1. Introduction

In the recent years, a demand for a compact, reliable and high power femtosecond laser is growing. It has several applications not only in a scientific field but also in an industrial field, such as optical comb generation, cutting a very thin glass plate, processing for transparent materials and so on. A Ti³⁺:Al₂O₃ laser system is the most developed laser system in an ultrashort-pulse region so far. Today few-cycle-pulse oscillators, several-J-energy and several-hundred-TW-peak-power amplifier systems are commercially available. However, there are deficiencies in their applications in the industrial field. Ti³⁺:Al₂O₃ laser system needs a green pumping and has a large quantum defect that result in requirement of an expensive pump source and a large cooling system. Those problems also make its high average power operation difficult even though Al₂O₃ has a high thermal conductivity. Hence, in the past decade, femtosecond lasers based on Cr³⁺ and Yb³⁺-ion-doped materials have been extensively investigated [1–15]. Yb³⁺-ion-doped crystalline materials have been particularly recognized as great candidates for a highly efficient high power femtosecond laser, because they possess several advantages. Their unique energy-level scheme composed of only two manifolds leads to very small quantum defect and absence of undesirable effects such as excited-state absorption and cross relaxation, that allow highly efficient operation with small thermal loads [16]. They also enable a high average power operation and their relatively long upper state lifetimes make them prominent gain media for amplification. In addition, their fluorescence peaks occur near the wavelength of 1 μm range where a valley of the fluorescence spectra of Ti³⁺:Al₂O₃ and Er³⁺-ion-doped laser exists. Therefore their wavelength ranges are important for some applications. On the other hand, their quasi-three-level laser scheme needs high intensity pumping to overcome a reabsorption effect. The performance of commercially available InGaAs-laser diodes (LDs) has been dramatically improved in this decade, so that they have strongly accelerated the development of Yb³⁺-ion-doped lasers. From a mode-locked ytterbium fiber oscillator, a pulse duration of as short as 28 fs has been reported [17]. Solid-state laser operations with sub-100 fs pulses [5–12] as well as high-power femtosecond solid-state laser operations [18–20] based on direct LD pumping also have been reported. While some single crystals and disorder crystals have so broad gain bandwidths that sub-100 fs pulses can be generated, they tend to suffer their low thermal conductivities, low mechanical toughness, low absorption cross-section values and/or anisotropy of the thermo-mechanical properties. Among Yb³⁺-ion-doped crystalline materials, the isotropic sesquioxides (Re₂O₃, Re=Y, Sc, and Lu) are very attractive for a high-power femtosecond laser since they have excellent thermal conductivities, moderately broad fluorescence spectra and isotropic construction [21]. A high-power femtosecond laser operation with a high efficiency based on Yb³⁺:Lu₂O₃ single crystal has been successfully achieved [22]. On the other hand, their fluorescence spectral bandwidths are not so much broad, so that their pulse durations were about 200 fs in experimentally [21,23], and have been considered to be limited to about 100 fs. Recently, we have reported sub-100 fs pulse operations with Yb³⁺:Sc₂O₃ and Yb³⁺:Lu₂O₃ [24, 25] where the spectral bandwidths of the pulses have become broader than their fluorescence spectral bandwidths. It is expected to arise from nonlinear effects such as self-phase modulation and Kerr-lensing. Such extremely broad spectral broadening over the gain-bandwidth limitation of a gain material has also been observed with Yb:YAG oscillators [26, 27]. In the case of sesquioxides, their high nonlinear refractive induces (5.32±1.33 ×10^-13 esu for Sc₂O₃ and 5.79±1.45×10^-13 esu for Y₂O₃) [28] can be helpful for the broadening. Therefore, Yb:sesquioxides can be considered as great candidates for a sub-100 fs high average power laser.

In this letter we have demonstrated ultrashort-pulse generation based on Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ ceramic multi-gain-media oscillator. A 66-fs pulse duration with a 1.5-W average power and a 53-fs pulse duration with a 1-W average power were obtained without water cooling. To our knowledge, this is the first demonstration of an ultrashort-pulse operation based on multi-gain-media oscillator and is the highest average power operation among sub-100 fs Yb³⁺-ion-doped solid-state lasers ever reported. We also demonstrated 68-fs pulse generation based on Yb³⁺:Y₂O₃ ceramic alone as a reference.

