A blue (400–500 nm) laser has very important applications in underwater optical communication for its low absorption coefficient in water^[1]. Also, the blue band is admirable in materials processing and information storage for its short wavelength. The generation of 420 nm blue light from Rb vapor has attracted more interest since the first report of blue light from pumping Rb vapor with a mixture of laser wavelengths of 780 and 776 nm, in 2002^[2]. However, improving the power conversion efficiency is still a challenge. 1.1 mW cw 420 nm light with a conversion efficiency of 2.6% was realized in 10−3mbar Rb in 2010^[3], then 9.1 mW cw 420 nm light with a power efficiency of 1.5% was successfully achieved in 2014^[4]. Single-wavelength conversion using 778 nm has also been realized^[5]. Pulsed 420 nm light was also reported^[6], with single-wavelength 778.1 nm pumping; the slope efficiency was 0.5% and the power efficiency reached about 0.5%. This is the highest power efficiency record for similar pulsed laser frequency upconversion.

Although the linewidth of a dye laser is usually several GHz, which does not match the atom absorption well, a pulsed nanosecond dye laser has a much higher peak power in comparison with cw-laser diodes. Therefore, the probability for non-resonant two-photon excitation will increase dramatically. In this work, a 45 cm-long, 170°C heated Rb vapor cell was used, 420 nm pulsed blue light with energy efficiency over than 1% was generated by a pulsed dye laser with a linewidth of 3 GHz. A conversion mechanism including hyper-Raman and four-wave mixing (FWM) is given for the explanation of experimental results.

The Rb cell used in this work was homemade with quartz glass. The filling process was completed in a vacuum glove box under the protection of pure helium. The concentrations of water vapor and oxygen were controlled under 1 ppm. After the filling, the Rb cell was pumped to a vacuum of 0.1 Torr by a rotary pump, and the opening was sealed by heating. Rb metal of natural isotope distribution was used. A frequency-doubled Q-switched Nd:YAG with a repetition rate of 10 Hz was used to pump a dye laser (Continuum ND6000). Laser dye styryl8 dissolved in dimethylsulfoxide was used to generate the pump laser of 778.1 nm. The wavelength of the dye laser was monitored by a wavelength meter with an accuracy of ±600MHz. Lasers out of the Rb cell were diffracted by a grating with a known diffracting efficiency to separate the blue-light and residual-pumping laser. The laser power was recorded by mJ- or μJ-grade pyroelectric detectors. Spectra of blue light were recorded by a fiber-coupled CCD spectrometer. Different cell temperatures from 130°C to 280°C were tried for highest conversion efficiency. The dye laser was restricted to a diameter of 1.5 mm by an iris to gain a near-flattop beam through experiments for consistency.

By monitoring the blue light energy while scanning the wavelength of the dye laser, 778.10 nm is found to be the optimized wavelength for generating 420 nm blue light (two-photon excitation from the 5S1/22 to the 5D5/22 state). At a pump energy of ∼0.85mJ/pulse, a reproducible high efficiency of blue light appears at about 170°C when changing the cell temperature. Lower temperatures lead to low efficiency and higher temperatures can cause fast cell damage.

Figure 1 shows the pulse energy of the output blue light versus temperature and atom density^[7]; here, the optimized pumping wavelength of 778.10 nm is used. The wavelength of the blue light is measured to be 420.3 nm. The data points in Fig. 1 show a logarithm tendency, which means by increasing atom density the blue light energy and the corresponding conversion efficiency go higher; however, the growth speed is slower than a linear growth.

