A fluorescence resonance energy transfer (FRET) system between fluorescence carbon dots (CDs, donor) and Eosin B (EB, acceptor) was constructed, and a new method for sensitive and selective determination of pefloxacin (PEFL) was proposed. Fluorescent carbon dots (CDs) were synthesized by pyrolysis treatment using Setcreasea purpurea boom as carbon source. The carbon dots exhibited preferable dispersibility in water, high stability and a quantum yield of 3.7%. The as-prepared CDs were characterized using high-resolution transmission electron microscopy (HRTEM), powder X-ray diffraction patterns (XRD), and Fourier transform infrared spectroscopy (FTIR). Theresults showed that the CDs demonstrated an amorphous structure with the presence of —OH and —COOH groups on their surface. Using Frster’s theory of resonance energy transfer, the FRET between CDs and EB was determined, and the FRET system was constructed between CDs and EB. Some important factors were optimized, such as reaction medium and acidity, reaction time, the concentration of donor and acceptor, salt effect. According to the result, in PBS buffer solution at pH=3.0, the energy transfer from CDs to EB occurred with λex=340 nm, which resulted in an enhancement of the fluorescence intensity of EB. Upon the addition of pefloxacin, the fluorescence intensity of CDs was significantly increased because of the interaction between pefloxacin and CDs. Under the optimized experimental conditions, the change of CDs fluorescence intensity (ΔF) with the PEFL concentration exhibited a linear relationship in range of 0.016 8~6.71 μg·mL-1. The detection limit was 0.072 5 ng·mL-1 based on the formula 3s/k (n=11). Most of common substances such as cations (Fe3+, Al3+, Ca2+, Zn2+, Cr3+, Co2+, Cu2+, Mn2+, etc. ), anions Cl-, NO-3, I-, S2-, SCN-, SO2-4, Br-, NO-2, IO-3, F-, ClO-3, SO2-3, etc.), medicines (isoniazide, ascorbic acid and heparin sodium) and melamine did not interfere with the detection PEFL. The proposed method was applied to determine PEFL in drugs. The recoveries were 100.4%~105.1%, and the relative standard deviations (RSD, n=5) were not more than 2.5%. This method has the advantages such as high sensitivity and good selectivity.