1. Introduction

Optically pumped semiconductor disk lasers (SDLs), also known as vertical external-cavity surface-emitting lasers (VECSELs) or optically-pumped semiconductor lasers (OPSLs) [1], have gained significant attention due to their wavelength flexibility [2, 3], high continuous wave (CW) output powers [4], and good beam quality. Because of the external cavity configuration, they are superior to edge-emitting semiconductor lasers for applications requiring intra-cavity access, such as frequency doubling with non-linear crystals [5, 6], optical parametric oscillators [7], mode locking with saturable absorbers [8], and intra-cavity laser cooling [9].

A typical SDL consists of an active region and a monolithically integrated semiconductor distributed Bragg reflector (DBR), with an external dielectric mirror to complete the cavity [1]. To maintain good material quality, both DBR and active region must be lattice matched to the substrate, a condition not every semiconductor material system may satisfy. Moreover, with relatively small refractive index contrast in semiconductor materials, tens of DBR pairs are often required to achieve high reflectivity, thus resulting in DBRs that are several microns thick. Such thick DBRs having multiple interfaces introduce a large thermal resistance, which hampers the thermal management of SDLs. Wafer fusion [10] has been demonstrated as a solution to the material system problem, by growing the active region and semiconductor DBR separately on different substrates and fusion bonding them together. This approach eliminates the lattice matching requirement between DBR and active region, and has achieved high powers [10, 11], but the DBR remains as the thermal transport bottleneck in these gain structures. Although there is an initial investigation of optically-pumped semiconductor lasers at cryogenic temperatures with only external dielectric mirrors [12], the overall potential of DBR-free SDLs was not fully explored or analyzed. Here, we demonstrate high power, room temperature, and broadly tunable SDLs without the use of semiconductor DBRs.

For a DBR-free SDL, the gain structure is epitaxially grown and separated from the substrate by a thin sacrificial layer, which allows lift-off and subsequent van der Waals bonding to a transparent window or prism. The laser cavity is completed by at least two external dielectric mirrors, allowing the gain structure to operate in transmission or total internal reflection (TIR) geometries. Compared with semiconductor DBRs with commonly available material systems, dielectric mirrors or TIR offer much broader high-reflection band. This property together with a broader integrated gain bandwidth introduced by the DBR-free geometry, enable a relatively large wavelength tunability in DBR-free SDLs. From the thermal management perspective, our modeling further indicates that, by sandwiching the gain structure between a pair of diamond substrates, the DBR-free geometry has the potential to lower the temperature rise in active region and significantly improve power scaling.

2. Laser geometries and performance

Since semiconductor substrates, like GaAs, have relatively low thermal conductivity and often introduce high optical losses, it is essential to perform epitaxial lift-off and bonding for fabricating a high power DBR-free SDL. Following the procedures in [13, 14], the gain chip is protected with black wax on the epitaxial surface, and lifted off from the substrate by lateral etching of an AlAs sacrificial layer using diluted hydrofluoric acid. The sample is rinsed with deionized water and subsequently van der Walls bonded to a high optical quality heatspreader (such as diamond or silicon carbide). The procedure involves applying a controlled amount of pressure, before the black wax is removed and the sample is carefully cleaned. The bonding quality plays a critical role in the laser performance and thermal rollover characteristics. All processing is performed at room temperature in a cleanroom environment.

2.1 Transmission geometry

The 2.4 µm thick active region, as shown in Fig. 1(a), consisted of 8 InGaAs quantum wells (QWs) emitting near 1160 nm forming a resonant periodic gain (RPG) structure inside the strain compensation GaAsP barrier with InGaP window layers. To increase the absorption for the pump laser, the thicknesses of the last two barrier layers were chosen to be λ and 3λ/2 respectively, compared with the standard λ/2 for the remaining barriers. The gain chip was bonded onto a 500 µm thick and 4 mm diameter chemical vapor deposition (CVD) grown single-crystalline diamond (Element Six, Inc.), which was clamped onto a copper block water cooled to 10 °C, with indium foil in between to improve thermal contact. Two curved dielectric mirrors completed the external cavity, a 10 cm radius of curvature (RoC) high reflector (HR) and a 25 cm RoC output coupler (OC). A fiber coupled 25 W 808 nm diode laser was focused down to approximately 200 µm in diameter on the gain chip, at about 20 degree angle of incidence from the diamond side, as shown in Fig. 1(b). The distance between the HR and the sample is 9.2 cm and 24.5 cm between the sample and OC. The cavity length and the overlap between pump and cavity mode on the gain chip were optimized for maximum output power.

