1. Introduction

The near-infrared (NIR) spectral region around 0.8 μm lacks efficient diffraction-limited laser sources capable of scaling to high average and peak powers. Two major laser types exist which have found many applications, though neither is well-suited for both peak- and average-power scaling. AlGaAs diode systems, while relatively simple in design and highly efficient, have low energy storage ability due to short recombination times, limiting peak power and requiring the use of low-brightness bar and stack configurations for power scaling [1]. Additionally, thermal handling requirements dictate low pulse energies and repetition rate in diode systems [2]. Titanium-doped Sapphire (Ti:sapphire) systems [3] can provide excellent beam quality and high pulse energies in this spectral region. These lasers achieve very high peak power due to the material’s large emission bandwidth, but these features come at the cost of significant complexity as the optical properties of the material itself result in inefficient systems. Despite several decades of research and development, diffraction-limited average powers of GaAs-based diodes and Ti:sapphire lasers have reached only tens of watts [4–6]. In contrast, 1 μm diode-pumped solid state lasers (DPSSLs), fiber lasers, and beam-combined systems thereof commonly achieve powers >1 kW with near-ideal beam quality [7,8].

A third source of 0.8 μm NIR laser emission which has potential for scaling to very high powers is in Tm³⁺-doped materials on the ³H₄ – ³H₆ transition. This transition can be accessed with low quantum defect by pumping with AlGaAs diode sources emitting at 0.78 – 0.8 μm. These systems have the potential to scale to high peak and average powers with relatively simple designs. Highly efficient lasing has been demonstrated in Tm³⁺-doped fluorozirconate (ZBLAN) fibers [9–11] but these systems are not promising candidates for scaling power or energy due to the difficulty in manufacturing high purity materials [12] and the poor thermal and mechanical properties of ZBLAN. Additionally the fiber geometry poses difficulties for peak power scaling as the small mode areas and long path lengths enhance nonlinear and dispersive effects. There has been little further development of ZBLAN materials for NIR Tm³⁺ lasers in the last decade. Bulk crystal hosts are more attractive for power scaling because of their higher thermal conductivity and larger apertures than ZBLAN fiber. In crystal hosts Y₃Al₅O₁₂ (YAG) and YLiF₄ (YLF), the NIR transition in Tm³⁺ is similar to the 1 μm ²F_5/2 – ²F_7/2 laser transition in Yb³⁺_, having the pump and laser levels residing in a single manifold, comparable emission bandwidth, and a 1-2 ms upper state lifetime suitable for high-power operation [13,14].

2. Thulium laser key physics

Lasers based on Tm-doped media are well-established as infrared sources operating at wavelengths around 1.9 μm. Figure 1 depicts a simplified diagram of the energy levels associated with this transition and the NIR transition. For 1.9 μm lasers, concentration quenching in highly doped media (typically 2 at.% or greater) enables emission of two fluorescence photons for each ~785 nm pump photon (the “2 for 1” effect), doubling the effective quantum efficiency in these systems. Radiative decay from ³H₄ to ³F₄ can also occur with emission wavelengths of 1.5 μm or 2.3 μm. Laser operation on these transitions has been achieved, but must contend with the long lifetime of the ³F₄ state which serves as a lower laser level for these transitions. Although the NIR transition of interest here does not directly decay to ³F₄, alternate decay paths such as multiphonon and cross relaxation must be carefully considered for lasers operating on this transition as each presents a potential source of loss to the desired NIR laser. Specifically, as the radiative lifetimes of ³H₄ and ³F₄ are of order 1 ms and 10 ms, respectively, continuous pumping with significant rate of branching to ³F₄ results in accumulation and trapping of the population at this level, reducing pump absorption.

 

Fig. 1 Energy level diagram of Tm³⁺ showing the numerous transitions stemming from the ³H₄ level and order of magnitude lifetimes in crystal hosts. Dotted lines indicate non-radiative transitions.

Download Full Size | PDF

Recently, quasi-cw NIR lasing of diode-pumped Tm:YAG at liquid nitrogen temperature was demonstrated with 36% slope efficiency [15]. The laser’s output duration was limited by population trapping at the ³F₄ level and quasi-CW pumping was necessary to allow the trapped population to spontaneously decay to the ground state. For durations greater than several tens of milliseconds, the laser’s output was ultimately extinguished through this bleaching mechanism. This demonstration confirmed that population trapping is a key physics issue in Tm-doped bulk crystal lasers operating on this transition and that cw operation requires mitigation of this trapping. Potential mitigation strategies may include selection of the crystal host to minimize non-radiative decay rates, co-doping to quench the ³F₄ lifetime [16], upconversion schemes to pump out of ³F₄ [17], or quenching by stimulated emission on the ³F₄ – ³H₆ transition [18]. There has been a report of operation on the NIR transition in bulk crystal Tm:YLF by upconversion pumping, but no data were reported [19]. Upconversion pumping is unattractive as it is less efficient than single photon schemes.

The effect of population trapping can be modelled with a rate equation analysis [15]. An analysis for trapping in Tm:YAG reveals a significantly higher population trapped in ³F₄ than in the upper laser level ³H₄. For example, in 1%-doped Tm:YAG, this analysis predicts the steady-state population in ³F₄ is 26 times that in ³H₄ if no mitigation is applied. The same analysis is extensible to other crystal hosts with substitution of the appropriate rate constants. Concentration quenching and multi-phonon relaxation are the two dominant non-radiative decay mechanisms. Concentration quenching can be minimized by keeping dopant concentrations to 1% or less [14], while the multi-phonon lifetime is largely determined by the host material. Due to their lower characteristic phonon energies, fluoride crystals (YLF, CaF₂, etc.) in general have lower non-radiative decay rates than oxide crystals (YAG, YVO₄, YAlO₃, etc.), and are likely candidates for improved efficiency and cw operation. Even so, residual trapping due to concentration quenching and alternative radiative decay paths can require additional mitigation strategies.

