We present a highly efficient picosecond diamond Raman laser synchronously-pumped by a 4.8 W mode-locked laser at 1064 nm. A ring cavity was adopted for efficient operation. With a low-Q cavity for first-Stokes 1240 nm, we have achieved 2.75 W output power at 1240 nm with 59% overall conversion efficiency. The slope efficiency tended towards 76% far above the SRS threshold, approaching the SRS quantum limit for diamond. A high-Q first-Stokes cavity was employed for second-Stokes 1485 nm generation through the combined processes of four-wave mixing and single-pass stimulated Raman scattering. Up to 1.0 W of second-stokes at 1485 nm was obtained, corresponding to 21% overall conversion efficiency. The minimum output pulse duration was compressed relative to the 15 ps pump, producing pulses as short as 9 ps for 1240 nm and 6 ps for 1485 nm respectively.
© 2014 Optical Society of America
Raman lasers that utilise stimulated Raman scattering (SRS) are practical devices that can frequency-shift the output of a conventional laser to a hard-to-reach spectral region [1, 2]. The cascading nature of SRS also allows the first-Stokes line to be shifted to second- or higher Stokes fields. The wavelength range can be further extended by combining other nonlinear process such as second-harmonic generation (SHG) or sum-frequency generation (SFG), enabling wavelength-versatile output from a single laser source . Recently, there has been progress in translating these wavelength conversion versatilities to the picosecond domain.
Picosecond Raman lasers are attractive sources for many nonlinear optical applications that require non-standard wavelengths. For picosecond operation where the pump laser pulse duration is similar or shorter than the Raman transition dephasing time T2 (typically T2 is of the order of 1-10 ps for crystalline Raman materials), the system operates in the so-called transient regime, in contrast to the steady-state regime for CW and nanosecond operation; in general, Raman gain in the transient regime is smaller for a given peak intensity than that in the steady-state regime .
Picosecond Raman lasers can be simple singe- or double-pass converters, which require high pump pulse energy of the order of hundreds of µJ or even mJ to reach the SRS threshold . The SRS threshold can be significantly reduced by placing the Raman material within the cavity of the pump laser, such as demonstrated in [6, 7]: this exploits the high intracavity pump power, but complicates the ultrafast pulse formation dynamics and may be problematic for sub-picosecond lasers. Alternatively, a separate external Raman cavity can be used, but for ultrafast pulse trains, this requires use of a synchronous pumping scheme. For this, the external cavity length is matched to the pump laser cavity such that an intense circulating intracavity Stokes pulse is maintained at steady-state by amplification by each successive pump pulse in the pulse train. While some studies have used Q-switched mode-locked pump lasers with µJ or mJ pulse energies , this method also enables Raman conversion of trains of CW-mode-locked pump pulses having nJ energies, and has been successfully employed to generate picosecond Raman output in the visible [9, 10] and UV  regions. Furthermore, the Raman cavity mirror coatings can be designed to cascade the first-Stokes to higher order Stokes fields. In 2010, Granados et al.  reported a cascaded synchronously-pumped CW mode-locked Raman laser. A 50 mm-long KGW crystal was pumped by a frequency-doubled Nd:YVO4 laser that yielded 7 W at 532 nm, with 28 ps pulses in an 80 MHz pulse train. Both first-Stokes (559 nm) and second-Stokes (589 nm) pulses were resonated in the synchronized standing-wave Raman cavity, resulting in 2.5 W output power for 559 nm and 1.4 W for 589 nm, with 35.7% and 20% overall conversion efficiency respectively. It was important to carefully manage the group velocity dispersion of the two Stokes wavelengths using dispersion elements to enable efficient conversion.
In this paper, we report picosecond Raman laser operation by using a synchronous pumping scheme in a ring cavity configuration. Using diamond as the Raman medium, we demonstrate several important features of this type of architecture including a selectable method for uni-directional forwards and backwards operation, and efficient second-Stokes generation in a singly resonant cavity. We show that the second-Stokes output is generated in the forward direction through the combined processes of four-wave mixing (FWM) and single-pass SRS, which enables Raman cavity mirror coatings to be simplified. Up to 1.0 W output power was obtained in the eye-safe region at 1485 nm, corresponding to 21% overall conversion efficiency. Backwards SRS was also observed in the high-Q cavity, and we discuss future prospects for strong pulse-compression.
