1. Introduction

Synchronously pumped optical parametric oscillators (OPOs) based on periodically poled lithium niobate (PPLN) have attracted significant interest because they provide ultrashort pulses in broad spectral ranges from the ultraviolet to mid-infrared [1–7], which are useful in time-resolved characterization and nonlinear imaging of various species [8–11]. The OPOs rely on quasi-phase matching (QPM) within PPLN crystals for an efficient frequency down conversion from a pump photon to signal and idler photons [12,13]. Owing to efficient conversion efficiency even in thin PPLN crystals, OPO operations have been demonstrated over broad signal wavelengths [3,6].

However, in typical femtosecond OPOs using the output coupling method, the output pulse energy is much weaker than the intra-cavity pulse because of the relatively low transmittance of output couplers [14,15]. The output pulse energy can become comparable with the intra-cavity pulse when one applies the cavity-dumping method, where a large portion of the intra-cavity pulse is extracted out of the cavity through an optical switching mechanism. By considering the efficiency as well as group velocity dispersion (GVD) of the dumping material, femtosecond OPOs generally adopt acoustic optic modulation (AOM) at Bragg cells of SiO₂ or TeO₂ crystals. Recently, cavity-dumped OPOs based on the linear-type cavity design were demonstrated by a few groups [16–18]. Though the linear-type cavity design is advantageous over the ring-cavity design in terms of dumping efficiency, it is unfavorable for pulse duration and wavelength tuning because the pulse dispersion and cavity-loss in gain and dumping crystals are doubled during the round-trip passage. However, to the best of our knowledge, there has been no report on cavity-dumped OPOs employing ring-cavity configurations.

In this paper, we report on a cavity-dumped ring-type OPO, which is synchronously pumped by femtosecond pulses from a mode-locked Ti:sapphire laser, and generates ultrashort pulses up to a pulse energy of 42 nJ in near-infrared. Because of the reduced pulse duration and cavity loss, a wide wavelength tuning capability of 1.02–1.65 μm, fully covering the mirror reflectance, was achieved by the simple displacement of a cavity mirror. By analyzing the wavelength tuning behavior with respect to the cavity length, we obtain cavity dispersion, and confirm its validity through a comparison with the calculated group delay dispersion (GDD) from each optical element.

2. Cavity configuration

Figure 1 shows a schematic of the OPO cavity fabricated in this work. A Ti:sapphire oscillator, which synchronously pumps a PPLN crystal through a concave mirror (M7) with a radius of curvature of 20 cm, has a center wavelength of 805 nm, an average power of 1.7 W, and a pulse duration of 140 fs. The PPLN crystal is 0.5 mm thick, and contains seven segments with different poling periods varying from 20.5 to 21.1 μm with 0.1 μm increment. All the results in this work were obtained with a poling period of 20.6 μm. To reduce the photorefractive effect and enable stable operation without heating, the crystal is doped with 1.3 mol % of MgO. The cavity is composed of two sets of folding mirrors with a focal length of 75 mm; one (M1 and M2) for the gain crystal and the other (M3 and M4) for the Bragg cell, a prism pair of SF10 for the dispersion compensation, and a pair of flat mirrors (M5 and M6). All the cavity mirrors (M1–M6) have high reflectance (>99.8%) in a wavelength range of 1.0–1.6 μm, and each pair was designed to have moderate negative values of the total GVD throughout the reflectance range. A BK7 window anti-reflection coated on one surface plays as an output coupler, for which the incidence angle was adjusted to have a reflectance of approximately 1%. We use AOM based on a TeO₂ Bragg crystal in order to dump optical beams out of the cavity. The rate of cavity-dumping was set at 804 kHz from a commercial AOM controller (Coherent 9200). When a radio frequency (RF) electrical signal is applied to a Bragg crystal, the refractive index grating is created, by which an optical beam is diffracted out of the cavity. Because a large portion of the intra-cavity pulse can be dumped out of the cavity, the dumping pulse is, in general, an order of magnitude stronger than the output coupled pulses in typical OPOs, where the output coupling efficiency is as low as several percent.

 

Fig. 1 Resonator design of cavity-dumped ring-type OPO: PPLN, MgO-doped PPLN with a poling period of 20.6 μm; AOM, TeO₂ crystal with a thickness of 3 mm; M1 to M4, dielectric mirror (R>99.8% at 1.06–1.60 μm) with a radius of curvature (R) of 150 mm; M5 to M6, plane dielectric mirror; OC, BK7 window with anti-reflection coating at one side; PZT, piezoelectric actuator; M7, pump beam focusing mirror with R = 200 mm.

