Fig. 1. Λ-type energy structure of the CPT experiment. (a) The simple three-level model. (b) The energy levels involved in the real experiment. The main transitions are the D1 transitions of ⁸⁷Rb.

Fig. 2. Experimental setup. (a) The main part of the CPT experiment. In order to change the single-photon detuning, the laser double passes an AOM before entering the Rb buffer-gas cell. After the cell, the laser is coupled into a fiber. (b) The detection part of our experiment. Lights with different frequencies are spatially separated by a VIPA. The pictures show the spatial distribution of lights. Three detectors are set to measure I₁, I₂, and I_total, respectively. PBS, polarization beam splitter; BS, beam splitter; PD, photodetector; DAVLL, dichroic atomic vapor laser lock; VCSEL, vertical cavity surface emitting laser; VIPA, virtually imaged phased array; AOM, acoustic-optical modulator.

Fig. 3. Numerical results with the simple three-level model. The simulation parameters are Γ/2π = 2000 MHz, γ₁/2π = 1000 Hz, and γ₂/2π = 5000 Hz. (a) Typical CPT signals for I₁, I₂, and I_total when Δ/2π = −300 MHz, Ω_ab = 0.002Γ, Ω_ac = 0.001Γ. They are shifted vertically for a better visibility. (b)–(d) The center frequencies versus Δ when Ω_ab/Ω_ac equals (b) 1:1, (c) 1:1.5, and (d) 2:1. The results show that f_ave is insensitive against the change of Δ and Ω_ab/Ω_ac.

Fig. 4. Experimental data. (a) Typical CPT signals for I₁, I₂, and I_total versus δ₀ when Δ/2π = −190 MHz and Ω_ab/Ω_ac = 1:0.55. They are shifted vertically for better visibility. Both I₁ and I₂ show strong asymmetry, while I_total is more symmetric. (b) The center frequencies versus Δ. Both f₁ and f₂ are very sensitive to Δ, but f_ave is insensitive to Δ. The slope of f_ave is smaller than that of f_total.

Fig. 5. Numerical results with the four-level model. The simulation parameters are Γ/2π = 2000 MHz, γ₁/2π = 1000 Hz, and γ₂/2π = 5000 Hz. Ω_ab/Ω_ac = 1:0.55.