2. Experimental setup
The optical schematic of experimental setup is shown in Fig. 1. For the Raman crystal we used CVD-grown diamond. Diamond was chosen for its high Raman gain coefficient and large Raman shift of 1332 cm−1 [13, 14]. Even with its relatively long phonon dephasing time of 7.1 ps , it is surpassed only by LiNbO3 for transient gain for few-ps pump pulses . Diamond also has outstanding thermal and mechanical properties that enable efficient Raman conversion for high power operation [15–17]. An 8 mm-long rectangular diamond crystal, with anti-reflective coating at 1240 nm (R = 0.06% per pass for each end-face), was oriented such that the pump beam propagated along the <110> axis and was polarized along the <111> axis in order to access the maximum Raman gain of the diamond crystal . The diamond crystal was pumped by a CW mode-locked Nd:YVO4 laser (Spectra Physics Vanguard 2000-HM532) that generated up to 4.8 W average power at 1064 nm with pulse duration of 15 ps, repetition rate of 80 MHz, and beam quality factor M2 of 1.1. The incident pump power could be attenuated by using a half-waveplate and a polarizer beam-splitting cube without otherwise changing the operating characteristics of the pump laser. Two plano-convex lenses (f1 = 500 mm and f2 = 200 mm) were used to focus the pump beam through the input mirror M1 into the center of the diamond crystal with a focal spot size ωP = 22 µm.
The external diamond Raman laser cavity was a bow-tie shaped ring cavity composed of two curved mirrors (M1 and M2, ROC = 200 mm) and two flat mirrors (M3 and M4). The coating specifications for each mirror are summarized in Table 1. Two different M4 mirrors (M41 and M42) were used for first-Stokes and second-Stokes operation respectively. For first-Stokes operation, the majority of the output power was obtained from M41 (T = 8.3% at 1240 nm) while M2 (T = 0.4% at 1240 nm) also leaked a small amount output power. For second-Stokes operation, M42 (R > 99.99% at 1240 nm) was used to increase the intracavity intensity of the first-Stokes field, while M2 operated as the output coupler (OC) for both first- and second-Stokes. A separation of approximately 205 mm between M1 and M2 produced a resonator TEM00 mode at the center of the diamond with a similar mode size (ω0 = 20 µm) to the focused pump. To optimize the laser performance, the Raman cavity had to be synchronized with the pump laser: the cavity length was tuned by changing the position of M4 with a high precision translation stage. The cavity length detuning ∆x was defined so that ∆x = 0 µm was the length that resulted in the lowest laser threshold, with a positive (negative) ∆x corresponding to a longer (shorter) cavity.
There are several advantages of using a ring cavity over the standing-wave cavity used previously in synchronously-pumped Raman lasers [9–12]. In a standing-wave cavity, the intracavity Stokes field is predominantly amplified when traveling forward together with the pump pulse along the Raman crystal; the return pass through the crystal can provide a small level of additional gain if the (potentially strongly) depleted pump power is also returned from the end mirror, but at the expense of almost doubled round-trip losses. A ring cavity design thus does not strongly reduce the round-trip Raman gain and can almost halve the round-trip losses compared to a standing-wave cavity. This is especially important in the present case where our cavity losses are mostly accounted for by scattering and surface losses associated with the Raman crystal. The synchronously pumped ring Raman laser will ordinarily provide unidirectional operation without requiring an intracavity optical diode, owing to enhanced temporal overlap of the pump and co-propagating Stokes pulses through the Raman crystal; this is discussed in more detail below. Finally, an external ring cavity gives no back-reflections to the pump laser system, avoiding the potential need for an isolator between the cavity and the pump laser. These advantages for ring cavities are of course in common with χ(2) optical parametric oscillators (OPOs), e.g , which share the similarity of being a gain system with no energy storage in the gain media.
