1. Introduction

Crystalline nonlinear materials are ubiquitous in almost all areas of laser technology as key functional components for manipulating the phase, polarization, amplitude and frequency of coherent beams. In the deep ultraviolet (DUV; λ<300 nm) the number of crystals available is limited to only a few candidates that simultaneously satisfy the requirements for low absorption and high nonlinear polarizability. For second order nonlinear processes, materials such as the barium and lithium borates are common place but for third order nonlinear processes such as frequency conversion by stimulated Raman scattering there are very few options. Solid-state Raman laser operation has been largely limited to the visible and infrared where the high gain crystals have good transmission. Indeed, the shortest wavelengths crystalline Raman lasers have been in the UV-A (315-400 nm) for single-pass Raman generation [1] and longer than 500 nm for Raman laser oscillators [2,3]. Pumping of Raman lasers in the DUV is interesting for generating Stokes wavelengths spanning the UV-C (<285 nm) and UV-B (280-315 nm) regions important for applications such as explosive vapor detection as well as more generally in biosensing, environmental sensing and covert optical communications. Although DUV generation by using cerium doped laser crystals [4,5], and harmonic conversion of visible optical parametric oscillators [6] and Raman lasers [7] are important options generation by short wavelength pumping of Raman lasers has advantages in wavelength versatility (discrete tunability and multiwavelength operation, see e.g. [8].) and in high beam quality conversion via Raman beam cleanup [9].

Diamond, which can be now synthesized with excellent optical properties and a range of properties well suited to pumping in the ultraviolet. It has a bandgap wavelength of 230 nm, and thus can be pumped at longer wavelengths including by industry-standard 266 nm and 355 nm lasers (4th and 3rd harmonic of Nd lasers) or excimer lasers operating at 248 nm and 308 nm. In contrast, high gain Raman crystals commonly in use (metal tungstates, metal vanadate, metal nitrates and silicon) are incompatible with such pump sources due to band edges at wavelengths longer than 350 nm (see Fig. 1 ). The extraordinary high thermal conductivity of diamond (~2000 W/m/K [10]) is promising for developing high average power devices. It has a reduced affinity for taking up UV absorbing impurity ions in the tightly bonded crystal lattice compared to other UV dielectric crystals [11], thus it is a material potentially less vulnerable to bulk photo-degradation. The Raman gain coefficient of diamond is the highest of all materials [12] and increases with the Stokes frequency as is well known for Raman materials (see e.g. [13].) as shown for selected materials in Fig. 1. The high Raman gain holds promise for creating lasers with low threshold and using short crystal lengths. On the other hand, operating close to the bandgap presents a challenge for efficient and stable pulsed operation due to increased multi-photon absorption.

 

Fig. 1 Gain coefficients and short wavelength transmission ranges for diamond compared with other exemplary Raman materials [12,20]. The gain coefficient plotted for diamond at 266 nm is that determined in this study.

Download Full Size | PDF

To date, diamond Raman lasers have been investigated for pumping at 532 nm [14–16] and 1064 nm [17–19] but the authors of unaware of investigations in the ultraviolet. Herein we report a diamond Raman laser pumped at 266 nm generating picosecond pulses at the first Stokes 275.7 nm. The combination of diamond’s high Raman gain coefficient and a synchronously pumped mode-locked laser architecture have enabled us to attain laser thresholds at nJ input pulse energies. The study highlights some important considerations for achieving efficient DUV operation.

2. Results and Discussion

Figure 2 shows a schematic diagram of the experimental arrangement. The pump laser was a diode-pumped cw mode-locked Nd:YVO₄ laser generating 28 ps pulses at a repetition rate of 78 MHz (Photonic Industries Inc. PS300-1064). The 1064 nm output was chopped to reduce thermal effects in the doubling and quadrupling stages, generating 100 μs duration bursts of pulses every 1.3 ms. The frequency doubling was performed using a noncritically phase-matched 3.5-cm-long lithium triborate (LBO) crystal. The 532 nm radiation was then line focused into a 4-mm-long β-barium borate (BBO) crystal, generating pulses of up to 12 nJ at 266 nm (0.93 W average power when the chopper was open) with a pulse duration of 25 ps and M² ≈2. The 266 nm pump beam was focused through M₁ into the diamond crystal with 15-cm lens L₁ to a 10 µm (1/e² radius) waist, with a confocal length of 3.0 mm.

 

Fig. 2 Experimental layout of the synchronously pumped diamond Raman laser operating at 275.7 nm.

