1. Introduction

High-average-power lasers employing geometry of thin-disk (TD) gain media have been transformed to robust and reliable photon sources emitting high-power beams from continuous wave (CW) to femtosecond pulses. Compared to other approaches, TD layout has several advantages, such as excellent heat extraction from the gain medium, which is in case of widely used Yb:YAG crucial to prevent re-absorption of laser photons and excitation of electrons back to upper laser level. Another important benefit is minimizing of the wavefront distortion effect caused by thermally-induced changes of refractive index (thermal lensing) due to thermal gradient dominantly in axial direction of TDs. All these privileges enable generation of CW [1] and pulsed [2] kW-class high-quality laser beams. One of problems limiting further scaling is a quality of thin-disk gain modules. For instance, material expansion caused by elevated temperature in the pumped area and mismatch of thermal expansion coefficients of TD and the heatsink materials result in bending and pump power-dependent change in their radius of curvature measured across the pump region [3]. This behaviour can affect order and stability of resonator modes, which can limit next power scaling of TD-based amplifiers as well as laser beam quality, which is crucial for many high-tech applications like, for example, diffractive optical element based multi-beam micromachining [4]. To partially suppress the impact of thermal effects, it is necessary to investigate techniques allowing better heat management of TD lasers. This includes optimization of the cooling scheme and geometry, and investigation of new and more efficient materials for the gain media, heatsinks and coatings [5].

In this paper, we present lasing performance and heat management of a novel epoxy-free TD gain module adopting crystalline-coating on the back side of an Yb:YAG single crystal. The crystalline coating acts a role of both a highly reflective (HR) mirror and an adhesive layer bonding the gain medium to a heatsink simultaneously. The technology of crystalline coatings adopts molecular beam epitaxy to grow crystalline multi-layers of high-index GaAs and low-index Al_0.92Ga_0.08As [6]. This method has already been used in several laser components and applications, such as SESAM [7], VECSEL [8] and others [9–10]. In contrast, currently the most frequently used technology is based on thin amorphous layers of oxides deposited by a physical vapour deposition process such as ion-beam sputtering (IBS). Such amorphous coatings have a thermal conductivity much lower than Yb:YAG or heatsink, such as <2 W m⁻¹ K⁻¹ for HfO₂ or Ta₂O₅ used for IBS-created layers [11]. In comparison, crystalline coatings have superior characteristics, in particular a higher thermal conductivity > 55 W m⁻¹ K⁻¹ [12], lower optical losses, and excellent surface roughness [13]. The TDs are frequently bonded onto a heatsink using an additional epoxy layer or a solder layer. Heat extraction is then strongly affected by the quality of the layer and its thermal conductivity. Epoxies used for gluing of gain media to heatsinks have extremely low thermal conductivity (units of W m⁻¹ K⁻¹), and the whole coating-epoxy structure creates a bottle-neck in efficient heat extraction from the gain module. On the other hand, these techniques are tolerant to surface imperfections such as scratches and digs (S/D) [4,14]. Since the crystalline coating technology allows for reaching excellent quality surfaces with low surface roughness value, the disk can be bonded by method called direct bonding to a polished heatsink substrate. No additional interlayers are required, and this method provides better heat extraction from the TD thanks to absence of epoxies or other chemicals. As for the IBS coated disks, direct bonding is hardly achievable in high quality without any defects because of the low surface quality of IBS coatings, as demonstrated in the past [15].

Frequently used heatsink materials are copper, copper-tungsten, or lately chemical vapor deposited (CVD) diamond [5,16,17]. The materials are distinguished by high thermal conductivity and low thermal expansion. However, the high-quality surface important for direct bonding can be difficult to achieve because of the hardness or grain structure of these materials. We demonstrate this process for Yb:YAG and silicon carbide (SiC) heatsinks. SiC having 490 W m⁻¹ K⁻¹ thermal conductivity [18] was selected as the heatsink material due to significantly higher thermal conductivity than Yb:YAG with <13 W m⁻¹ K⁻¹ [19] and thermal expansion coefficient close to YAG crystals [20], and hardness and chemical properties suitable for surface polishing to flatness and roughness suitable for direct bonding of crystalline mirror-coated TD gain media. To the author‘s knowledge, this is the first time the lasing performance of crystalline coated TD was tested.

2. Experimental setup

In this experiment, a 7 at.% Yb:YAG disk with a diameter of 11 mm and a thickness of 220 µm was used. The front side was coated with IBS antireflective (AR) layers, and the back side of the disk was coated with the commercialy produced crystalline HR coatings with the thickness of several µm. Both coatings were designed for spectral range from 940 to 1040 nm. The TD was contacted onto the SiC substrate with a diameter of 14 mm and a thickness of 3 mm by the direct bonding technique. A structure of the final TD module is shown in Fig. 1(a).

