfect sky line subtractionand low QE of the detector at wavelengths longer than0.8 µ m. IMS J0122+1216 shows significant deep absorp-tion features and we classify it as a broad absorption line(BAL) quasar. This property can be noticed in its NIRspectrum more clearly.Table 3 lists the redshifts and absolute magnitudes ofthe continua at rest-frame 1450˚ A ( M ) of the quasars. µ m] F ν [ µ J y ] IMS J032407.70+042613.3 (z Lya =4.70) Lyr Lyb Lya N V OI Si IV + O IV] C IV µ m] F ν [ µ J y ] IMS J012247.33+121623.9 (z Lya =4.83) BAL µ m] F ν [ µ J y ] IMS J143704.82+070808.3 (z Lya =4.94) µ m] F ν [ µ J y ] IMS J222514.39+033012.6 (z Lya =5.35) µ m] F ν [ µ J y ] IMS J102201.90+080122.2 (z Lya =5.36) µ m] F ν [ µ J y ] IMS J015533.28+041506.8 (z Lya =5.35) Figure 8. Optical spectra of the 6 quasars at 4.7 < z < The redshifts of IMS J2225+0330 and IMS J1022+0801were measured from the Ly α emission lines by fittingGaussian profiles. However other spectra show a sharpdrop bluewards of Ly α . In these cases, their redshiftswere measured by fitting the spectra (the orange linein Figure 8) from the redshifted and IGM-attenuatedSDSS composite quasar template. The redshift errorsestimated from these optical spectra contain the uncer-tainties from the spectral resolution of each instrument(typically ∼ iv or Mg ii emission lines from theNIR spectra (see section 5) in Table 3. The redshift errorestimated from the NIR spectra due to the spectral reso-lution is about 0.002 for Magellan/FIRE and about 0.007for Gemini-N/GNIRS. The redshifts estimated from theoptical spectra and the NIR spectra show discrepancies,and we believe that this is caused by the ambiguous Ly α shapes, which can be heavily affected by the Ly α forestand the blending with the N v emission line.Richards et al. (2009, 2015) provide photometric red-shifts (z phot ) for three out of the six quasars, IMSJ0122+1216, IMS J1437+0708, and IMS J2225+0330.0 Jeon et al. µ m] F λ [ - e r g s - c m - A - ] CIV CIII] MgII IMS J032407.70+042613.3Magellan/FIRE µ m] F λ [ - e r g s - c m - A - ] CIV CIII] MgII IMS J012247.33+121623.9Magellan/FIRE µ m] F λ [ - e r g s - c m - A - ] CIV CIII] MgII IMS J222514.39+033012.6Gemini-N/GNIRS µ m]051015202530 F λ [ - e r g s - c m - A - ] CIV CIII] MgII IMS J015533.28+041506.8Magellan/FIRE Figure 9. NIR spectra of IMS quasars from Magellan/FIRE andGemini-N/GNIRS, smoothed to the instrumental resolution. Theblue lines denote the errors of the spectra and the red vertical linesindicate the locations of emission lines at the redshift determinedfrom optical spectra. For the spectrum from Gemini-N/GNIRS,the gray bars show regions of strong atmospheric absorption. Their estimate for IMS J0122+1216 (z phot = 5 . +0 . − . from Richards et al. 2015) does not agree with our spec-troscopic redshift (z Ly α = 4.83), while IMS J1437+0708(z phot = 5 . +0 . − . from Richards et al. 2009 or5 . +0 . − . from Richards et al. 2015) and IMSJ2225+0330 (z phot = 5 . +0 . − . from Richards et al.2015) are in agreement with our estimates (z Ly α = 4.94and 5.35, respectively). The discrepancy between z phot and z spec for IMS J0122+1216 is likely because the objectis a BAL quasar.We calculated the M values using the average fluxat 1440˚ A – 1460˚ A from the optical spectra in Table 3.The uncertainties were estimated from the rms contin-uum flux density. For z = 5.0 quasars, the observedwavelength of the rest-frame 1450˚ A is located at 8700˚ A ,where the sky emission lines are significant. Due to the difficulty of subtracting the sky from the relatively lowS/N spectra, these values are crude and the actual mag-nitude uncertainties could be higher than our error esti-mates. Our IMS quasars are within the M range of − . − . Individual Properties of Quasars IMS J0324+0426 (z Ly α =4.70, z CIV =4.68,z MgII =4.73): This quasar has a strong Ly α emissionline. It also shows relatively strong Lyman β (Ly β ), O i ,Si iv +O iv] , and C iv emission lines, and a weak N