3. Laser experiments and results

As mentioned above, fluoride crystals are more favorable hosts for operation at the NIR transition in Tm³⁺. The measured absorption and emission spectra of Tm³⁺:YLF at 300 K and 80 K are graphed in Fig. 2. The absorption features are directly accessible with AlGaAs diode systems. The effective stimulated emission cross sections were calculated using a Füchtbauer–Ladenburg analysis, taking into account the spectrally dependent fluorescence intensity, a radiative lifetime of 2 ms [16], and a spontaneous decay branching ratio of the ³H₄ – ³H₆ transition of 0.9 [13]. The room temperature absorption and effective emission cross sections are in reasonable agreement with a previous report [13].

 

Fig. 2 Absorption (top) and emission (bottom) spectra of Tm³⁺:YLF at 80 K and 300 K. The laser was pumped at 781.5 nm and operated at 816 nm. Note the dual vertical axes (respective plots are color-coded) and the different exponential orders of the absorption axes.

Download Full Size | PDF

Although Tm:YLF has lower nonradiative decay rate than Tm:YAG, residual population trapping still affects laser performance. Co-lasing on the 1.9 μm transition can be used to continuously quench the trapped population by stimulated emission. As this transition is not directly accessed from the ³H₄ level, co-lasing does not compete with the NIR transition. Also, from the rate equation analysis [15], the ³F₄ population is proportional to the ³H₄ population, and consequently, the 1.9 μm output power is expected to remain constant with increasing pump power above laser threshold for both transitions. This analysis assumes no significant concentration quenching and hence requires low dopant concentrations as previously mentioned. In the presence of significant quenching, the effective lifetime of the ³H₄ level is reduced and higher pump powers will be required to achieve lasing threshold. To validate these predicted dynamics, a proof-of-principle dual-wavelength laser was constructed using a 0.5%-doped Tm:YLF crystal fabricated to ~15 mm length and cryogenically cooled to liquid nitrogen temperature. Cryogenic cooling was used to reduce the Tm³⁺ population in the lower laser level and enhance absorption features, reducing threshold and enabling operation with lower-power pump sources.

The laser was constructed as an end-pumped hemispherical cavity, with a flat dichroic and 250 mm radius-of-curvature output coupler. The dichroic was anti-reflection coated at the pump wavelength (781.5 nm) and high-reflection coated at the NIR and infrared Tm³⁺ emission bands. The output coupler was 3% transmissive at both lasing wavelengths and 97% reflective at the pump wavelength. At 80 K the single-pass pump absorption was 47%, resulting in double-pass absorption of ~70% in the cavity. The crystal was pumped using a Ti:sapphire laser with output wavelength and polarization tuned to maximize absorption in the crystal. The pump beam was focused to a diameter of 160 μm in the crystal.

The laser performance is graphed in Fig. 3 for cw pumping at liquid nitrogen temperature. The output wavelengths were separated in transmission using a polished silicon window at near normal incidence, reflecting the NIR output and transmitting 52% of the 1.9 µm output with accounting for Fresnel losses. The operating wavelengths were 816 nm and 1876 nm, with respective slope efficiencies 46% and 4% versus incident pump power. The low output coupling fraction and large number of intracavity optical elements (6 AR-coated surfaces on dewar windows and crystal facets) are likely contributors to intracavity loss to account for the discrepancy between the stokes efficiency (96%) and slope efficiency with respect to absorbed power (~70%). Additionally, upconversion effects were visible as blue fluorescence in the crystal during laser operation. Resonance was not possible in this cavity at the ~795 nm emission peak because the pass-band of the dichroic mirror extended nearly to 800 nm. The fitted NIR threshold was 246 mW, and the maximum NIR output was 339 mW, limited by the 1 W Ti:sapphire pump laser. The NIR output polarization was orthogonal to the pump polarization, as expected based on Fig. 2 (π pump orientation and σ emission orientation). The low-efficiency 1.9 μm signal agrees well with predictions based on rate analysis. A plane-wave analysis predicts the slope efficiency for the 1.9 µm output should be 0, and the low but nonzero observed slope efficiency can be attributed to the Gaussian beam nature of the pump and laser outputs.

 

Fig. 3 Output power as a function of incident pump power for the dual wavelength Tm³⁺:YLF laser system. The fitted 816 nm slope is 46%.

Download Full Size | PDF

To investigate operation at higher temperatures, the laser was run at full pump power (1 W) while monitoring temperature and output signal as the liquid nitrogen evaporated. Lasing was achieved for temperatures up to 200 K, however this does not represent a physical limit to the laser operation. As mentioned previously, cryogenic operation lowers threshold by depopulating the lower laser level and increasing the absorption and stimulated emission cross sections. It is expected that with the use of higher-power pump sources (AlGaAs diode arrays, for instance), the laser can be run at or near room temperature, albeit with higher threshold power than in this proof-of-principle demonstration.

4. Summary

Efficient cw laser operation of the NIR transition in Tm³⁺:YLF at 816 nm has been demonstrated. This is an improvement over a previous demonstration in Tm:YAG which required quasi-cw pumping to mitigate population trapping at the ³F₄ manifold. Trapping mitigation in these experiments was mitigated by cryogenic cooling, use of a YLF crystal host to reduce nonradiative decay rates and the use of a dual-wavelength resonator to continuously quench the trapped population by stimulated emission on the ³F₄ – ³H₆ transition. Bulk crystal systems have a clearer pathway to scaling and optimization than do previously demonstrated non-crystal hosts. The ability to operate crystal lasers continuously on this transition expands their potential for use as high peak and average power sources.