 

Fig. 1 (a) Sample design and (b) schematic diagram of the transmission SDL geometry.

Download Full Size | PDF

For the pump laser, 20% of the power was reflected and 13% was transmitted by the active region. Laser threshold was at about 1 W and no thermal roll-over was observed with the available pump power, as seen in Fig. 2(a). No obvious thermal degradation was observed after the high pump power operation. The slope efficiency was 11%, with respect to the incident pump power. We believe the low efficiency was due to Fresnel and scattering losses from the uncoated diamond-air interface. Utilizing an antireflection coated diamond (on one side) or a Brewster’s angle geometry is expected to dramatically improve this. The output beam profile is shown in Fig. 2(a) inset. We did not observe a significant change in the edge photoluminescence of the sample before and after lift-off and bonding, but closer investigation of the bonding quality and possible sample degradation is needed.

 

Fig. 2 Laser characterization. (a) Power conversion graph for the transmission geometry with a linear fit, showing a slope efficiency of 11% (CW). Inset is the beam profile characterized with a silicon CCD camera at 5 W pump power, showing horizontal and vertical profiles with overlaid Gaussian fits. (b) DBR-free SDL wavelength tuning at 10.5 W incident pump power with two HRs. The spectra were collected with a 0.2 nm resolution optical spectrum analyzer.

Download Full Size | PDF

The free-running SDL operates at λ≈1160 nm with multiple modes, caused by the etalon effect of the uncoated diamond. By inserting a 0.5 mm thick quartz birefringent filter (BRF) at Brewster’s angle inside the cavity, the laser wavelength can be tuned from 1118 nm to 1196 nm, corresponding to 78 nm tuning range at 10.5 W incident pump power, as shown in Fig. 2(b). To the best of our knowledge, this is the widest tuning range achieved near 1160 nm for a SDL [3, 6, 15]. With an extra 300 µm thick fused silica etalon, the laser operated with a linewidth less than 0.2 nm, limited by the resolution of the optical spectrum analyzer. The reasons for the observed wide tuning range (with potentially even wider range possible) in DBR-free systems are multifold. First, with the gain structure directly bonded to diamond, the reflectivity at the semiconductor-diamond interface is negligible. This leads to a low finesse subcavity and subsequently a reduced modulation of the gain spectrum. Second, as stated earlier, dielectric mirrors offer wider high reflection band than semiconductor DBRs. Most importantly, with a thin and longitudinally inhomogeneous gain structure like an RPG, the overlap between the cavity standing wave and QWs strongly depends on the position of the RPG in the cavity. We evaluated this so-called “integrated gain factor” [16] for an arbitrary distance z from the node on the cavity end mirror to the first QW. The results reveal that the effective gain bandwidth is significantly broader when the RPG is further from the end mirrors (z>>λ) compared with in the proximity of an end mirror (z = 3λ/4). Assuming 90% of peak gain, with 11 QWs, the calculated bandwidth (potential tuning range) of a DBR-free SDL is 76 nm, compared with 34 nm for a typical VECSEL. Similar gain position dependent laser linewidth phenomena due to spatial hole burning have been reported in solid state lasers [17], which is consistent with this simple picture. With the relatively flat and broad gain spectrum accessible, DBR-free SDLs not only offer a wide tuning range, but also could potentially be further utilized for ultrashort pulse generation and amplification.

2.2 Total internal reflection geometry

The DBR-free gain structure lends itself nicely for exploiting TIR either in standing wave or monolithic ring cavity geometries. As a proof of concept, we implemented a standing-wave cavity, in which the cavity mode experiences TIR at the semiconductor-air interface, when the active region was bonded onto the hypotenuse surface of a right angle fused silica prism, as shown in Fig. 3(a). The RPG structure consisted of InGaP window layers and 12 InGaAs/GaAsP QWs, designed to operate at 1020 nm. The SDL was pumped with a CW Ti:sapphire laser at 810 nm, modulated with approximately 1% duty cycle using a mechanical chopper, to reduce thermal effects because of the low thermal conductivity substrate and lack of active cooling. The pump spot size was 80 µm in diameter on the active region. Two 5 cm RoC HRs completed the cavity. All prism surfaces were uncoated.

 

Fig. 3 Schematic diagrams of pulsed TIR based SDL setup (a) and proposed monolithic ring cavity (b) and (c).