v emission line. In the NIR spectrum, C iv , C iii] , andMg ii emission lines are prominent. Wang et al. (2016)reported z=4.72. IMS J0122+1216 (z Ly α =4.83, z CIV =4.81): Weclassify this as a BAL quasar because of deep absorptionfeatures bluewards of Ly α , O i , Si iv +O iv] , and C iv lines. It has a strong Ly α emission line, and a weak Ly β emission line. We are not able to identify other emissionlines due to these deep absorptions. The NIR spectrumhas strong C iv , C iii] , and Mg ii emission lines. The leftside (shorter wavelengths) of these lines are severely ab-sorbed. Yi et al. (2015) analyzed this quasar and deriveda redshift of z=4.76 while Wang et al. (2016) reportedz=4.79. IMS J1437+0708 (z Ly α =4.94): Its spectrum wasobtained from NTT/EFOSC2 with R ∼ 130 and it hasthe highest S/N ratio among the optical spectra. How-ever it does not show any prominent emission lines exceptthe Ly α . Wang et al. (2016) reported z=4.93. IMS J2225+0330 (z Ly α =5.35 and z MgII =5.26): This source was observed by two telescopes, the KPNO4-m telescope and NTT, and the two spectra were com-bined in average. It has a smooth Ly α emission line anddoes not show any other emission lines. In the NIR spec-trum, the C iv , C iii] , and Mg ii emission lines are strongbut the C iv emission line has a rough shape due to thestrong atmospheric absorption. Wang et al. (2016) re-ported z=5.24. IMS J1022+0801 (z Ly α =5.36): This quasar hasthe weakest Ly α emission line among the six observedquasars. No other emission lines are visible due to lowS/N. This quasar was recently discovered independentlyby Yang et al. (2017), reporting the spectroscopic red-shift of z = 5 . IMS J0155+0415 (z Ly α =5.35, z CIV =5.27): Theoptical spectrum shows a weak Ly α emission line andother emission lines are not detected. In the NIR spec-trum, it has prominent Si iv +O iv] , C iv , and C iii] emis-sion lines. The Mg ii emission line is hidden due to tel-luric absorption. Wang et al. (2016) reported z=5.37. SELECTION COMPLETENESS To calculate the expected number of quasars for eachselection method, we derived the quasar selection com-pleteness, which can be affected by various effects. Thecompleteness from color selection is defined as the frac-tion of quasars inside specific color cuts among allquasars within specific redshift and magnitude bins.First, applying various quasar templates, we calculatedthe completeness using the fraction of quasars that fallwithin each selection box, as a function of redshift and M (Section 4.1). Then we apply this completeness MS III Completeness from Color Cuts To measure the fraction that a quasar with a given red-shift, M , and intrinsic SED meets our selection cri-teria, we follow approaches from previous studies (e.g.,Willott et al. 2005; Venemans et al. 2013). The compos-ite quasar template from Vanden Berk et al. (2001) isredshifted to various values, assuming that the spectralproperties of quasars do not evolve significantly with red-shift (e.g., Kuhn et al. 2001; Fan et al. 2004; Jun et al.2015), except wavelengths blueward of the Ly α line.Fluxes in these shorter wavelengths are absorbed by neu-tral hydrogen (H i ) in the IGM, and the absorption be-comes stronger toward higher redshift because the frac-tion of H i increases with redshift (Gunn-Peterson effect;Gunn & Peterson 1965). We applied this attenuation ef-fect to our redshifted spectra using the IGM attenuationmodel of Madau et al. (1996). We redshifted the spec-trum to 4 ≤ z ≤ M in the range − < M < − 20 with steps of∆ M = 0 . 5. Then we calculated model magnitudesfor each band.The most important factor in the observed color distri-bution is the continuum slope of quasars. We considered13 cases of models for each redshifted spectrum with con-tinuum slopes of − . ≤ α ν ≤ − . F ( ν ) ∝ ν α ν )with steps of ∆ α ν = 0 . 1. This range was derived basedon the range of α ν values from the SDSS DR12 quasarcatalog (Pˆaris et al. 2016) that includes about 230,000quasars with a mean value of α ν = − . σ dis-persion of 0.6 (68.3% confidence level). De Rosa et al.