 

Fig. 1. (a) Scheme of crystalline-coated TD bonded onto the SiC heatsink. b) Experimental setup of the in-situ Yb:YAG TD measurement.

Download Full Size | PDF

We tested the module in two different setups to verify the effect of different pumping wavelengths because of the steep increase of the absorption losses of GaAs/Al_0.92Ga_0.08As layers in range from 900 to 1000 nm [6]. The first setup consisted of a TD head allowing for 36 passes through the gain medium with a pump region of about 4.1 mm in diameter (FWHM). The disk was pumped by a commercially available fiber-coupled laser diode (LD) emitting at 940 nm with maximum available pump power of 615 W in CW. The second setup was built with a 24-passes TD head arrangement with a pump region of about 5.1 mm (FWHM), and a fiber coupled LD providing pump light at 969 nm (zero-phonon line of Yb:YAG) with the highest available CW power of 2.3 kW. In both cases, the back side of the SiC heatsink was sprayed with cooling water maintained at 18°C through a nozzle with flowrate of about 3.6 l min⁻¹. For the experiment, TD head was inserted into a simple V-shaped multimodal cavity consisting of a concave mirror with radius of curvature of 6 m and a flat output coupler (OC) with transmittance of 5%, respectively. A thermal imaging camera FLIR A65 was used to determine the temperature distribution in the TD surface pump region in-situ. We also constructed a Michelson interferometer using a 533 nm probe beam to evaluate the thermally-induced TD deformation during the laser operation. Since the thin disk has a small wedge in order of several minutes, we used the beam partially reflected off of the AR surface for the measurement which was separated from the other reflections in propagation. The entire experimental setup is shown in Fig. 1(b).

For comparison, an 8 at.% doped Yb:YAG disk with a diameter of 12 mm, a thickness of 200 µm and IBS HR and AR coatings was tested with the same setup pumped at 969 nm. This disk was bonded via an epoxy layer with thickness of approximately 1 µm onto the identical SiC heatsink like the crystalline-coated disk.

3. Experimental results

Using a 940 nm laser diode for pumping, the laser with crystalline-coated disk reached 370 W of output power and optical efficiency of 60% at pump power of 615 W and corresponding pump intensity of 4.5 kW cm⁻² in the first pass through the thin disk. In case of pumping at zero phonon line (969 nm), we reached 665 W of output power and optical efficiency of 58% at the maximum pump power of 1.15 kW corresponding to pump intensity up to 5.5 kW cm⁻². Both power characteristics and optical - optical efficiency curves are shown in Fig. 2.

Higher efficiency at 940 nm pumping is reached due to a different pump scheme with higher number of pump beam passes, which allowed us to reach higher absorption of pump radiation and as well.

We also monitored surface temperature of the TD for both pump wavelengths. Figure 3 shows the maximum temperature on the disk surface as a function of incident pump intensity. With 940 nm pump radiation wavelength, the maximum temperature of 113 °C was reached at 4.5 kW cm⁻² and 615 W. In contrast, the highest temperature achieved in case of the 969 nm pumping was only 107 °C at 5.5 kW cm⁻² and 1.15 kW. A pumped area with a diameter of 4.5 mm was analyzed by the in-house developed Michelson interferometer to determine a dioptric power in the case of the unpumped and zero phonon line-pumped TD. The unpumped crystalline-coated disk was optically flat with an initial dioptric power of almost 0.012 m⁻¹. With increasing pump power, the disk becomes more convex up to a dioptric power of approximately 0.057 m⁻¹, as shown in Fig. 3. These values correspond to a change of radius of curvature from roughly 200 m down to 35 m, which is still considered a flat disk for many linear cavities of ultrashort pulse lasers. Nevertheless, in case of longer linear or ring cavities this change is significant and should be consider for their designs.

 

Fig. 2. Output power and efficiency characteristics of the crystalline-coated and directly bonded in comparison to IBS coated and epoxy bonded Yb:YAG disks pumped at 940 and 969 nm.

Download Full Size | PDF

 

Fig. 3. Maximal temperature and dioptric power of the crystalline-coated thin disk directly bonded to SiC heatsink, and comparison with and IBS coated epoxy bonded Yb:YAG disk, both for 940 and 969 nm pump wavelengths.

Download Full Size | PDF

As shown in Fig. 2, the epoxy bonded disk reached only 605 W of output power and optical efficiency of 53% at 5.5 kW cm⁻² of incident pump intensity. Similarly, its maximum temperature measured in the center of a pumped area was higher by approximately 18°C compared to the directly bonded one. Regarding the deformations, the unpumped disk was optically flat with a dioptric power of 0.002 m⁻¹ but with rising pump power, it is becoming convex up to a dioptric power above 0.18 m⁻¹, which corresponds to the change in radius of curvature from 248 m to 11 m. This additional deformation compared to the crystalline-coated disks happened most likely due to the thermal expansion of the epoxy layer. To evaluate the effect of different parameters of the disks, we measured the absorbed pump power after 12 pump passes in case of 969 pumping. The absorbed power was higher of about 3% in case of the epoxy bonded disk; therefore, the disks are considered to be comparable.