We have in previous work [10, 11] used Brewster-cut diamond crystals. The advantage of this is that with the Brewster crystal the cavity astigmatism can be completely compensated, and surface losses can be extremely low; the disadvantage is that the threshold of the laser is increased by a factor of 2.4 by the elongation of the cavity waist in the tangential plane. In the present work using a plane-cut crystal, the cavity folding angle was kept as small as possible, approximately 5 degrees, to minimise astigmatism and AR coatings limited the surface losses to 0.06% per face.
3. Low-Q first-Stokes (1240 nm) cavity
To generate first-Stokes output, we built a low-Q first-Stokes cavity using M41 as the output coupler (OC). The first-Stokes intracavity field was unidirectional, lasing in the “forward” direction defined as that co-propagating with the incident pump beam. Note that in contrast to OPOs, synchronous Raman lasers can lase in the backwards direction, as the underlying Raman gain coefficient is the same for forwards- and backwards-SRS in crystals. Geometric factors will in general favor forward operation: the overlap time for a 15 ps pump pulse and any point moving with the Stokes pulse envelope is 64 ps for a forward Stokes pulse (corresponding to the full 8-mm crystal length) but only 7.5 ps for a backwards Stokes pulse (corresponding to the collision time of the pulses). We see then the geometry will strongly favor forwards operation unless the crystal length is significantly shortened. This is discussed again below where we nevertheless observe backward Stokes oscillation.
Figure 2 shows the first-Stokes threshold, output power and pulse duration as a function of cavity length detuning ∆x, where the output power and pulse duration were measured at the maximum pump power of 4.8 W. Changing ∆x results in a very slight change in the cavity alignment, minimised in our geometry where the cavity has a very small folding angle; we re-optimised the cavity alignment at each ∆x to minimise this effect. First-Stokes output was obtained within a detuning range of ∆x = –400 to + 45 μm. The minimum SRS threshold was 1.53 W at ∆x = 0 μm (by definition), where the maximum output power of 2.75 W at 1240 nm was also obtained. Power transfer data for ∆x = 0 μm is plotted in Fig. 3. We have achieved an overall conversion efficiency ηo = 59%. Far above the SRS threshold (above 3.5 W pump power), the residual pump power started to clamp and the slope efficiency ηs tended towards a limit of 76%; this is not far below the quantum limit of 86% for diamond indicating a very efficient laser. The beam quality factor M2 for the 1240 nm output was measured to be 1.55 at the maximum output power.
The first-Stokes generation in this work was much more efficient than the synchronously-pumped Raman lasers reported previously, e.g. ηo = 25.6%, ηs = 42% in , ηo = 29%, ηs = 41% in  and ηo = 10.3%, ηs = 25% in . The improvement is mainly attributed to the ring-cavity configuration for the present work, as well as superior coatings for both the cavity mirrors and Raman crystal used in this work. To indicate the significance of intracavity losses affecting the laser performance, we used another diamond having similar length (10 mm) and orientation but higher end-face reflectivity (R = 0.4% per end-face, compared to R = 0.06%), and obtained a reduced output power at 1240 nm of 2.1 W and ηo = 43.8% (compared to 2.75 W and ηo = 59%).
The pulse duration of the first-Stokes output was measured with an autocorrelator assuming that the pulses were Gaussian in time, and the measurement results are shown in Fig. 2 (b). The minimum pulse duration of 9 ps at 1240 nm was obtained at ∆x = + 45 μm. The effect of pulse-compression in this work was however much less prominent compared to previous reports [9–11]. Compression relies on the group delay difference between the pump and Stokes pulses transiting the crystal , which is relatively low in the current experiment: the diamond crystal is shorter than the KGW crystal employed in , and provides less dispersion compared to that for visible- or UV-pumped lasers reported in refs [10, 11].
