Vertical external-cavity surface-emitting lasers (VECSELs) are ideal sources for the generation of near diffraction-limited beams [1] in the multi-watt regime [2]. For high output power operation, several techniques were put into practice such as in-well pumping, pump recycling, or optimizing the gain structure design [3–7]. The development of high-power VECSELs was further advanced by using an intracavity heat spreader, such as silicon carbide (SiC) [8] or diamond [9]. Efficient single-side cooling of the gain region could be enabled in this way. The distributed Bragg reflector (DBR) behind the gain mirror consists of multiple semiconductor pairs in order to provide a sufficiently high reflectivity, especially when only low refractive index contrast materials are available. As a result, DBRs are typically a few micrometers thick and have a comparatively small thermal conductivity [10–12]. Therefore, it is beneficial to use an alternative laser architecture in which the gain region is operated in transmission and the DBR function is taken by an external mirror; this allows implementing the double-side cooling for the gain region as recently proposed for membrane external-cavity surface-emitting lasers (MECSELs) [13–15].

The first MECSEL incorporating two diamond heat spreaders was successfully demonstrated in the red spectral range in 2016 [16] and various types of MECSELs were realized [17–20] shortly after. In addition to the more efficient heat dissipation, one further advantage can be seen from the epitaxial point of view. Without the DBR, only lattice matching between the gain region and the substrate needs to be considered, which gives more freedom to various semiconductor material systems.

VECSELs in the 8XX nm wavelength region have been realized for 830 nm to 870 nm [21,22], also with passive mode locking [23] or linewidth narrowing [24,25]. On the short 8XX nm range, a spectral gap was left (illustrated as red area in Fig. 1) although there should be no fundamental limits to realize VECSELs in this wavelength regime.

In this Letter, we present a MECSEL extending the wavelength coverage in the 810 nm to 830 nm filling the gap. With a tuning range of more than 20 nm, such a laser source opens new perspectives in numerous fields. These include biology research, spectroscopy, and metrology, e.g., in water-vapor differential absorption lidar [26], in which a good beam quality is required. The external cavity and optical pumping yield a strong benefit and enable power scaling [27] in a MECSEL with the adjustable pump spot diameter and cavity mode area on the gain membrane. Our investigations focus on two aspects: power scaling and thermal lensing in a MECSEL.

Our MECSEL structure was fabricated by molecular beam epitaxy using a V80H-10 VG Semicon solid source reactor. On top of a 50.8 mm GaAs $ (100) \pm {0.5^ \circ } $ substrate, a GaAs buffer layer was deposited, followed by a 150 nm thick AlAs process layer. The active region had a resonant design for an emission wavelength of 825 nm. It contained nine GaInAsP quantum wells (QWs) that were equally allocated to three groups. GaInP was used for the barrier/spacer layers to achieve a sufficient confinement of the charge carriers. A 20 nm thick AlGaInP window layer enclosed the active region on both sides to prevent electron diffusion to the semiconductor–heat spreader interface. For efficient thermal management, the substrate and the AlAs layer were removed via wet chemical etching [16], and the gain membrane was bonded between a pair of uncoated 4H-SiC heat spreaders. Due to its lower thermal conductivity, SiC is not favorable to diamond. SiC was used here because of its lower cost. A schematic illustration of the experimental setup is shown in Fig. 2. The gain membrane sandwich was mounted to a copper heat sink with indium foil to improve the thermal contact. The heat sink plates were cooled by water/glycol, while the temperature was set and stabilized to 20°C via thermoelectric cooling. A Coherent V-18 laser emitting at 532 nm was used to optically pump the gain membrane from two sides at an incident angle of 14°. The resulting pump spot was slightly elliptical, and its diameters in tangential ($ {D_{{\rm p, \tan}}} $) and sagittal planes ($ {D_{{\rm p,sag}}} $) were calculated with the measured distance between the lens and the membrane. To obtain the absorbed pump power, the reflected and transmitted powers were subtracted from the incident pump power. The pump absorption for the gain media was about 70% and the reflection about 26% of the incident pump power.