2. Multi-gain-media oscillator

The center positions of the fluorescence spectral peaks of Yb:sesquioxides are slightly different to each other. Therefore, a broader gain bandwidth can be formed by a combination of them. The effective gain cross section σ_gain of Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ ceramic multi-gain media can be written as

where σ_el, σ_al, σ_e2 and σ_a2 are emission and absorption cross sections of Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃, respectively. β₁ and β₂ indicate population inversion ratios of Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃, respectively. α indicates ratio of the Yb³⁺-ion number in the Yb³⁺:Y₂O₃ gain part against the total Yb³⁺-ion number interacting with the laser mode. The effective gain cross sections for different ratios α and β are shown in Fig. 1. Its FWHM strongly depends on the ratios α and β. With proper α and β, the FWHM around 1035 nm becomes broader than 25 nm, more than 1.5 times broader than that of Yb³⁺:Sc₂O₃ (11.6 nm) or Yb³⁺:Y₂O₃ (17 nm). In addition, this gain bandwidth broadening can be carried out with keeping their own properties such as thermal conductivities, mechanical toughness and absorption cross-section values.

3. Experimental set-up

The experimental setup is schematically shown in Fig. 2. We constructed a Z-shaped astigmatically compensated cavity. As the gain media, an Yb³⁺:Sc₂O₃ (1 mm thick, C_Yb=2.5 at.%) and Yb³⁺:Y₂O₃ (1.5 mm thick, C_Yb=1.8 at.%) ceramics were used simultaneously. They correspond to α of 0.46. They were physically contacted to each other without any coating and arranged at the Brewster angle. Due to their close refractive index values [28], their Brewster angles are almost the same as each other and additional loss of the contact was very small. They were mounted in a copper holder without any additional cooling. The Yb³⁺:Sc₂O₃ ceramic was situated at the near side of the pump source and the Yb³⁺:Y₂O₃ ceramic was situated at the opposite side. As the pump source a broad-stripe LD [emission area of 1 × 95 μm (vertical × horizontal), 8 W, 976 nm, Δλ~5 nm, Lumics GmbH] was used. The measured beam quality M² of the pump beam were ~1 in the vertical direction and >20 in the horizontal direction. The pump beam was focused into the ceramics to 1/e² diameters of about 25 μm × 100 μm by four beam-shaping lenses. The maximum incident pump power was 7.4 W. The folding mirrors (M1, M2, Layertec GmbH) have 100 mm radii of curvature (ROC) on both surfaces to eliminate a concave lens effect and are antireflection coated for wavelength below 980 nm and high-reflection (99.9%) coated above 1020 nm. In the continuous wave (cw) operation, the laser mode diameters inside the gain ceramics are calculated to be about 40 × 38 μm in air. To achieve stable and self-starting mode-locked operation, a semiconductor saturable absorber mirror (SESAM, BATOP GmbH) with a 1% saturable absorption and a 10 ps recovery time was used. The laser beam was focused onto the SESAM by a concave mirror (M3, ROC = 400 mm). The estimated laser mode field diameters focused onto the SESAM are about 220 × 190 μm. As a dispersion compensation element, an SF10 Brewster prism pair (P) with the tip-to-tip separation of 40 cm was inserted in the output arm. A 10% transmittance output coupler (OC) with a wedge of about 30 minutes was used.

 

Fig. 1. (a) Absorption and emission cross sections of Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ used in the estimation. (b)-(d) Estimated effective gain cross sections of their multi-gain media for (b) α= 0.3, (c) α= 0.5 and (d) α= 0.7 with different β’s.

Download Full Size | PDF

 

Fig. 2. Experimental setup of the mode-locked multi-gain-media oscillator.