4. High-Q first-Stokes (1240 nm) cavity
We also operated the laser with a different output coupler M42 that had a HR coating at 1240 nm. With this setup, the round trip loss at 1240 nm from the mirrors was dominated by the 0.4% transmission through M2, and the intracavity intensity of the first-Stokes field, monitored through the power leaking from M2, was significantly increased. We observed forward generation of second-Stokes at 1485 nm, as well as backward generation of first-Stokes at 1240 nm in certain detuning regions. The performances of the second-Stokes and backward first-Stokes are discussed separately in the following subsections.
4.1 Second-Stokes (1485 nm) output
We found that this laser could efficiently generate second-Stokes output, despite the fact that the second-Stokes radiation was extremely weakly resonated, with only 0.02% of the second-Stokes power at the exit of the crystal circulated for the next round trip. The laser performance as a function of ∆x is summarized in Fig. 4, showing that the second-Stokes 1485 nm was operating in a narrower ∆x region (from −105 μm to + 22 μm) compared to the first-Stokes described in the previous section. The maximum 1485 nm output power was 1.0 W at ∆x = 0 μm with 4.8 W pump power, corresponding to ~21% overall conversion efficiency. The 1485 nm output pulse duration was around 10 ps for most of the negative ∆x cavity detuning range and dropped steeply for positive detuning. The shortest pulse duration was 6 ps at ∆x = + 20 μm. The measured beam quality factor M2 was ~2.2 for the maximum output power.
We also observed a small collinear output at 935 nm corresponding to the anti-Stokes wavelength. To further investigate the behaviour of this laser, we measured the second-Stokes, anti-Stokes and first-Stokes power coupled from M2 as well as residual pump power as a function of pump power, shown in Fig. 5. The presence of anti-Stokes generation points to parametric four-wave-mixing, where two pump photons create a Stokes and anti-Stokes pair. With a strong pump field, and a collinear resonated Stokes field, this process can be strong. In the collinear geometry, the parametric process is not phase matched, and the wave vector mismatch can be written ∆kAS = 2kP - ∆kS - ∆kAS where kP, kS, kAS are the wave vectors of the pump, Stokes, and anti-Stokes fields. For diamond, with a pump wavelength of 1064 nm, ∆kAS = 8065 m−1 , corresponding to a dephasing length of 390 µm. Along the 8-mm-long diamond crystal then, the anti-Stokes will be created and back-converted several times as the nonlinear polarization at the anti-Stokes frequency comes in and out of phase with the propagating anti-stokes wave; the anti-Stokes output that we see is a result of generation over just part of a dephasing length at the exit of the crystal. Note that phase-matched generation of Stokes and anti-Stokes beams that are both non-collinear with the pump beam is negligible in this case, owing to the presence of the strong collinear Stokes beam.
Parametric generation of the second Stokes also occurs via phase-mismatched FWM, where the second-Stokes wave vector mismatch ∆kSS = 2kS - ∆kP - ∆kSS is 6857 m−1 in our case. Second-Stokes generation should be much stronger than for anti-Stokes: the former is proportional to PPPS2, whereas the latter is proportional to PP2PS where PP and PS are the pump and Stokes intracavity power, and in this laser the cavity-enhanced Stokes field is substantially the stronger of the two except very close to threshold. In contrast to the anti-Stokes wave, the second-Stokes wave will of course also see SRS gain pumped by the Stokes field. We suggest that the parametric generation of second-Stokes seeds the SRS process, generating the second-Stokes field in a single-pass; the weakly-resonated second-Stokes field is negligible in this case. This interpretation is also supported by the threshold-less behaviour of the second-Stokes, seen clearly in the inset of Fig. 5(a), and in the anti-Stokes. The combination of FWM and single-pass SRS for generating the second-Stokes output represents a practically convenient method for efficient generation. The cavity mirrors are not required to have reflectivity at the second-Stokes wavelength, nor is there a requirement for intracavity dispersion management as was used in . We note that FWM seeding of cascaded Stokes generation has been studied previously in different SRS experiments, for example .