 

Fig. 1. Realized VECSELs and MECSELs between 785 nm and 855 nm at heat sink temperatures $ {T_{{\rm hs}}} \ge {10^ \circ }{\rm C} $.

Download Full Size | PDF

 

Fig. 2. Schematic illustration of the experimental setup. The used configurations are listed in Table 1.

Download Full Size | PDF

We used two v-shaped cavities (cavity 1 and cavity 2) to characterize the MECSEL at different pump spot and cavity mode diameters. Each cavity was composed of laser mirrors M1, M2, and M3 with the radius of curvature $ r $ and the transmission $ T $ as listed in Table 1. $ {L_1} $ denotes the distance between M1 and the membrane, $ {L_2} $ between membrane and M2, and $ {L_3} $ between M2 and M3. Accordingly, the cavity was adjusted close to its stability limit, and the membrane was positioned at the beam waist of the cavity. In the small pump spot configuration (cavity 1), the mode diameter ($ {D_{{\rm m, \tan}}} $) was about the same size as the pump diameter in the tangential plane. This was different in the large pump spot configuration (cavity 2). Here, the pump spot diameter was adjusted at the highest deliverable pump power of 18.5 W, and an optimum for high output power was found when the cavity mode diameter was smaller than the pump mode with a ratio of about $ 0.66 \pm 0.14 $. This value is in very good agreement with the simulated values of Laurain et al. [28].

The output power characteristics of the MECSEL using a pump spot diameter of $ (88 \pm 4)\,\, {\rm \unicode{x00B5}{\rm m}} $ (cavity 1) and $ (209 \pm 6)\,\, {\rm \unicode{x00B5}{\rm m}} $ (cavity 2) are compared in Fig. 3. The output power was measured behind M3. In the case of the large pump spot diameter (cavity 2), a significant amount (about 13.3%) of the total output power was coupled out through M1. A possible reason is that the transmission of M1 ($ r = 250\,\, {\rm mm} $, $ T \lt 0.2 \% $) in cavity 2 was higher than the value given in the data sheet. Thus, the output from M1 (cavity 2) was always taken into account and added to the total output power, whereas in cavity 1, the output power was not measurable behind M1 ($ r = 100\,\, {\rm mm} $, $ T \lt 0.2 \% $).

Table 1. Cavity Configurations with the Radius of Curvature $ r $ and Transmission $ T $ Given in the Data Sheet^a

View Table

 

Fig. 3. Performance of the MECSEL operated in cavity 1 and 2 at $ {T_{{\rm hs}}}{ = 20^ \circ }{\rm C} $. The inset shows the emission spectrum at 8 W absorbed pump power.

Download Full Size | PDF

In the small pump spot configuration (cavity 1), the threshold was reached at an absorbed pump power of 0.77 W. It can be seen that the output power increased linearly, and the maximum output power of about 0.72 W was attained with 4.52 W of absorbed pump power before thermal rollover. In the large pump spot configuration (cavity 2), the threshold increased to 2.83 W and the maximum output power was doubled to 1.4 W. The differential efficiency indeed decreased slightly from 19.4% to 17.7%, but thermal rollover started at a much higher absorbed pump power of 11.45 W, as heat load generated by the pump was spread over a larger area.

A typical emission spectrum of the MECSEL, measured by a StellarNet BLUE-Wave miniature spectrometer, is illustrated in the inset in Fig. 3. It can be seen that the emission is centered at $ {\sim} 825\,\, {\rm nm} $. For tuning the wavelength, a 2 mm thick birefringent filter without temperature control was placed at Brewster’s angle between M2 and M3. By rotating the birefringent filter around the axis normal to its surface, the lasing wavelength could be tuned between 810 nm and 832 nm, as shown in Fig. 1.