Download Full Size | PDF

4. Results

In a cw operation (replacing the SESAM by a high-reflection mirror), an output power of about 1.9 W was obtained from the cavity. For the mode-locked operation, at first we optimized the cavity with respect to the average power. Under this optimization an average power of 1.7 W was obtained. The output pulses, however, showed serious multi-pulse instability and the pulse duration was limited to several hundred fs. Next, we optimized the cavity with respect to the pulse duration. The positions and the folding angles of the concave mirrors (M1, M2) were carefully adjusted. Under the optimization the resonator became in an almost unstable state in the vertical direction and was not fully astigmatically compensated. However, above the incident pump power of ~7.3 W (this threshold power strongly depends on the alignment) the multi-pulse instability was strongly suppressed; in other words, the multi-pulses were converted to a single pulse. The allowance of the distance between the folding mirrors for this suppression was only several 100 μm. During the optimization the average power once decreased to about 1.3 W. At the same time of the multi-pulse suppression, the average power increased from 1.3 W to ~1.5 W and the spectral bandwidth broadened from several nm to ~20 nm. The significant change in the laser mode field diameter was also observed at the same time. The sech²-fit pulse duration of 66 fs with the average power of 1.5 W at the repetition rate of 89 MHz was obtained (Fig. 3). The center wavelength was 1041 nm and the spectral bandwidth was 19.7 nm. The time bandwidth product was 0.36. The peak intensity was about 0.25 MW and the optical-to-optical efficiency was about 18.8% (against the LD power of 8 W). As shown in Fig. 4, the measured laser-mode profiles of the leaking beam at the point X (see Fig. 2. The distance from M3 was about 40 cm) was an ellipse before the multi-pulse suppression and it became a circle after that. By changing the insertion depth of the prisms and the distance of the folding mirrors (M1, M2), the shortest pulse duration of 53 fs with a 1-W average power was also achieved (Fig. 5). The spectral bandwidth was 27.3 nm and the time bandwidth product was 0.40 (Fig. 6). About the stability, the short pulse operation kept for longer than 30 minutes. However, it showed an instant instability during the operation and the stable time decreased as the pulse was shorter. It is also very sensitive to the alignment. The mode-locked operation has also been investigated with a couple of SESAMs with different specifications (BATOP GmbH , absorption depths of 0.5% and 1.3% with a recovery time of 500 fs). Pulses as short as sub 60 fs have been also obtained with both SESAMs. Therefore, the pulse durations were almost independent from the SESAMs.

 

Fig. 3. Pulse train of continuous wave Kerr-lens mode-locking in a μs-time scale.

Download Full Size | PDF

We also demonstrated a mode-locked operation based on an Yb³⁺:Y₂O₃ ceramic (3mm thick, C_Yb=2 at.%) alone. The Yb³⁺:Y₂O₃ ceramic was also oriented at the Brewster angle and the cavity setup was almost the same but a different pump source (5 W fiber-coupled LD, 50-μm core diameter, NA of 0.15, center wavelength of 975 nm, Δλ~5 nm) and a different output coupler (5% transmittance) were used. By this setup, the similar behaviors were observed and 68-fs pulses with an average power of 540 mW were obtained at the center wavelength of 1036 nm (Fig. 7). The spectral bandwidth was 20 nm. The repetition rate was about 99 MHz. However, it was stable for only less than 1 minute at the pulse duration of 68 fs, and therefore it seems to be difficult to generate much shorter pulse with the setup. We also have investigated Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ multi-gain media oscillation with the same pump source and OC. The pulse duration of as short as 56 fs with the average power of 380 mW has been achieved.

 

Fig. 4. The measured laser mode profiles of the leaking beam (at the point X in Fig. 2). (a) Before multi-pulse suppression: laser mode field diameters were 2900 × 1800 μm (vertical × horizontal). (b) After multi-pulse suppression: laser mode field diameters were 2060 × 1800 μm.

Download Full Size | PDF

 

Fig. 5. Autocorrelation trace of the 53-fs pulses with the average power of 1 W. The experimental data (circles) and sech2-fitting curve (solid curve) are shown.

Download Full Size | PDF

5. Discussion

The asitigmatism compensation of the cavity in the multi-pulse suppression regime (Fig. 4) is considered to be caused by the Kerr-lens effect. The cavity was aligned at an almost unstable state in the vertical direction where the cavity became very sensitive to the Kerr-lens effect, so that the significant change of the beam profile was apparent only in the vertical direction. Similar behaviors have been reported with Kerr-lens mode-locked operation with a Brewster configuration [3,6,24,25]. The multi-pulse suppression increases the pulse energy and shortens the pulse duration several times, so that the Kerr-lens effect is strongly enhanced. From a different viewpoint, the Kerr-lens effect causes deeper gain and/or loss modulation with higher saturation energy than those by a SESAM, suppressing the multi-pulse instability [29]. The increase of the average power from 1.3 W to 1.5 W even with the low Q cavity (10% output coupling) is also the evidence of the large gain and/or loss modulation. Therefore, we believe the present operation is almost governed by Kerr-lens mode locking. As is in ref. 24, Kerr-lens mode locking without a SESAM will also be possible, with the penalty of reduced long-term (> 15 minute) stability.