Download Full Size | PDF

3. Results and discussion

Because the signal beam singly passes through the gain and the dumping crystal in one round-trip, the ring-type cavity-dumped OPOs can be advantageous over the linear-type ones in the group delay (GD) and signal loss. In linear-type OPOs, the pulse dispersion from the dumping crystal is doubled and further signal loss occurs as the signal passes the gain crystal in the direction opposite to the pump pulse. Though a prism pair is used to compensate for the positive material GVDs, higher-order terms in the dispersion are not easy to compensate [19]. Because of these advantages in the pulse dispersion and cavity loss, ring-type OPOs can achieve an extended wavelength tuning range. In synchronously pumped OPOs, in which the round-trip time of the signal pulse must precisely match with the inter-pulse period of the pump pulse, the GVD determines the signal pulse wavelength under the synchronization criterion. In another perspective, if the physical length of the cavity is varied, the signal beam will seek a new wavelength to compensate for the cavity detuning. In order to adjust the cavity length systematically, we mounted mirror M6 on a translation stage equipped with a PZT actuator.

We acquired the signal spectrum by displacing the position of mirror M6. Figure 2(a) shows the contour plot of the signal intensity as functions of the position of M6 and the wavelength in the horizontal and vertical axes, respectively. The plot indicates continuous wavelength tuning from 1.65 to 1.02 μm according to the M6 displacement from 20 to 175 μm. The wavelength dependence of signal tuning, which is closely correlated with the GDD, will be further discussed in later parts. It is remarkable that the tuning range fully covers the reflectance range of the cavity mirrors, and that this broadband tuning is easily fulfilled by ramping the PZT voltage without any detuning of the PPLN or pump wavelength.

 

Fig. 2 Evolution of the pulse spectrum according to the position displacement of M6. The inset shows normalized output spectra at several M6 position displacements.

Download Full Size | PDF

The average output powers of the dumped beam and output coupled beam were measured simultaneously with the signal spectrum. In Fig. 3, the pulse energy of the dumped beam and the dumping efficiency are plotted as a function of the wavelength. The intra-cavity pulse energy is attained by dividing the output coupled pulse energy by the coupling ratio (1%), and the dumping efficiency is the energy ratio of the dumped pulse with respect to the intra-cavity pulse. Both the pulse energy and dumping efficiency are maximum around 1.28 μm, reaching 42 nJ and 50%, respectively. A pulse energy of approximately 100 nJ with ~90% dumping efficiency was reported from linear-type cavity-dumped OPOs with similar configurations [16,18]. The low pulse energy and dumping efficiency in this system is due to the single-pass diffraction of the ring-cavity geometry in which the efficiency is up to four times lower than that of the double-pass diffraction of the linear-cavity geometry. We believe that the efficiency and pulse energy will increase if an amplifier is used to increase the RF power to the Bragg cell.

 

Fig. 3 Pulse energy and dumping efficiency as a function of the wavelength of the cavity-dumped beam, acquired by displacing M6.

Download Full Size | PDF

The dumping becomes less efficient as the wavelength approaches the short or long wavelength limit of the tuning range. In order to explain this wavelength dependence, we need to consider the diffraction efficiency of the Bragg grating created in the TeO₂ crystal by the acoustic-optic modulation. Taking into account the 380 MHz carrier frequency and the 4200 m/s velocity of the sound wave in TeO₂, the grating spacing (d) is about 11 μm. Then, the Bragg condition of $\sin \theta =$ λ/2d determines the incidence angle of $\theta$ = 3.32° for a signal wavelength of λ_m = 1.28 μm at maximum efficiency. If the signal wavelength is changed for a fixed incidence angle, the diffraction efficiency worsens according to the phase mismatch between diffracted beams. The wavelength dependence of diffracted intensity is approximated by the equation $I(\lambda )=A\times \frac{{\sin }^2(N{\lambda }_m/\lambda )}{{\sin }^2({\lambda }_m/\lambda )}$, where N is the number of the grating period and A is the normalization factor. We used an RF pulse with a duration of 10 ns, and this leads to approximately N = 3 periods of the refractive index grating. The calculated profile generates a diffraction peak of about 400 nm broadness, which is well consistent with the experimental wavelength dependence in Fig. 3. As we change the incidence angle $\theta$ by rotating the TeO₂ crystal, the peak dumping wavelength shifts accordingly, and this further confirms that the wavelength dependence follows the Bragg diffraction condition.

Our ring-cavity system is superior in terms of the wavelength tuning capability. In a previous report on a linear-type cavity-dumped system based on a PPLN crystal, wavelength tuning from 1.0 to 1.5 μm was possible by additionally optimizing the PPLN period and the pump wavelength at each step. In comparison, our ring-type system exhibits wider tuning range without adjusting any parameters except the cavity length. Though the best tuning performance was obtained from the poling period of 20.6 μm, other periods generated comparably wide wavelength tuning. This broad wavelength tuning less sensitive to the poling period results from relatively small PPLN thickness. As we tested for a thicker crystal of 1 mm, the tuning range narrowed confirming the effect of thickness on the wavelength tuning.