(2014) analyzed a sample of four quasars at z > α ν range.We also considered variable rest-frame equivalent widths(EW ) of the Ly α emission line: 8 cases of 50 ≤ EW ≤ 85 with steps of ∆EW = 5 (Fan et al. 2001). In total,we generate a database of 104 model quasars of which thecontinuum slopes and Ly α EWs are uniformly sampledwithin given ranges and calculate the average selectedfraction as a function of redshift and M .Figure 10a shows the completeness distribution as afunction of redshift and M , for the selection usingthe r − i − z − J − K and r − is − iz color-color dia-grams (selection method A), and Figure 10c shows thecompleteness distribution when using r − i − z − J − K and is − iz − J color-color diagrams (selection methodB). In Figures 10b and 10d, we plot the completeness asa function of redshift for the two methods, for the caseof M = − 29. The completeness in Figure 10b risessteeply from 0% to 100% between z = 4.60 and z = 4.70,remains at 100% up to z = 5.15, and drops below 80%for z > ≤ z ≤ ≤ z ≤ M > − . M limit corresponds to our magnitude cut, z < . M of rizJK and r-is-iz -30 -29 -28 -27 -26 -25M z Completeness (%) (a) M =-29 C o m p l e t e n e ss (b) rizJK and is-iz-J -30 -29 -28 -27 -26 -25M z Completeness (%) (c) M =-29 C o m p l e t e n e ss (d) Figure 10. (a): Completeness as a function of redshift and M for r − i − z − J − K and r − is − iz selection. The red boxes indicatethe redshifts and M of our six new quasars. (b): Complete-ness as a function of redshift from (a) when M = − 29. (c):Completeness for r − i − z − J − K and is − iz − J selection.(d): Completeness from (c) when M = − 29. The colors of thecontours indicate 0% and 100% completeness for white and black,respectively. our six newly discovered quasars (Table 3) with red boxesin Figure 10a. Expected Quasar Number from Our Surveys We calculated the expected number of quasars fromour survey by extrapolating the luminosity function of z ∼ kz factor that accounts for the decline in num-ber density as a function of redshift. We adopted twovalues of k : k = − . 47 from Willott et al. (2010a) and k = − . 71 from McGreer et al. (2013). Then, we extrap-olated the luminosity function of z ∼ M limit (column4). The expected number of quasars for each quasarselection are listed in columns 5 and 6 for the case of k = − . 47 and k = − . 71, respectively, with the 1 σ errors caused by the uncertainties in break magnitude M ∗ and bright end slope β from Willott et al. (2010a).We only considered the completeness from our color cuts,and assumed that the efficiency of each selection in itsredshift range (column 3) and the M limit (column5) is 100%.Our quasar survey discovered 20 quasars including 6new quasars at 4.60 ≤ z ≤ ≤ z ≤ < z < < z < W ISE photometry (theyare fainter than quasars at 4.60 ≤ z ≤ PHYSICAL PROPERTIES OF QUASARS In this section, we present the physical properties offour IMS quasars, IMS J0324+0426, IMS J0122+1216,IMS J2225+0330 and IMS J0155+0415, based on the2 Jeon et al. Table 7 Expected number of quasars from our surveySelection Method Area (deg ) Redshift Range M Limit Expected Number Selected Number(1) (2) (3) (4) (5) a (6) b (7) r − is − iz − . . +67 . − . . . − . is − iz − J − . . +15 . − . . +19 . − . a For k = − . b For k = − . data obtained with optical and NIR spectroscopy. Inour NIR spectra, we identified both the C iv and Mg ii lines for IMS J0324+0426 and IMS J0122+1216, only theMg ii line for IMS J2225+0330, and only the C iv linefor IMS J0155+0415. After modeling the continuum andemission lines of C iv and Mg ii , we estimated contin-uum slopes α ν (where α ν is for F ( ν ) ∝ ν α ν ), line widths(full width at half maximum; FWHM), continuum lu-minosities at the rest-frame wavelengths of 1350˚ A and3000˚ A ( λL λ (1350) and λL λ (3000)) for each emission line(Section 5.1). From these measurements, we calculatedthe