Frequency tunable femtosecond sources such as optical parametric oscillators (OPOs) are of great interest in various applications, e.g., coherent anti-Stokes Raman spectroscopy (CARS) [1], multiphoton microscopy [2], frequency comb metrology [3], and terahertz generation [4]. They are available in various wavelength ranges: the ultraviolet (UV) [5], visible (VIS) [6], near-infrared (NIR) [7], and midinfrared (MIR) region [8]. VIS OPOs are of particular importance, because the visible spectral range cannot be covered by a tunable source with typical laser gain media. Until now, it has been quite difficult and elaborate to realize a fast frequency-tunable laser source with high output powers and a broad wavelength range in the visible range. One approach is to generate fiber-optic Cherenkov radiation in a tapered photonic crystal fiber. A resulting tuning range of 414–612 nm with output power below 5.5 mW was reported in Ref. [9].

Over the past decades, crystal materials like beta barium borate (BBO), periodically-poled lithium niobate and bismuth triborate were used in synchronously pumped OPOs. Some of these reports addressed the intracavity second harmonic of the signal or sum-frequency generation of the signal and pump inside near-infrared OPOs [10–14]. As a result of the high fundamental power in these cases, up to a few hundred milliwatts of average output power have been achieved. However, the tuning range was limited to the phase-matching bandwidth of the second harmonic and sum-frequency-generation process, so that high output powers have been achieved only over a small wavelength range.

Others have reported pumping an OPO with blue light from the second harmonic of a Ti:Sa laser to create tunable femtosecond laser pulses directly in the visible range [6,15,16]. However, due to the limited pump power, the output power was small. Recently, this idea was taken up again utilizing harmonics of Yb-doped thin disk oscillators and fiber lasers [7,17] for higher pump powers.

Comparing the different wavelength tuning schemes in OPOs, methods like varying the internal crystal angle [6] or the crystal temperature [12] require seconds before stable operation at the new wavelength is reached. A much quicker way to tune the wavelength is by changing the cavity length and therefore the group delay between the pump and a chirped signal pulse–in particular when combined with an ultrabroadband noncollinear phase-matching scheme. Based on this method, the authors’ group has demonstrated tuning speeds beyond 100 nm/ms [18] in a near-infrared noncollinear parametric oscillator (NOPO).

Noncollinear phase matching in BBO can be realized by a constant angle ±α between the pump and signal beams in two different configurations. At first, the Poynting vector walk-off compensation (PVWC) geometry is characterized by a small angle between the Poynting vectors of the pump and signal with a better spatial overlap. A major disadvantage is a crossing of the second-harmonic phase matching curve of the signal, which is responsible for a strong dip in the tuning power (here at 615 nm). Second, at −α, in the tangential phase-matching geometry (TPG), the Poynting vectors of the idler and pump point in similar directions, which causes a stronger walk-off of the signal. More information about these cases can be found in the literature [19,20].

In this work, we demonstrate a high-power, high-repetition-rate quickly tunable noncollinear optical parametric oscillator, which covers almost the entire visible spectral range (VIS-NOPO). It is directly pumped by the third harmonic of an Yb-fiber laser at 345 nm with up to 3 W of pump power. Due to the noncollinear phase-matching scheme, the signal wavelength can be changed from deep blue up to the red, from 440 to 720 nm by varying the cavity length only. Long-time stable output powers up to 452 mW with 1.4% rms power fluctuations are obtained. Without extra pulse compression effort, the shortest signal pulse duration is 286 fs.

Figure 1 shows a schematic sketch of the experimental setup. It consists of an Yb-fiber pump laser with an average output power of up to 22 W at 1040 nm, a repetition rate of 50.2 MHz, and pulse durations of 250 fs. Next, the third harmonic of the Yb-fiber laser at 345 nm is created in two steps. The second harmonic of the pump laser is generated in a BBO crystal (θ=23.2°, L=1.5mm) in one arm and combined with the fundamental from a second arm via sum-frequency mixing in a second BBO crystal (θ=32.5°, L=1mm). The generated third harmonic at 345 nm can be scaled up to 3 W of average output power with an overall efficiency of 17%.