For the estimation of internal losses and laser gain in cavity 2, further output characteristics with outcoupling mirror transmission 1.0% and 0.1% were recorded and are displayed in Fig. 4. By including the transmission of M1, the total transmission $ {T_{{\rm M1} +\,\, {\rm M3}}} $ from M1 and M3 corresponded to 1.4% and 0.7%, respectively. In the $ {T_{{\rm M3}}} = 2.5 \% $ configuration, $ {T_{{\rm M1} + {\rm M3}}} $ was about 2.9%.

 

Fig. 4. Performance of the MECSEL with different total transmission values $ {T_{{\rm M1} + {\rm M3}}} $ in cavity 2. The inset displays the output power at 8 W absorbed pump power for different $ {T_{{\rm M1} + {\rm M3}}} $.

Download Full Size | PDF

The inset in Fig. 4 illustrates how the performance of the MECSEL varied for all outcoupler transmission values at 8 W absorbed pump power. Two data points with zero output $ {P_{{\rm out}}} = 0\,\, {\rm W} $ were additionally included to an outcoupler transmission of $ {T_{{\rm M1} + {\rm M3}}} = 0 \% $ and $ {T_{{\rm M1} + {\rm M3}}} = 5.4 \% $. The latter case was tested in the setup and no lasing occurred due to the high outcoupling losses. By setting the threshold to $ {T_{{\rm M1} + {\rm M3}}} = 5.4 \% $, realistic values for the unsaturated gain $ {g_0} $ per pass were obtained, which was not the case for $ {T_{{\rm M1} + {\rm M3}}} $ much lower than 5.4%. A fit function $ {P_{{\rm out}}} = C \cdot T[{g_0}/({L_{\rm i}} + T) - 1] $ [29], with a fit parameter $ C $, was used to estimate the unsaturated gain $ {g_0} $ per pass and the internal losses $ {L_{\rm i}} $ arising from absorption, scattering, and diffraction. $ {g_0} $ resulted as $ (12.1 \pm 2.7)\% $ and $ {L_{\rm i}} $ as $ (6.8 \pm 2.7)\% $. The fit curve reveals an optimum at $ {T_{{\rm M1} + {\rm M3}}} = 2.3 \% $. This relatively small value can be related to the small amount of QWs.

The beam quality was studied in cavity 2 with a Thorlabs CCD Camera Beam Profiler BC106N-VIS/M, which was placed behind M1 at a fixed distance. Due to the v-shape of the cavity, the beam profile was elliptical, as shown in Fig. 5. It turned into a nearly circular and smaller spot when the pump power was increased. On the one hand, this change is partly initiated by the growth of the mode diameter with increasing pump power, as the pump laser possesses a Gaussian power distribution. On the other hand, we suspect the thermal lensing setting in [30,31]. The thermal lens can result from heating the gain medium and the heat spreaders. In this case, the pump laser as a Gaussian heat source creates a nonuniform temperature distribution. Associated with the thermo-optical effect, a refractive index gradient effectively acting as a lens is created. Although the thermal lens from the gain medium and the heat spreaders cannot be distinguished in the following $ {M^2} $ measurements, the thermal lens from the gain medium is expected to take the dominating part [31]. This is because the thermo-optical coefficient in semiconductor materials is typically up to two orders of magnitude larger than in dielectric materials. Furthermore, pump power absorption leads to higher temperature rise in the gain medium. To further investigate the effect of thermal lensing, $ {M^2} $ measurements were performed with a Thorlabs $ {M^2} $ Measurement System M2MS. The $ {M^2} $ value of about 1.1 in the tangential plane did not change substantially, as shown in Fig. 6(a). A decrease in the $ {M^2} $ value in the sagittal plane is visible between laser threshold and 7 W absorbed pump power. This change is a matter of the thermal gradient, which gets stronger in this pump power regime. However, the $ {M^2} $ value increases when the absorbed pump power exceeds 7 W. It should be noted that there might be a loss of the thermal gradient because the high temperature area on the membrane grows with absorbed pump power. As the beam parameter product remained nearly unaffected, it was consistent that the beam divergence angle decreased from 2.5° to 1.1°, and the beam waist diameter created inside the $ {M^2} $ measurement system enlarged from 33 µm to 71 µm, as shown in Fig. 6(b).