4.2 Backwards generation at the first-Stokes (1240 nm)
With the high-Q cavity used for second-Stokes operation, backwards propagating first-Stokes output was also obtained for small positive length detunings of ∆x = 0 to + 55 μm. The output power and threshold for the backward first-Stokes output as a function of ∆x are characterized in Fig. 4(a) and 4(d) respectively. The maximum output power for the backward 1240 nm was 0.45 W at ∆x = + 27 μm. The pulse duration along the whole operating region was > 30 ps, limited by the scanning range of the autocorrelator, much longer than that for the forward first-Stokes. We must explain why backwards operation was observed given our observation above that geometric factors should favour forward operation by a large factor. The answer lies in the complexities of the synchronization effects, acting differently for forwards and backwards operation depending on ∆x.
In general, the forward SRS operates best for cavity lengths that are matched, or a little shorter that the pump laser (negative detuning), as shown in Fig. 2 and Fig. 4(a). For negative detunings, the steady-state Stokes pulse must be preferentially amplified in its tail to reshape the pulse to cause an effective delay, counteracting the fact that it is advanced on each round trip owing to the cavity length mismatch. Preferential amplification of the tail is consistent with transient Raman scattering , and so large negative detuning can be tolerated . Efficient operation of the forward-Stokes leads to strong depletion of the pump field as shown in Fig. 3 and Fig. 4(b), and so backward-Stokes operation does not reach its threshold. Positive detuning on the other hand require preferential amplification of the leading edge of the Stokes pulse: this is not easily achieved causing forward-Stokes laser output to fall below threshold for significant positive detuning. In this regime, however, the suppressed forward operation appears to enable backward-Stokes operation to achieve threshold.
The pulse forming dynamics for the colliding pump and backwards Stokes fields clearly permits operation for larger positive detuning. In future, we plan to adapt the model procedure described in  to quantitatively examine operation in this regime and investigate ways backwards operation can be enhanced. For example, the Raman cavity could be forced to operate in the pure backward direction by introducing an intracavity optical isolator to prevent the forward Stokes field from oscillating. Strong compression in backward SRS amplifiers is well studied in gases, with order-of-magnitude or more compressions possible , and potential compression down to pulse durations a short as 0.1T2  corresponding to 720 fs for diamond, and 38 fs for LiNbO3 . Backwards operation in a synchronously pumped ring laser, in which the leading edge of the backward Stokes pulse collides with the undepleted pump pulse on each round trip, thus may provide a promising method for enabling steepened and substantially shortened pulses.
In summary, we have demonstrated a highly efficient picosecond diamond Raman laser operating at both 1240 nm and 1485 nm. Up to 2.75 W output power and 59% conversion efficiency were achieved for the first-Stokes 1240 nm in a low-Q Raman cavity. The slope efficiency tended towards 76% when the pump was far above the SRS threshold. The second-Stokes 1485 nm was efficiently generated through FWM and single-pass SRS without being resonated in a high-Q Raman cavity. We have generated 1.0 W output power at 1485 nm with 21% overall conversion efficiency. The ring laser design is a convenient design for enabling separate investigation of forwards and backwards operation. The system has great potential in power scaling for the Stokes generation owning to the high slope efficiency far above the SRS threshold and to the superior thermal properties of diamond.
This work was funded by an Australian Research Council Linkage Project LP110200545, in association with M Squared Lasers Ltd. Aravindan Warrier is supported by a Macquarie University iMQRES scholarship.