black hole mass ( M BH ) from the C iv emission line( M BH , CIV ) or from the Mg ii emission line ( M BH , MgII )through different relations from McLure & Jarvis (2004),Vestergaard & Peterson (2006), and Jun et al. (2015)(Section 5.2). For the virial factor in these blackhole mass estimators, we adopted f = 5 . ± . Analysis of NIR Spectra We modeled the quasar NIR continuum assuming twocomponents, a power law component and a componentthat describes the pseudo-continuum due to the blendedforest of Fe ii emission lines as given below: F ( λ ) = a × λ α λ + b × FeII( λ, v ) (1)where α λ is the continuum slope (in this case, α ν = − α λ − F ( λ ) ∝ λ α λ ), and v and b are the width andstrength of Fe ii templates. We used two Fe ii templatesfrom Vestergaard & Wilkes (2001) and Tsuzuki et al.(2006). A scaled and broadened Fe ii template was usedfor modeling the Fe ii emissions from our spectra. In thecase of the C iv emission line, only Vestergaard & Wilkes(2001) provide the Fe ii template in this wavelengthrange. We modeled the two components simultaneously.The quality of the continuum subtraction depends onthe determination of the continuum fitting ranges. Weselected narrow fitting windows which minimize the con-tributions from other components. Since the qualitiesof the C iv emission line in the IMS J0324+0426 spec-trum and the Mg ii emission line in the IMS J2225+0330spectrum are not sufficient to constrain the Fe ii emis-sions, we failed to find the Fe ii component. Since IMSJ0122+1216 shows significant broad absorption featuresbluewards of the C iv and Mg ii emission lines, we nar-rowed the fitting window ranges to exclude the absorp-tion part.Since most of the uncertainties in the continuum slopeor the line width result from the fitting range of the con-tinuum modeling, we adopted 36 different fitting rangeswithin the given wavelength windows and performedmodel fitting for each different sub-wavelength range to calculate the uncertainties. Since we cannot vary thecontinuum fitting range of C iv of IMS J0122+1216, weset the uncertainty of this line width as 5% of the linewidth instead of the uncertainty derived from the variouscontinuum ranges. This fraction is identical to the ratioof line widths and their uncertainties, for all other lines.After subtracting the best-fit continuum from eachspectrum, we fit the C iv and Mg ii emission lines. Weused single and double-Gaussian profiles considering thepresence of asymmetric profiles characterized by red orblue wings. For the fitting ranges, we set 1500˚ A –1600˚ A for the C iv line and 2700˚ A – 2900˚ A for theMg ii line, except for the C iv of IMS J0122+1216,which is affected by broad absorption. In this case,we set the fitting range to 1530˚ A – 1590˚ A . The Mg ii lines of IMS J0324+0426 and IMS J0122+1216 are wellfit by double-Gaussian profiles due to their asymmetricshapes, whereas the other lines can be fit using a single-Gaussian profile. One of the double-Gaussian compo-nents of IMS J0324+0426 is a narrow line (violet line inFigure 11b) with FWHM = 800 ± 40 km s − . To ob-tain the line width FWHM, the measured FWHM obs was corrected for the instrumental resolution FWHM ins :FWHM = p (FWHM obs ) − (FWHM ins ) .We used an IDL procedure, mpfit.pro to find the best-fit models to the observed spectra that uses the χ min-imization method for both the continuum and the emis-sion line. We included 1 σ errors of the spectra for eachfitting. From the best-fit model, we obtained the best-fitestimates for each parameter, such as the power law slopeand the line width. The uncertainties for each parameterwere calculated as follows. The error for each parameteris dominated by the scatter of the various best-fits whenaltering the fitting range for the continuum. We com-pared the best-fit parameters for each trial and we setthe average and standard deviation of the values as thebest-fit parameter and its error. Ultraviolet Luminosity and M BH In Figure 11, the best-fit