 

Fig. 6. Spectrum of the 53-fs pulses (black curve) and normalized emission spectra of the Yb³⁺:Sc₂O₃ (blue curve) and Yb³⁺:Y₂O₃ (red curve) are shown.

Download Full Size | PDF

 

Fig. 7. Autocorrelation trace of the 68-fs pulses from Yb³⁺:Y₂O₃ ceramic and their spectrum (inset, black curve) and fluorescence spectrum (inset, red curve). The experimental data (points) and a sech²-fitting curve (solid curve) are shown

Download Full Size | PDF

In the previous reports of Yb³⁺:Sc₂O₃ and Yb³⁺:Lu₂O₃ ceramic lasers, the pulse duration were limited to be 92 fs and 65fs, respectively [24,25] (At present, pulses as short as 70 fs were obtained from Yb³⁺:Sc₂O₃ ceramic). In the case of the Yb³⁺:Y₂O₃ mode locking, the pulse duration was also limited to be 68 fs as shown above. In the case of the Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ multi-gain-media oscillator, pulses as short as 53 fs were obtained. The spectrum of the 53-fs pulses was centered at 1042 nm and broadened to 1010 nm and 1075 nm (Fig. 6). In the case of single gain medium, as seen in the inset of Fig. 7 and in ref. 24, strong spectral broadening was observed only to the longer wavelength side from the gain peak due to the existence of reabsorption loss. On the other hand, at the present multi-gain-media mode-locked oscillator, the existence of a 1031 nm gain peak of Yb³⁺:Y₂O₃ helped the spectral broadening to the shorter wavelength side from the gain peak at 1041nm of Yb³⁺:Sc₂O₃. From a different point of view, Yb³⁺:Sc₂O₃ helped the spectral broadening to the longer wavelength side from the 1031 nm gain peak of Yb³⁺:Y₂O₃. At the shorter wavelength side, the folding mirrors (M1 and M2) do not keep the high reflectance (>99.9%) below 1020 nm and also have large high-order dispersion near 1010 nm, which limits the spectral range of the shorter wavelength side. At the longer wavelength side, the spectrum reached another gain peak of Yb³⁺:Y₂O₃ at 1076 nm. This is greatly interesting, because it indicates that the presented operation made use of not only the different gain materials but also the separating different gain peaks simultaneously with the help of the large spectral broadening effect and the multi-gain media. Additionally, Yb³⁺:Sc₂O₃ also has another gain peak at 1094 nm. Although the emission cross sections at the longer wavelength side are about 3~5 times smaller than those of the shorter wavelength side, the reabsorption at the shorter wavelength side can modify the effective gains to comparable values by optimizing the Yb³⁺-ion concentration. Therefore, the effective gain bandwidth of Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ multi-gain media can become much broader. It should be remarked that even though a Kerr-lens effect can broaden the spectrum with a large modulation depth, the broad gain bandwidth also plays an important role to sustain a short pulse duration without multi-pulsing. This is because, if the gain bandwidth is narrow, the broadened pulse spectrum decreases the effective gain, and once the decrease becomes comparable to the modulation depth of the Kerr-lens effect, multi-pulse operation occurs. While Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ were physically contacted in the present experiment, a composite ceramic technique will enable a monolithic multi-gain media in the future [30,31]. Additionally, the physical contact of the multi-gain media is dispensable, because their stimulated emissions are optically connected even if the gain media are spatially separated [32]. We can select their configuration for different aims.

6. Conclusion

We have demonstrated for the first time to our knowledge femtosecond pulse generations from a directly LD pumped Yb³⁺:Sc₂O₃ and Yb³⁺:Y₂O₃ ceramic multi-gain-media oscillator. The 66-fs pulses with an average power of 1.5 W and 53-fs pulses with an average power of 1 W were obtained. The 68-fs pulses with the average power of 540 mW were also obtained from Yb³⁺:Y₂O₃ ceramic. We believe that the generation of much shorter pulses with a higher average power will be possible by further optimization. The multi-gain-media system will be very helpful for ultra-short-pulse lasers and amplifiers.