References and links
1. H. M. Pask, “The design and operation of solid-state Raman lasers,” Prog. Quantum Electron. 27(1), 3–56 (2003). [CrossRef]
2. P. Cerný, H. Jelinkova, P. G. Zverev, and T. T. Basiev, “Solid state lasers with Raman frequency conversion,” Prog. Quantum Electron. 28(2), 113–143 (2004). [CrossRef]
3. H. M. Pask, P. Dekker, R. P. Mildren, D. J. Spence, and J. A. Piper, “Wavelength-versatile visible and UV sources based on crystalline Raman lasers,” Prog. Quantum Electron. 32(3-4), 121–158 (2008). [CrossRef]
4. B. I. Carman, F. Shimizu, C. S. Wang, and N. Bloembergen, “Theory of Stokes pulse shapes in transient stimulated Raman scattering,” Phys. Rev. A 2(1), 60–72 (1970). [CrossRef]
6. M. Weitz, C. Theobald, R. Wallenstein, and J. A. L’huillier, “Passively mode-locked picosecond Nd:YVO4 self-Raman laser,” Appl. Phys. Lett. 9829), 091122 (2008). [CrossRef]
7. D. S. Chunaev, T. T. Basiev, V. A. Konushkin, A. G. Papashvili, and A. Y. Karasik, “Synchronously pumped intracavity YLF-Nd-KGW picosecond Raman lasers and LiF:F2–ampliﬁers,” Laser Phys. Lett. 5(8), 589–592 (2008). [CrossRef]
8. G. G. Grigoryan and S. B. Sogomonyan, “Synchronously pumped picosecond Raman laser utilizing a LiIO3 crystal,” Soviet J Quantum Electron. 19(11), 1402–1404 (1989). [CrossRef]
12. E. Granados, H. M. Pask, E. Esposito, G. McConnell, and D. J. Spence, “Multi-wavelength, all-solid-state, continuous wave mode locked picosecond Raman laser,” Opt. Express 18(5), 5289–5294 (2010). [CrossRef] [PubMed]
13. R. P. Mildren, “Intrinsic optical properties of diamond,” in Optical Engineering of Diamond (Wiley, 2013), 1–34.
14. T. T. Basiev, A. A. Sobol, P. G. Zverev, V. V. Osiko, and R. C. Powell, “Comparative spontaneous Raman spectroscopy of crystals for Raman lasers,” Appl. Opt. 38(3), 594–598 (1999). [CrossRef] [PubMed]
16. A. McKay, H. Liu, O. Kitzler, and R. P. Mildren, “An efficient 14.5 W diamond Raman laser at high pulse repetition rate with first (1240 nm) and second (1485 nm) Stokes output,” Laser Phys. Lett. 10(10), 105801 (2013). [CrossRef]
17. R. P. Mildren, A. Sabella, O. Kitzler, D. J. Spence, and A. M. McKay, “Diamond Raman laser design and performance,” in Optical Engineering of Diamond (Wiley, 2013), pp. 239–276.
19. Q. Fu, G. Mak, and H. M. van Driel, “High-power, 62-fs infrared optical parametric oscillator synchronously pumped by a 76-MHz Ti:sapphire laser,” Opt. Lett. 17(14), 1006–1008 (1992). [CrossRef] [PubMed]
21. T. T. Basiev, M. E. Doroshenko, L. I. Ivleva, S. N. Smetanin, M. Jelínek, V. Kubeček, and H. Jelínková, “Four-wave-mixing generation of SRS components in BaWO4 and SrWO4 crystals under picosecond excitation,” Prog. Quantum Electron. 43(7), 616–620 (2013).
22. R. L. Carman, F. Shimizu, C. S. Wang, and N. Bloembergen, “Theory of Stokes pulse shapes in transient stimulated Raman scattering,” Phys. Rev. A 2(1), 60–72 (1970). [CrossRef]
23. J. R. Murray, J. Goldhar, D. Eimerl, and A. Szoke, “Raman pulse compression of excimer lasers for application to laser fusion,” IEEE J. Quantum Electron. 15(5), 342–368 (1979). [CrossRef]
24. S. F. Morozov, L. V. Piskunova, M. M. Sushchik, and G. I. Freĭdman, “Formation and amplification of quasi-soliton pulses in head-on stimulated scattering,” Sov. J. Quantum Electron. 8(5), 576–580 (1978). [CrossRef]
25. D. S. Chunaev and A. Y. Karasik, “Temporal characteristics of picosecond stimulated Raman scattering in oxide crystals,” Laser Phys. 16(12), 1668–1671 (2006). [CrossRef]