continuum and emission linemodels are shown for each emission line. In Table 8we list the best-fit estimates of the power law slope( α ν, CIV and α ν, MgII ) and the line width (FWHM CIV and FWHM MgII ) and their errors for each emission line.There is no significant difference in the derived powerlaw slope and line width parameters when using differentFe ii templates from Vestergaard & Wilkes (2001) andTsuzuki et al. (2006). Note that the IMS 2225+0330spectrum has low S/N and the uncertainty of the linewidth estimated using the method in Section 5.1 is un-derestimated. The 1 σ error from the Gaussian fitting isabout 15%.The power law slopes of quasars vary significantly be- MS III IMS J032407.70+042613.3 (CIV) F λ [ - e r g s - c m - A - ] Single Gaussianz CIV =4.68 (a) IMS J032407.70+042613.3 (MgII) F λ [ - e r g s - c m - A - ] Double Gaussianz MgII =4.73 (b) IMS J012247.33+121623.9 (CIV) F λ [ - e r g s - c m - A - ] Single Gaussianz CIV =4.81 (c) IMS J012247.33+121623.9 (MgII) F λ [ - e r g s - c m - A - ] Double Gaussian (d) IMS J222514.39+033012.6 (MgII) F λ [ - e r g s - c m - A - ] Single Gaussianz MgII =5.26 (e) IMS J015533.28+041506.8 (CIV) F λ [ - e r g s - c m - A - ] Single Gaussianz CIV =5.27 (f) Figure 11. The best-fit continuum and emission line modeling for the C iv and Mg ii emission lines of IMS quasars. In each panel, thespectrum (black) with errors (gray) is overplotted with the best-fit model (red), which consists of the power-law component (green), thebest-fit Fe ii template (blue; we used the Fe ii template from Vestergaard & Wilkes 2001 as an example), and each emission line (magenta).Estimated redshift from each emission line is denoted except for the Mg ii emission line of IMS J0122+1216. (a): C iv emission line ofIMS J0324+0426. We cannot find a solution for the Fe ii template fitting. (b): Mg ii emission line of IMS J0324+0426. We used adouble-Gaussian model for the line fitting (violet and orange lines) and one of the double-Gaussian components is a narrow line (violetline). (c): C iv emission line of IMS J0122+1216. (d): Mg ii emission line of IMS J0122+1216. Two Gaussian components (two orangelines) are used to fit the line shape. (e): Mg ii emission line of IMS J2225+0330. We cannot find a solution for the Fe ii template fitting.(f): C iv emission line of IMS J2225+0330. Table 8 Power-law slopes, line widths, and continuum luminosities estimated from the NIR spectraName α ν, CIV α ν, MgII FWHM CIV FWHM MgII λL λ (1350) λL λ (3000)(km s − ) (km s − ) (10 erg s − ) (10 erg s − )IMS J0324+0426 1 . ± . − . ± . 78 6070 ± 300 2660 ± 280 6 . ± . 22 3 . ± . · · · − . ± . 31 6240 ± 310 4210 ± 160 5 . ± . 08 6 . ± . · · · . ± . · · · ± · · · . ± . − . ± . · · · ± · · · . ± . · · · tween sources. For example, Davis et al. (2007) found − . < α ν < < z < < z < < z < − < α ν < > − . < α ν < λL λ (1350) and λL λ (3000) in Table 8 are alsocalculated from the optical and NIR spectra. ForIMS J0324+0426, we used the optical and NIR spec-tra for the λL λ (1350) and λL λ (3000), respectively. The λL λ (1350) of IMS J0155+0415 and the λL λ (3000) ofIMS J0122+1216 were estimated from their NIR spec-tra. Since the continuum spectra near the 3000˚ A of IMSJ2225+0330 show low S/N due to the strong atmosphericabsorption, we used fit spectra using the redshifted SDSScomposite quasar template. In the case of the λL λ (1350)of IMS J0122+1216, the continuum near 1350˚ A shows deep drops in its optical spectrum. Therefore, we usedthe fit spectrum when we estimated the redshift in Sec-tion 3.2. The λL λ (1350) and λL λ (3000) were calculatedfrom the average flux in the 1340˚ A – 1360˚ A and 2950˚ A – 3050˚ A , respectively. The uncertainty in the contin-uum luminosity was estimated from the scatter on thecontinuum flux in each window.In Table 9, we list the virial black hole mass estimatesobtained from C iv and Mg ii emission lines ( M BH , CIV and M BH , MgII ) using relations presented in Jun et al.