Download Full Size | PDF

The diamond laser used a 5.7 mm long Type IIa synthetic diamond crystal (Element Six, UK) with low birefringence (< 10⁶; as specified by the manufacturer) and selected for low UV absorption. The manufacturer estimated the concentration of nitrogen, the dominant UV absorbing impurity, to be less than 20 pbb. The crystal was Brewster cut and oriented so that the p-polarized pump was parallel to the [110] axis. The diamond laser cavity was a z-fold astigmatically compensated configuration comprised of two curved mirrors (M₁ and M₂, with 100 mm concave radius) and two plane mirrors (M₃ and M₄). The separation of the curved mirrors was initially set close to 100 mm, and then tuned to match the cavity waist and pump spot size to optimize the laser efficiency. Synchronous pumping relies on matching the inter pulse period of the pump laser with round trip time of the Raman laser resonator to build-up an intense circulating picosecond pulse in the Raman resonator over many pulses, and the laser behavior is very different to standard external-cavity Raman lasers. The two long arms of the Raman cavity were both approximately 91 cm, and synchronization was achieved by adjusting the cavity length via mirror M₃. A 6-mm-thick UV-grade uncoated silica plate placed at an angle in the cavity was used as a variable output coupler. There was also leakage of 1% at 276 nm through mirrors M₁ and M₂; the reported output powers include these leakages as well as all 4 reflections off the silica plate, in order to represent the single beam power that could be produced using low-transmission mirrors and a single output coupler.

The cavity length and output coupling were systematically adjusted to obtain maximum Stokes output energy. For a total output coupling of 15% (4% leakage and 11% off the silica slide), output pulse energies of up to 0.96 nJ were measured for a maximum available pump energy of 9.35 nJ, which corresponds to an optical-to-optical conversion efficiency of 10.3%. The lasing threshold was 5.8 nJ, which is 4-5 orders of magnitude lower than the 100 μJ threshold for diamond Raman lasers pumped at the 532 nm with 10 ns pulses [15] and approximately 4.5 times lower than the 26 nJ threshold for synchronous pumping with 30 ps, 532 nm pump pulses [16]. Figure 3 shows that the output pulse energy increased with input pulse energy initially with a slope efficiency of 25%. The slope efficiency decreased fractionally as the pump power was increased, consistent with our calculations below for intensity-dependent two-photon absorption (TPA) in the diamond crystal.

 

Fig. 3 Measured (squares) and simulated (line) laser output at 275.7 nm as a function pump pulse energy.

Download Full Size | PDF

In order to understand the factors affecting the laser threshold and slope efficiency, we used a numerical laser model adapted from that described in [21]. The model solves the equations for transient Raman scattering and was augmented for this study to include the effects of TPA. The model simulates a sequence of single passes through the crystal, using the output Stokes field from one pass as the input Stokes field for the following pass; the simulation is terminated when the Stokes pulse has reached its steady state profile. The performance as a function of cavity length is calculated by retarding or advancing the Stokes pulse before it is recycled into each successive round trip.

We first characterized the diamond material and cavity losses to determine as many of the input parameters as possible for the numerical model. The non-linear TPA loss coefficient β for the diamond sample was determined using the same 266 nm pump laser, by pumping the crystal with a range of peak power densities up to ~500 MW/cm² and recording the absorption, yielding a fitted β = 1.65 ± 0.21 cm/GW at 266 nm. This is in reasonable agreement with measurements for both natural and CVD diamond at similar wavelengths [22]. For the 276 nm field, the β value of 1.56 cm/GW was determined using the wavelength scaling formula in [23] along with an intermediate value for the process of absorbing one photon from each field. The linear cavity losses, which include absorption and scatter in cavity mirrors and the diamond, were determined using a cavity ring down loss measurement by interrupting quickly the 266 nm pump beam, and recording the decaying output of the diamond laser with an oscilloscope. By fitting the decaying Stokes pulse train recorded after the pump beam was completely blocked we obtained a 1/e ring-down time of 50 ns, corresponding to a round-trip cavity loss at 276 nm of 23% - comprising the 15% output coupling, and an additional 9.5% loss. Since we measured the loss by fitting the low intensity tail of the decaying output of the laser, nonlinear effects such as TPA did not affect the measurement. The additional loss is mainly attributed to absorption in the mirrors—in the UV, absorptions of 0.5% are usual for simple high reflecting coatings, and can be as high as 2% for complex dichroic coatings.

With these measurements, the set of input parameters for the model are the pump pulse duration (25 ps, with an assumed Gaussian temporal profile), the cavity mode waist (10 µm), the output coupling (15%), diamond length (5.7 mm), linear cavity losses (9.5%), and the above β values. A phonon dephasing time for diamond of 6.8 ps [24] was used. The single remaining free parameter, the Raman gain, was adjusted in the model to achieve best fit with the experimental data.