4. Numerical simulations

To separate influence of epoxy bonding and direct bonding via crystalline coatings on the performance of the TD, and to find which parameters have a dominant impact on the temperature distribution in the TDs, we have performed numerical simulations of heat extraction from the TD. For the simulation, we have chosen conditions equivalent to the experiment with the 969 nm wavelength and the beam with a flat-top profile with a diameter of 5.1 mm. We have also taken into account slightly different disk thickness and doping concentration (±0.2 at.%) compared to 7 at.% crystalline-coated disk used in some of the experiments. Smaller thickness than 220 µm for the higher doped disks was chosen to compensate the small difference in doping and equalize absorption. The scheme of the gain module including the thermal conductivities of the individual layers and the coating thickness used for the simulations is shown in Fig. 4. The resulting maximal temperature in the TD was obtained from a solution of a two-dimensional, axially symmetric (cylindrical symmetry) heat transport equation using the finite element method. A good matching between the experimentally observed data and the values predicted for directly bonded disk by the numerical model was obtained (Fig. 5) when we have chosen a constant cooling temperature of 30°C on the water-cooled heatsink boundary and the fraction of the absorbed power that is turned into heat equals to 6.7%. This value is mainly consisted of quantum defect which is equal to 5.9% in case of 969 nm pumping and by other mechanisms involved, such as heating cause by the non-radiative transitions and indirectly by the fluorescence or by the amplified spontaneous emission [21].

 

Fig. 4. Scheme of the thin-disk gain module including the individual layers and their properties for (a) crystalline [22] and (b) IBS HR coatings [11].

Download Full Size | PDF

 

Fig. 5. Experimental and simulated thin disk temperature of the crystalline-coated disk directly bonded onto SiC and CVD diamond heatsink, respectively.

Download Full Size | PDF

4.1 Influence of the heatsink

Since commercially available disks are typically bonded onto CVD diamond with > 3 times higher thermal conductivity than SiC (> 1800 W m⁻¹ K⁻¹ for CVD diamond [23]). We used the numerical model to predict TD temperature of the crystalline-coated disk directly bonded onto CVD diamond heatsink with a diameter of 14 mm and thickness of 3 mm and compared the data with the SiC-bonded one. In Fig. 5 is shown that in the case of a diamond heatsink, the disk temperature is lower by roughly 12°C at 1.15 kW of incident pump power. This temperature difference cannot justify in many cases (lower pump power systems) using the >20 times more expensive diamond heatsinks. Although a more detailed numerical analysis of the heat extraction from the TDs including thermo-mechanical parameters will be definitely needed, e.g. to explore impact of changing different geometrical and materials parameters.

4.2 Influence of the coating

We used the numerical model with the same boundary conditions to demonstrate how the temperature profiles differ for the uncoated, IBS-coated and the crystalline-coated TDs with the same geometry, all directly bonded to the SiC and CVD diamond heatsink. As shown in Fig. 6, an uncoated disk results in almost identical temperatures as the crystalline-coated one, which means that the crystalline layers with high thermal conductivity deteriorate the heat extraction negligibly. On the contrary, oxide layers sputtered by the IBS with lower thermal conductivity decrease the heat transfer significantly. At 1.15 kW of pump power, the temperature of the IBS coated TD is higher by about 18°C for both cases of heatsinks. It can be dedicated the temperature gradient is lower in case of CVD diamond heatsink which reduces the effect of thermal lensing. Nevertheless, another study is needed to calculate its dioptric power and other parameters.

 

Fig. 6. Numerical prediction of thin disk temperature cross section in the pumped area of an uncoated, an IBS coated, and a crystalline-coated thin disks, all directly bonded onto SiC or CVD diamond heatsink at different pump powers. Shaded area represents the pump region.

Download Full Size | PDF

5. Conclusions

We presented thermal and power characteristics of crystalline-coated Yb:YAG thin disk directly bonded by this coatings to a silicon carbide heatsink. It has been demonstrated for the first time that the crystalline coatings have great potential to be used for thin disks due to their higher thermal conductivity, lower optical losses, and direct bonding capability compared to the widely used ion-beam-sputtered coatings. Measurements of thermal and thermo-mechanical parameters of thin disks including in-situ measurement of thin disk dioptric power using a Michelson interferometer were performed at both 940 nm and 969 nm pump wavelengths. The disk remained nearly flat even for the pump intensity as high as 5.5 kW cm⁻². Finite element method simulations confirmed the enhanced thermal performance of the crystalline-coated thin disk due to their better thermal properties and demonstrated the potential of silicon carbide heatsinks together with the crystalline coatings. The model predicted almost comparable performance of the silicon carbide bonded disks as much more expensive synthetic diamond bonded ones.