(2015). The uncertainties of the masses propagate fromthe uncertainties of the FWHM and the continuum lu-minosity. The Eddington luminosities ( L Edd ) estimatedfrom the two mass estimators are listed in Table 9. Com-paring the two mass estimates ( M BH , CIV and M BH , MgII )for IMS J0324+0426 and IMS J0122+1216, M BH , CIV islarger than M BH , MgII by 0.8 dex and 0.2 dex, respec-tively. We note that M BH values from C iv show a4 Jeon et al. Table 9 M BH , L Bol , L Edd , and Eddington ratiosName M BH , CIV M BH , MgII L Bol (1350) L Bol (3000) L Edd (CIV) L Edd (MgII) Edd. ratio Edd. ratio(10 M ⊙ ) (10 M ⊙ ) (10 erg/s) (10 erg/s) (10 erg/s) (10 erg/s) (1350, CIV) (3000, MgII)IMS J0324+0426 7.60 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± · · · ± · · · ± · · · ± · · · ± ± · · · ± · · · ± · · · ± · · · larger scatter with respect to those from Mg ii or H β /H α (e.g., Jun et al. 2015; Karouzos et al. 2015). For exam-ple, the intrinsic scatters of the M BH , CIV and M BH , MgII from Jun et al. (2015) is 0.40 dex and 0.09 dex, respec-tively. Therefore the large discrepancy between M BH , CIV and M BH , MgII can be understood as a result of thelarge scatter in M BH , CIV estimators. Hence, we takethe Mg ii based values to be more reliable. The M BH values are roughly consistent with each other, when us-ing different estimates (e.g. McLure & Jarvis (2004) orVestergaard & Peterson (2006)) that use the same emis-sion line, within the error bars and the intrinsic scatterin the M BH estimators. Accretion Rate of Newly Discovered Quasars Bolometric luminosities ( L Bol ) and Eddington ratiosare given in Table 9, where L Bol are computed from λL λ (1350) and λL λ (3000) by multiplying 3.81 and 5.15,respectively (Shen et al. 2008).For IMS J0122+1216, the L Bol values that are calcu-lated from λL λ (1350) and λL λ (3000) do not agree witheach other. Since the λL λ (1350) is estimated from thebest-fit model spectrum, we adopt λL λ (3000) as morereliable. In the case of IMS J0324+0426, the L Bol (1350)has a larger uncertainty due to significant contaminationfrom sky emission lines.The Eddington ratios from M BH , CIV and λL λ (1350)are smaller by a factor of a few than those using M BH , MgII and λL λ (3000). The discrepancy is most likely causedby the difference in the derived M BH values. As we men-tioned earlier, C iv -based M BH values are in general moreuncertain than Mg ii -based values, and therefore we con-sider Mg ii -based Eddington ratios to be more reliable.Figure 12 shows M BH as a function of L Bol . Tocompare our sources with low redshift quasars, weused the SDSS samples of quasars (Shen et al. 2011).Quasars with M BH , MgII information were selected andthey cover a redshift range of 0.35 < z < ∼ ∼ ∼ M BH values are derived using Mg ii estimators. The Ed-dington ratios, L Bol /L Edd = 0.01, 0.1, and 1, are indi-cated with black solid lines. Our sources are plotted withthe red filled circles from M BH , MgII and L Bol (3000) ex-cept IMS J0155+0415. We can see that the high redshiftsample occupies a region of the parameter space differentfrom that of the low redshift sample with similar L Bol :the Eddington ratios of these high redshift quasars aresignificantly larger than those of the low redshift sample.In particular, our high redshift quasars have Eddingtonratios around 1, suggesting that these quasars are grow- ing vigorously. The Eddington ratio of IMS J0155+0415is an exception, because it was estimated from M BH , CIV and L Bol (1350), which are less reliable than M BH , MgII and L Bol (3000), respectively. Willott et al. (2010b) showsimilar results that the luminosity-matched quasar sam-ples at z = 2 and z = 6 have different Eddington ratiodistributions. However, to compare the Eddington ratiodistribution of low redshift quasars to their high redshiftcounterparts, less luminous samples will be needed. In-trinsic Eddington ratios of normal high redshift quasarscan be studied by discovering quasars