The model predictions for the Stokes output power are plotted in Fig. 3. The threshold matches the experimental data when using a Raman gain coefficient of 100 ± 10 cm/GW. The coefficient is approximately 2 and 8 times higher than the values for pumping at 532 nm (50 cm/GW [16]) and 1064 nm (12.5 cm/GW [25]) respectively. The higher value is consistent with an expected increase for shorter wavelengths but we note is markedly lower than the expected value from the empirical relation in Ref. [13] which predicts a coefficient over 10 times higher than for pumping at 532 nm. We also note that recent studies have shown that for pump polarization along the [111] crystal axis (which is rotated 35° with respect to the [110] axis used in the present work) increases the Raman gain by approximately 30% [19]. Accordingly, lower thresholds are therefore likely using crystals cut that allow such access.

The modeled efficiency and dependence of output energy on input energy also closely match the measured values. The lower slope efficiency (25%) than that reported for the similar 532 nm pumped system (45% [16]) is attributable to the higher absorption losses of the resonator mirrors and the impact of TPA. The slope efficiency decrease at higher energies is due the increased TPA loss. Better-optimized mirror coatings might reduce their combined absorption to as little as 3% (or 2% for a three-mirror cavity), but TPA is an unavoidable consequence of the high peak powers required for Raman conversion.

We observed that the output of the diamond Raman laser decreased by approximately a factor of 2 over a period of 10 min at average input pump powers of 60 mW. As the laser performance decreased, a corresponding flaring of the pump beam profile exiting the diamond crystal was observed which indicated significant refraction by the diamond. Close inspection of the diamond crystal surfaces using an optical interferometric profiler revealed very shallow pits on the entrance and exit facets.

Figure 4 shows the detail and profile of one of the affected areas. The typical pit depth after 10 minutes of operation was approximately 50 nm, with intermediate measurements at one minute intervals confirming that the removal rate was roughly constant with exposure duration. The average removal rate per pulse corresponds to 1.4 × 10⁻⁸ nm or 395 carbon atoms, which is many orders of magnitude fewer than the surface atoms exposed to the pump beam. The etching also occurred in the absence of the Stokes beam.

 

Fig. 4 Interferometric optical profile image of the surface pit after exposure to 3.6x10¹⁰ pulses of energy 10 nJ. The cross-section through the deepest point is also shown.

Download Full Size | PDF

A similar etching process, occurring below the threshold for explosive ablation, has been reported for 248 nm incident pulses of duration 15 ns in Ref. [26] and described by the authors as a ‘nanoablation’ process that is two-photon assisted. In our case, by reducing the pump intensity by a factor of 10, we found that the ablation rate was reduced by a factor of approximately 30, also consistent with a multi-photon process. Since relaxation times of free carriers and surface temperature gradients are much shorter than the interpulse period [27,28], multi-pulse effects are presumed to be negligible. More study is required to better understand the etching process and to determine a method for prevention (for example by using hydrogen termination of the diamond surface or using dielectric crystal coatings). Nevertheless, the results highlight an opportunity to develop a method for highly controlled removal of surface atoms at near single atom per pulse rates. For the incident pulses of ten times lower energy, the ablation rate was 5 × 10⁻¹⁰ nm per pulse or 13 carbon atoms per incident pulse and 10⁶ times slower than the nanosecond laser etching reported in ref. [26]. DUV ultrafast etching of diamond may therefore provide a method to meet challenges in microstructuring of diamond micro-devices such as waveguides and microlenses.

4. Discussion and Conclusions

This study has demonstrated very low threshold and DUV operation of a diamond Raman laser. The threshold, which is several times lower than for a synchronously pumped system pumped at 532 nm [29] and the lowest for any crystalline Raman laser as far as we are aware, is a direct consequence of the high diamond Raman gain coefficient at short wavelengths (~100 cm/GW at 266 nm). The maximum efficiency of 10.3% is lower than that typically observed for Raman lasers (~50%), however there is substantial scope for this to be increased. Increased overall efficiency is likely by using input pulse energies more than twice above threshold and cavity mirrors with reduced absorption losses. However, the effects of TPA are significant at 266 nm and thus efficiencies are limited to below that seen at longer wavelengths. Due to reduced β at longer wavelengths [23], pumping at wavelengths further from the bandgap, such as 355 nm, is also likely to enable higher efficiencies. We expect that the considerations here will also apply in designing UV pumped diamond Raman lasers pumped using nanosecond (e.g., Q-switched) and continuous wave pump lasers, and also in adapting multiwavelength and wavelength selectable Raman designs applied to visible wavelengths [2,30] into the UV. By leveraging the high thermal conductivity and low thermal expansion coefficient of diamond these systems also have significant potential to be scaled to high average powers without thermal degradation.

In conclusion, we have reported a 275.7 nm first Stokes diamond Raman laser synchronously pumped at 266 nm at 78 MHz. There is significant promise for creating low-threshold and wavelength versatile diamond Raman lasers at wavelengths >230 nm. Two-photon assisted etching of the diamond surface has been identified as a challenge that needs to be solved for long term stable operation.