The pump beam is then focused into a Brewster-cut BBO crystal (θ=39°, L=2mm) inside the VIS-NOPO cavity. A noncollinear angle of α=4.5° in the PVWC geometry is set for ultrabroadband phase matching. The cavity consists of six plane and two curved (ROC=−150mm; d=25.4mm) doubly chirped mirrors (DCMs) with high reflectivity and defined dispersion in the range of 400–700 nm. The mirrors together with the gain crystal flatten the group-delay dispersion (GDD) curve close to zero, and a BK7-substrate (L=3mm) placed in Brewster’s angle in the cavity introduces a well-defined net GDD. An output coupler with 14% transmission in the spectral window from 480 to 720 nm is placed at the end of the cavity. The other end mirror is mounted on a piezo-driven translation stage, which is used to tune the cavity length.

Figure 2 shows the recorded signal spectra, output powers, and pulse durations of the VIS-NOPO as a function of the cavity length. Changing the cavity length by 400 μm results in a signal tuning range from 440 to 720 nm. The green line in Fig. 2(a) represents the calculated GD of all intracavity elements. NOPO tuning traces this curve quite nicely; even beyond 700 nm, it follows the drift in the DCM dispersion curve.

In Fig. 2(b), pulse durations from a second-harmonic intensity autocorrelator (sech2-shape assumed) and output powers are shown. The output power rises up to a maximum of 452 mW at 530 nm; between 604 and 620 nm, the signal power decreases drastically due to the generation of the parasitic second harmonic from the PVWC geometry. Except that dip, the output power stays above 170 mW over a range from 475 to 700 nm. Due to the sensitivity of the autocorrelator, the pulse duration could only be determined reliably at higher output powers. It is roughly constant at 490–600 nm and 620–700 nm around 300 to 400 fs without extra pulse compression efforts.

Figure 3(a) shows selected representative signal spectra; Fig. 3(b) shows the corresponding Fourier limits and the measured pulse duration. The quiet constant pulse width over the whole tuning range is a result of the net dispersion curve shown in Fig. 3(c). Here, the green solid line reflects the calculated GDD of all cavity elements. Due to the cavity dispersion, the signal pulses are positively chirped. However, external pulse compression by additional extra-cavity DCM bounces would result in pulse durations below 120 fs over the whole tuning range.

Figure 4 illustrates the scaling of the signal output power by variation of the incident pump power. The resulting curve shows no sign of saturation at a slope efficiency of 23.7% and a pump threshold of Pth=960mW. In these numbers, the reflection losses of the pump on the crystal surface of 19.4% are not subtracted. This measurement promises further power scalability of the NOPO concept. During the experiments, no direct damage of the crystal material was observed. However, the two-photon absorption of the BBO sets an upper limit to the pump intensity inside the crystal. Moreover, the coating on the curved DCM could be damaged by high intensities. The beam profile of the signal at the maximum output power is recorded and inserted into Fig. 4. It shows a slightly elliptical beam shape, which is attributed to a similar ellipticity in the pump beam.

In Fig. 5(a), the power evolution at 530 nm at the highest pump and signal power is plotted over a period of 1 h, whereas in Fig. 5(b), the phase noise power spectral density (PSD) of the signal and pump is shown over a range from 10 Hz to 10 MHz. The latter is obtained by a spectrum analyzer (Agilent E4440a). The slight degradation in output power is caused by a simultaneous degradation in pump power during the measurement. The determined rms noise of 1.4% is smaller than the pump rms noise of 4.1%. Further, the PSD is smaller at low frequencies; at high frequencies, the NOPO cavity inherits the phase noise almost entirely from the pump. Higher stabilities might be achievable with better passive stability and an active control of the pump and stabilization of the NOPO cavity length.

In conclusion, we have demonstrated a high-power, quickly tunable noncollinear optical parametric oscillator that covers nearly the complete visible spectral range. It is directly pumped by the third harmonic of an Yb-fiber laser. The signal wavelength can be tuned from 440 to 720 nm by changing the cavity length only. Output powers of up to 450 mW with less than 1.4% rms fluctuations over a period of 1 h and pulse durations down to 270 fs are obtained. The parasitic generation of the second harmonic of the signal causes the output power to drop to a minimum of 8 mW between 604 and 620 nm. This effect could be eliminated by utilizing the TPG, but this would result in a bad beam profile of the signal. Scalability of the NOPO concept is possible by proper cavity design to counteract the two-photon absorption of the UV light in the BBO crystal and the damage threshold of the curved mirror coatings. This makes the VIS-NOPO a great option for ultrafast spectroscopy and the generation of high-power frequency combs. For these applications, the beam pointing stability, beam quality, and tuning speeds would need to be investigated in more detail. Moreover, high tuning speeds of over 100 nm/ms could enable new methods for two-color spectroscopy and imaging.