For getting as high of a blue light efficiency as possible, the cell temperature is fixed to 170°C and the pumping wavelength is fixed to 778.10 nm. After efforts on improving cell manufacturing, the energy of the blue output and the conversion efficiency are given in Fig. 2. The energy efficiency is 0.58%–1.74%. When the pulse energy of the dye laser is less than 0.5 mJ, the energy efficiency is higher than 1%. For a simple calculation, the pumping volume in the cell is about 0.80cm3, containing 2.1×1014Rb atoms. Assuming that one atom absorbs 2 photons, the corresponding absorbed laser energy should be 0.10 mJ. Since the absorption is observed to be descending with the ascending pumping energy, the descending efficiency in Fig. 2 is believed to be mainly related to the very high pumping energy. For example, when the pumping pulse is about 1.5 mJ the absorption is lower than the energy fluctuation of the pumping laser and cannot be determined. Slight “fogging” is observed on both windows of the Rb cell after ten h of heating at 170°C; if it can be avoided, the efficiency would be further improved. Each point in Fig. 2 is the average of 1500 pulses; there is still a slight fluctuation in the curves. The fluctuation is quite possibly derived from changes in the surrounding environment such as instability of temperature, cell fading, or a slow changing of the dye laser pointing direction.

Figure 3 shows the intensity profile of the blue light as the pumping laser is detuned from 778.10 nm. The intensity profile shows wide asymmetric wings. This phenomenon has not been reported by others. The red wing is spreading more than the blue wing, and a shoulder peak exists in the red side. When the pyroelectric detector is replaced by a spectrometer, two blue wavelengths, 420.3 and 421.7 nm, are observed and are shown in Fig. 4. They correspond to the 6P3/22−52S1/2 and 6P1/22−52S1/2 transitions in the Rb atom. Two wavelengths have similar wings to Fig. 3. Though larger fluctuations exist in spectroscopy, asymmetric wings can be clearly observed.

Figure 5 is the recorded absorption curve with wavelength scanning. A broad asymmetric absorption peak is observed. Under these experimental conditions Rb atoms are expected to show Doppler broadened absorption with a linewidth of about 1 GHz. Therefore, the much broader absorption peaks must represent kinds of nonlinear absorption.

The most widely accepted mechanism for blue light generation from Rb vapor under 780 + 776 nm, or single 778 nm pumping is FWM, but not a direct one. Taking the 778 nm as an example, Rb atoms can be excited to the 5D5/22 energy level by absorbing two 778 nm photons. The lower 6P3/22 level is not populated at first, so population inversion will form between the 5D5/22 level and the lower 6P3/22 level. When the population inversion is large enough, an amplified spontaneous emission (ASE) of 5.233 μm is generated. Finally, resonating FWM with 778 nm and 5.233 μm happens to generate 420.3 nm forward-going light resonating with the transition from 6P3/22 to 5S1/22^[8]. Degenerate two-photon absorption can be expressed with the formula dIdx=−NRb·σ2photon·I2,(1)where I is the light intensity, x is the distance, NRb is the Rb atom density, and the Doppler free two-photon absorption cross sections σ2photon=1.2×1018cm4·W−1 from Ref. [6]. By integrating Eq. (1) we get: 1I0−1Il=−NRb·σ2photon·l,(2)where l represents the cell length and its value is 45 cm. In Fig. 5, the max absorption results in a residual of 0.371 mJ from the original ∼0.526mJ in the experiment (pumping laser with an spot diameter of 1.5 mm and an FWHM pulse duration of 5 ns). Here, NRb=2.60×1014cm−3. The nominal σ2photon is calculated to be 6.0×10−24cm4W−1 for this experiment, which is much smaller than the expected value. The pumping laser linewidth is probably one reason responsible for this discrepancy. The factory-measured dye laser linewidth is 3 GHz (0.1cm−1), 3 times that of the Doppler-broadened atom line.

