Identifications of quasars at intermediate redshifts (2.2<z<3.5) are inefficient in most previous quasar surveys as their optical colors are similar to those of stars. The near-IR K-band excess technique has been suggested to overcome this difficulty
. Our study also proposed to use optical/near-IR colors for selecting z<4 quasars. To this method, we selected 105 unidentified bright targets with i<18.5 from the quasar candidates of SDSS DR6 with both SDSS ugriz optical and UKIDSS YJHK near-IR photometric data, which satisfy our proposed Y-K/g-z criterion and have photometric redshifts between 2.2 and 3.5 estimated from the 9-band SDSS-UKIDSS data. 43 of them were observed with the 2.16m telescope of NAOC in 2012. 36 of them were identified as quasars at 2.1<z<3.4. High success rate of discovering these quasars in the SDSS spectroscopic surveyed area demonstrates the robustness of both the Y-K/g-z selection criterion and the photometric redshift estimation technique. We also used the above criterion to investigate the possible star contamination rate to the quasar candidates of SDSS DR6, and found that it is much higher in selecting 3<z<3.5 quasar candidates than selecting lower redshift ones (z<2.2). The significant improvement in the photometric redshift estimation by using the 9-band SDSS-UKIDSS data than using the 5-band SDSS data is demonstrated and a catalog of 7,727 unidentified quasar candidates with photometric redshifts between 2.2 and 3.5 is provided. We also tested the Y-K/g-z selection criterion with the SDSS-III/DR9 quasar catalog, and found 96.2% of 17,999 DR9 quasars with UKIDSS Y and K-band data satisfy our criterion. With some samples of red and type II quasars, we found that 88% and 96.5% of them can be selected by the Y-K/g-z criterion respectively, which supports that using the Y-K/g-z criterion we can efficiently select both unobscured and obscured quasars. (abridged)