 

Fig. 5. Beam profile at 3.4 W absorbed pump power with cavity 2 (left). The pump power was increased by a factor of two (center) and three (right).

Download Full Size | PDF

 

Fig. 6. (a) $ {M^2} $ and (b) beam waist diameter of the outcoupled beam in the $ {M^2} $ measurement system. (c) Calculated dioptric power of the thermal lens.

Download Full Size | PDF

The dioptric power of the thermal lens was estimated by a ray matrix algorithm that simulates the mode diameter of the Gaussian beam within the MECSEL cavity [32,33]. A thin biconvex lens in the simulation was directly positioned in front of the SiC heat spreader. It was assumed that the thin lens changes the intracavity and the external beam divergence angle by the same factor. For the external beam, we used the beam divergence angle near threshold at which the thermal lensing effect was weakest as a reference. The resulting dioptric power of the thermal lens is illustrated in Fig. 6(c) and saturated at about $ (10.3 \pm 0.1) \,{{\rm m}^{ - 1}} $. This reveals that the thermal gradient does not alter much before thermal rollover.

To clarify whether the thermal lens can be attributed to the thermo-optical effect, we calculated the dioptric power by considering the refractive index rise at the optical center of the lens. The thermo-optical coefficient for identifying the refractive index rise was indeed not measured, but for simplification, it was assumed that the gain membrane consists only of GaInP with a linear thermo-optical coefficient of $ {\sim} (2.0 \pm 0.3) \cdot {10^{ - 4}} \,\,{{\rm K}^{ - 1}} $ [34]. The temperature rise from laser threshold to thermal rollover was about $ \Delta T = 46.83\,\, {\rm K} $ according to the calculated thermal resistance of $ {R_{{\rm th}}} = 4.09\,\, {\rm K/W} $ originating from spectral shift measurements of the MECSEL emission [35]. The refractive index value at the optical center of the lens was calculated by summing up the refractive index of GaInP and the refractive index rise. For the refractive index profile, the following assumptions were made: according to the Gaussian heat source, 87% of the heat is concentrated mainly within the full width at half maximum [30] allowing quadratic approximations for small radial distances. This relation was adapted to describe the refractive index in variation with distance from the optical center. This yields a dioptric power of about $ (8.4 \pm 2.2) \,\,{{\rm m}^{ - 1}} $, which is on the same order of magnitude as in the experimental findings. In the future, the appearance of thermal lensing with different pump spot diameters, mode ratios, heat spreader materials, and thicknesses [31] could be analyzed, which could be important for future semiconductor disk laser development. Also relevant are separate thermal lensing effects from a single standing membrane or the heat spreader itself.

In conclusion, a room temperature operating MECSEL providing wavelength coverage from 810 nm to 832 nm was demonstrated. The near diffraction-limited $ {M^2} $ value remained nearly unchanged in the presence of thermal lensing. The power scalability of a MECSEL was shown by increasing the cavity mode and pump spot diameter. The results revealed only a small decrease in the differential efficiency, and in turn, the maximum output power could be doubled to 1.4 W. The pump spot diameter we used was about 209 µm. This is lower than what is usually used in VECSELs, and it can further be increased. Moreover, thicker SiC or diamond heat spreaders and an outcoupling mirror with optimal transmission value can be used to achieve even higher output power. Additionally, the pump laser should not be necessarily a diffraction-limited diode-pumped solid-state laser. High-power laser diodes in combination with the technique of power scaling can be applied to further push the limits of this new category of semiconductor lasers to higher powers with much lower cost.