Download Full Size | PDF

With two HRs, the laser output power was low, shown in Fig. 4, and as expected, thermal roll-over occurred at low pump powers. This could be solved by using high thermal conductivity prisms and applying active cooling power from the top, bottom and even part of the side surfaces. The laser operated at 1046 nm near threshold, as shown in the inset of Fig. 4. The much longer than expected lasing wavelength is attributed to gain modulation in the gain structure subcavity, due to relatively larger index mismatch between semiconductors and glass (compared with diamond).

 

Fig. 4 Power conversion graph for TIR geometry. The inset are spectra below and above threshold.

Download Full Size | PDF

TIR based DBR-free SDLs can be compact. With the active region bonded onto a high refractive index (n>2) equilateral prism in Fig. 3(b), a self-aligning monolithic ring cavity can be formed using TIR at the three faces. The pump-induced thermal lens effect in the active region could stabilize the cavity, while laser light could be coupled out evanescently using a prism. With a saturable absorber (semiconductor, graphene, etc.) bonded to another prism surface, a high repetition rate mode locked laser can be envisioned within a small prism. A rectangular prism can equally be used in this application if refractive index is larger than $\surd 2$. Compared with a triangular prism, the rectangular prism has more stringent angle precision requirements as opposing angles should be supplementary. Both prisms require high surface parallelism.

3. Thermal analysis

Due to quantum defect and non-radiative recombination, a significant amount of the pump energy is transformed to heat in SDLs. The temperature rise further increases the non-radiative losses, thus lowering the slope efficiency and leading to thermal rollover. Therefore, as in all lasers, thermal management is essential for power scaling of SDLs.

Here we use commercial finite element software (COMSOL Multiphysics 4.4) to perform and compare thermal analysis for conventional as well as DBR-free gain structures. To simplify the simulation, we assume azimuthal symmetry, with a Gaussian pump beam incident at the center of a circular gain chip, consisting of a 0.44 µm thick InGaP window layer and 2.25 µm thick GaAs active region, which absorbs the pump power, 40% of which is converted to heat. The gain chip is assumed to be 1.6 mm in diameter, and bonded to a 5 mm diameter, 1 mm thick diamond heat spreader, mounted to a copper heat sink with a 2 mm diameter clearance hole using 50 µm thick indium foil. For the standard thin device VECSEL geometry, a 3.7 µm GaAs/AlAs DBR is added, which is assumed to absorb any pump light transmitted by the active region. The VECSEL chip is treated as soldered to a 10 mm diameter 1 mm thick diamond plate, heatsunk to a copper plate (unlike in [18]) using indium foil. Fixed temperature boundary condition is applied to the backside of the copper, and thermal conductivity parameters are taken from [19]. For the thin device VECSEL geometry with two heatspreaders, the simulation conditions are a combination of the above.

The transmission geometry DBR-free SDL is similar to the intra-cavity heatspreader approach, but without the extra heat source in the DBR. Because the cooling power can only be applied to the diamond heatspreader in an annulus, where it is not in the laser path, heat flow is radial (three-dimensional), compared with the mostly one-dimensional heat flow in a traditional VECSEL. As shown in Fig. 5, with the pump power density held constant, by increasing both pump power and spot size, the maximum temperature rise in the gain region for the DBR-free geometry is lower than for the traditional VECSEL geometry. Additionally, this can be significantly improved by placing heatspreaders on both sides of the active region. Using two diamonds lowers the temperature rise further, making it a promising approach for future power scaling, even better than the thin device geometry with intracavity and extracavity diamonds. In the simulation, we assume perfect bonding and neglect the interface thermal resistances. This could influence experimental results and the potential impact is under investigation.

 

Fig. 5 Maximum temperature raise for different geometries as a function of incident pump power with constant pump density (80 kW/cm²). The absorption coefficient for the pump at the gain region is 1000 mm⁻¹ and pump at DBR region is 457 mm⁻¹. For the simulation diagrams, the red lines denote where the fixed temperature boundary conditions are applied.

Download Full Size | PDF

4. Conclusions

We demonstrate CW lasing of a transmission geometry DBR-free SDL with 2.5 W output power and more than 78 nm tuning range around 1160 nm. A proof-of-principle experiment showed pulsed lasing of a total internal reflection based geometry, which opens up the possibility for monolithic ring cavity based SDLs. Preliminary thermal analysis suggested that the double diamond transmission geometry has significant advantages over existing VECSEL geometries in thermal management. DBR-free geometries provide new opportunities for the development of high power and widely tunable SDLs with more material systems.