However, the two-photon absorption cannot explain everything, especially the allowed large tuning range in Figs. 3 and 4 and the broad absorption band in Fig. 5. These remind us of the possibility of stimulated hyper-Raman scattering^[9]. Atoms can absorb two photons simultaneously, then hyper-Raman scattering happens, generating directional infrared radiation just like the ASE from the transition from 5D5/22 to 6P3/22. But there is a large difference between hyper-Raman and ASE. Hyper-Raman radiation has variable Stokes wavelengths with the following wave frequency: 2hνpump−hνstokes=E6P23/2,1/2→5S21/2.(3)

The gain factor for stimulated hyper-Raman scattering can be expressed as^[10,11]G=NRbI2ωse630ℏ5(5ε0c)3Γ·{〈6p|r|5d〉〈5d|r|5p〉〈5p|r|5s〉Δω1Δω2}2·KΘ,(4)where Δω1 is (ω5P2J−5S21/2−ωpump+iγ1), Δω2 is (ω5D2J−5S21/2−2ωpump+iγ2), γ stands for the damping coefficient related to the atom life, NRb is the Rb atom volume density, Θ is the constant related to the pumping polarization and the Stokes wave propagating direction, K is a constant related to the degenerate levels, and the 〈i|r|j〉s stand for the transition dipole moments.

Hyper-Raman scattering happens between atom energy levels and “virtual energy levels.” A virtual energy level is statistical combination of real energy levels. Being nearer to real levels, a virtual level gives larger hyper-Raman transition possibility. For this work, asymmetric peaks in Figs. 3 and 4 are inferred, deriving partly from the asymmetric efficiency-to-wavelength curve of stimulated hyper-Raman scattering, since the pumping photon (778.10 nm) energy is slightly higher than the energy for the 5P2J to ground state (780 and 795 nm). Transition of the 5P2J to ground state has the largest transition dipole moment in Rb atom transitions, and the energy divergence of a pumping photon to 5P2J is apparently not symmetric in the tuning range. Finally, the asymmetric divergence contributes to the asymmetric stimulated hyper-Raman gain G in Eq. (4) and the asymmetric Stokes beam strength. The FWM efficiency is also affected by the resonating degree of the mixing waves with real energy levels just like Eq. (4). An asymmetric FWM efficiency-to-wavelength curve is believed to be another reason for asymmetric wings in Figs. 3 and 4.

Finally, as an explanation, FWM (generating 420.3 or 421.7 nm blue light, or generating them simultaneously) happens with three waves, including two degenerate waves around 778 nm and a mid-infrared beam. The mid-infrared beam is generated from ASE or stimulated hyper-Raman scattering. Hyper-Raman scattering is generally suppressed at the wavelength exactly resonating with transitions to certain energy levels, forming a dip in the curve shape of the Stokes output versus the pumping wavelength^[10,12,13]. ASE at 5.233 μm derives from two-photon absorption at the pumping wavelength; resonating 5D5/22 (778.10 nm) is believed to contribute mostly to the peak tips in Figs. 3 and 4, instead of forming a dip. A wavelength-scanning result related to the two-photon absorption process and ASE is given in Ref. [5]. Since quartz glass is not transparent for infrared light around 5.2 μm, the ∼5.2μm beam could not be measured in the experiment. In all tests, the blue light has an identical polarization direction with the pumping laser, and the blue light is only copropagating with the pumping laser. Such polarization and propagating make us believe that it comes from FWM. In all of the experiments, surprisingly the frequency upconversion did not show any fluctuation related to phase matching even at varying temperatures, though the vapor cell is 45 cm-long. One possible reason is the low dispersion of low-pressure Rb vapor. Furthermore, according to^[10], hyper-Raman scattering has the property of giving a much broader Stokes linewidth than both the pumping linewidth and the Doppler-broadened atom linewidth. So another reason may be that the broad hyper-Raman Stokes line always covers the needed mid-infrared wavelength for an exact collinear phase-matching for FWM. An investigation of phase matching on the six-wave mixing process is given in Ref. [14], where the relative spectra-peak heights show a positive correlation with the phase matching. Further investigation is needed to elucidate the phase-matching mechanism.

In conclusion, high-efficiency generating of blue light is realized in Rb vapor. With a cell temperature of 170°C, a cell length of 45 cm, and a pumping laser energy less than 0.5 mJ/pulse, the conversion energy efficiency is larger than 1%. The upconversion is explained as two steps; (1) two-photon ASE or stimulated hyper-Raman scattering and (2) FWM.