from deeper sur-veys (e.g., Kashikawa et al. 2015; Kim et al. 2015) andEddington ratio distributions at high redshift when lessluminous quasars are included can be different (e.g., Kimet al. in preparation). SUMMARY We conducted a quasar survey at 5 . z . . r − i − z − J − K color cuts, then we exploited the W ISE colors to nar-row down the candidates. The candidates were also ob-served with the CQUEAN is and iz filters that over-come the limitations of previous filter systems. We thencarried out optical spectroscopic observations to con-firm our high redshift quasar candidates and discoveredsix new quasars. Four of them were observed by NIRspectroscopy to measure their physical properties ( M BH , L Bol , L Edd , and Eddington ratio) via spectral modelingof their continuum and emission lines. We comparedEddington ratios of our sources to those of low and highredshift quasars, and found that the Eddington ratio ofour quasars at z ∼ M < − > M ⊙ , and near-Eddington limit luminosities, support the scenario ofrapid growth of supermassive black holes in the earlyuniverse.This work was supported by the National ResearchFoundation of Korea (NRF) grant, No. 2008-0060544,funded by the Korean government (MSIP). The Gem-ini data were taken through the K-GMT Science Pro-gram (PID: KR-2015B-005) of Korea Astronomy andSpace Science Institute (KASI). Based on observationsobtained at the Gemini Observatory acquired throughthe Gemini Observatory Archive and processed using theGemini IRAF package, which is operated by the Asso-ciation of Universities for Research in Astronomy, Inc.,under a cooperative agreement with the NSF on behalfof the Gemini partnership: the National Science Foun-dation (United States), the National Research Coun-cil (Canada), CONICYT (Chile), Ministerio de Cien-cia, Tecnolog´ıa e Innovaci´on Productiva (Argentina), and MS III 44 45 46 47 48 49log L Bol (erg/s)7.07.58.08.59.09.510.010.5 l og M B H ( M O • ) L Edd Edd Edd IMS J0324+0426IMS J0155+0415 (CIV)IMS J0122+1216 (BAL)IMS J2225+0330 SDSS Quasars at 0.35 < z < 2.25Quasars at z ~ 7Quasars at z ~ 6Quasars at z ~ 5This study Figure 12. M BH as a function of L Bol . Red filled circles are our sources, purple crosses are quasars at z ∼ ∼ ∼ < z < L Bol /L Edd = 0.01, 0.1, and 1 are plotted with black solid lines. The names of thefour newly discovered quasars are written next to the red filled circles. Names with ’BAL’ and ’CIV’ are for the less reliable M BH values(either a BAL quasar, or M BH estimated from the C iv line.) Minist´erio da Ciˆencia, Tecnologia e Inova¸c˜ao (Brazil).This paper includes data taken at The McDonald Ob-servatory of The University of Texas at Austin. Basedon observations at Kitt Peak National Observatory, Na-tional Optical Astronomy Observatory (NOAO Prop.ID: 2012B-0537, 2013A-0506, 2013B-0534; PI: Y. Jeon),which is operated by the Association of Universitiesfor Research in Astronomy (AURA) under cooperativeagreement with the National Science Foundation. Theauthors are honored to be permitted to conduct astro-nomical research on Iolkam Du’ag (Kitt Peak), a moun-tain with particular significance to the Tohono O’odham.This paper includes data gathered with the 6.5 me-ter Magellan Telescopes located at Las Campanas Ob-servatory, Chile. M.H. acknowledges the support fromGlobal Ph.D. Fellowship Program through the NationalResearch Foundation of Korea (NRF) funded by the Min-istry of Education (NRF-2013H1A2A1033110). H.D.J issupported by an appointment to the NASA PostdoctoralProgram at the Jet Propulsion Laboratory, administered by Universities Space Research Association under con-tract with NASA. D.K. acknowledges the fellowship sup-port from the grant NRF-2015-Fostering Core Leaders ofFuture Program, No. 2015-000714 funded by the Koreangovernment. We thank the anonymous referee for usefulcomments, which improved the content of this paper. Facilities: Mayall (RCSPL), NTT (EFOSC2), Magel-lan:Baade (FIRE), Gemini:North (GNIRS)REFERENCES