The pulse energy versus pump power curve presented in Fig. 4(a) indicates a pump threshold of 250 mW at a signal wavelength of 1.28 μm. In order to evaluate the temporal characteristics of cavity-dumped pulses, we carried out the auto-correlation measurement. From the result in Fig. 4(b), we find that the pulse duration is about 265 fs, with assuming a sech²-shaped pulse. With the spectral bandwidth at the full-width at half-maximum being 14 nm, the time-bandwidth product becomes 0.66, which is approximately two times larger than the transform-limited case. We think that the pulse duration can be reduced close to a transform limit by optimizing the GVD with extra-cavity prisms.

 

Fig. 4 (a) Pulse energy as a function of pump power and (b) autocorrelation signal in the time domain for cavity-dumped pulses at a wavelength of 1.28 μm. The spectrum is shown in the inset.

Download Full Size | PDF

According to the synchronization criterion, the round-trip time at each wavelength is identical to the pump pulse period. The cavity length displacement required to satisfy this condition then provides the round-trip time expected for a fixed cavity length. By neglecting the dispersions of free space, the wavelength versus M6 displacement plot in Fig. 2(a) is directly converted to the GD versus wavelength curve in Fig. 5(a). By differentiating it with the angular frequency, we obtain the GDD curve plotted in Fig. 5(b). The plot indicates that the OPO operates in the negative GDD regime, where a short-wavelength photon is delayed behind a long-wavelength one. While this kind of dispersion analysis was previously used to extract an averaged GDD in a finite range [20], we present the detailed wavelength dependence over a broad wavelength range from the reliable continuous wavelength tuning capability.

 

Fig. 5 (a) GD as a function of wavelength, deduced from the wavelength versus M6 displacement relation. (b) Comparison of experimentally extracted GDD from wavelength-dependent GD with that theoretically calculated by summing every contribution within the cavity. (c) Theoretical GDD calculated for each optical component.

Download Full Size | PDF

In order to check the validity of the GDD extraction procedure, we calculate the GDD by considering all the optical components in the passage of the signal beam, of which the results are presented in Fig. 5(c). In the case of the prism pair, measured values were used for the inter-prism distance (29 cm) and the optical pass length inside the prisms. Dispersions of the cavity mirrors (M1–M6), as expected from fabricated specific multilayer coatings, were provided by a vender (Layertec). Despite the π-phase shifted design in the dielectric mirror pair, residual oscillations are apparent in the plot. The total GDDs, contributed from the PPLN and TeO₂ crystal, prism pair, cavity mirrors, and BK7 output coupling mirror, are denoted by a red line in Fig. 5(b). The comparison in Fig. 5(b) indicates relatively good consistency with the experimental result, especially in the oscillatory feature of cavity mirrors. We think that the consistency can be further improved by taking into account higher-order terms in the calculation of the prism pair, slightly oblique incidence at cavity mirrors, and the uncertainty in the mirror coating. The plot indicates a negative GDD of approximately −1,000 fs² per round-tip of the signal beam. We could reduce the amount of negative GDD by decreasing the inter-prism distance; however, in that case, the wavelength tuning feature exhibits deteriorated signal oscillations at multiple wavelengths.

4. Conclusion

We report on the widely tunable operation of a ring-type cavity-dumped OPO, in which the gain medium of the PPLN crystal with a poling period 20.6 μm is synchronously pumped by a mode-locked Ti:sapphire laser. By ramping the voltage applied to a PZT actuator attached to a cavity mirror, the OPO wavelength can be continuously tuned from 1.02 to 1.62 μm in the near-infrared range. This superior tunability over linear-type systems relies on the reduction in GVD and cavity losses in both the gain and dumping crystal. A pulse energy of 42 nJ and dumping efficiency of 50% were obtained near a signal wavelength of 1.28 μm. The threshold pump power was approximately 250 mW, and the pulse duration was measured to be 265 fs without extra-cavity dispersion compensation. We demonstrate that the signal wavelength versus cavity length curve, enabled by the synchronous pumping criterion, enables us to diagnose the GDD properties of a laser cavity. We think that higher pulse energy and dumping efficiency can be achieved by amplifying the RF signal applied to the dumping crystal. We expect that the easy wavelength tuning capability of our system will find applications in nonlinear optical imaging and time-resolved studies of various materials. Moreover, the displacement-induced wavelength tunability can be used to characterize the GVD properties of optical materials or dielectric coatings.