1. INTRODUCTION

Semiconductor lasers are used in various fields including telecommunications, sensing, and medicine due to their compactness, high efficiency, and high beam quality at low output power. However, it is difficult to obtain high output power while keeping high beam quality, namely high brightness, with semiconductor lasers. This is because, to obtain high power, the lasing area has to be expanded, which leads to lateral-multimode oscillation and deteriorated brightness. Thus, semiconductor lasers have yet to be directly used for applications that need high brightness, such as laser processing and welding. Photonic-crystal surface-emitting lasers (PCSELs) [1–21] pave the way toward achieving high brightness with semiconductor lasers, since a two-dimensional (2D) large-area band edge resonant mode in the photonic crystal is used as the lasing cavity [1–3]. In addition, the photonic crystal can diffract the light in the surface-normal direction, which allows the device to function as a surface-emitting laser source. Very recently, we have employed a double-lattice photonic-crystal cavity to expand the lasing area to a diameter of more than 500 µm, and we obtained an output power of more than 10 W while maintaining a high beam quality under pulsed operation [20]. In addition, we have also conducted early demonstrations of continuous wave (CW) operation of double-lattice PCSELs [20]. However, a detailed investigation of thermal management has not been discussed. Thermal management is essential for CW operation of large-area PCSELs since a great amount of heat becomes concentrated in the resonant area during high power operation. In this paper, we discuss the thermal management of large-area PCSELs, starting with an analysis of temperature under CW operation. Such thermal management allowed us to obtain a high output power of ${\sim}{{8}}\;{\rm{W}}$ under CW operation of a single-chip PCSEL.

2. PCSEL DEVICE STRUCTURE

A schematic diagram of a PCSEL is shown in Fig. 1(a). Note that the schematic is in the epi-down configuration, where the epi-growth direction is shown by an arrow in the figure. 2D photonic-crystal air holes are introduced in the vicinity of the active layer. The photonic crystal and the active layer are sandwiched by $n$-type and $p$-type cladding layers. Light is confined in the vertical direction by cladding layers while interacting with the photonic crystal. We utilized the band edges of the photonic crystal, where the group velocity of light is zero, to form a large-area cavity mode. By using the band edges at the Γ point, the light can be also diffracted in the surface-normal direction. Moreover, new types of double-lattice photonic-crystal cavities can be utilized to realize broad-area emission while maintaining a high beam quality [20,21]. Efficient surface-normal emission can be obtained by introducing asymmetry to the photonic-crystal lattice points, such as a height difference between the two air holes, an asymmetric air hole shape, or a detuned lattice-point separation. In addition, a $p$-type distributed Bragg reflector (DBR) is introduced next to the $p$-cladding layer of the PCSEL to serve as a backreflector. The DBR helps to reflect downward-emitted light upward. Adjusting the phase of the upward-emitted light and reflected light enables us to efficiently extract light from the device. An $n$-type window electrode and a circular type $p$-electrode are formed on a $n$-GaAs substrate and a $p$-capping grown on the $p$-DBR, respectively, for the purpose of current injection. The lasing area is defined by the circular $p$-electrode size. The laser is emitted through the n-GaAs substrate as shown in Fig. 1(a).

 

Fig. 1. (a) Schematic diagram of a photonic crystal laser (PCSEL) in the epi-down configuration, where the epi-growth direction is shown by blue arrows. A 2D photonic crystal is introduced in the vicinity of the active layer, and the photonic crystal and active layers are sandwiched by cladding layers. (b) (left) Top-view scanning electron microscopic (SEM) image of one type of double-lattice photonic crystal after etching; (right) cross-sectional SEM image after air-hole-embedded crystal regrowth. The cross-sectional image shows the elliptical air holes cleaved in the horizontal direction.

Download Full Size | PDF

A detailed explanation of the device fabrication process is as follows. A tertiary-butyl arsine (TBAs)-based metal-organic vapor-phase epitaxy (MOVPE) method was used for crystal growth. First, a $n \text{-} {{\rm{Al}}_{0.7}}{{\rm{Ga}}_{0.3}}{\rm{As}}$ cladding layer, an InGaAs/AlGaAs multiple quantum well (MQW) layer, and a $p$-GaAs layer, which was used to fabricate the photonic-crystal air holes, were grown on a $n$-GaAs substrate. Here, the photoluminescence wavelength of MQW was chosen to be ${\sim}{{935}}\;{\rm{nm}}$. Next, a double-lattice photonic crystal was formed in the $p$-GaAs layer by using electron-beam lithography and reactive ion etching methods. The lattice constant of the photonic crystal was set to be ${\sim}{{276}}\;{\rm{nm}}$, where the wavelength of the Γ-point band edge matches to the emission wavelength of the MQW. A scanning electron microscopic (SEM) image of the double-lattice photonic-crystal air holes after etching is shown in Fig. 1(b). Here we utilized pairs of elliptical and circular air holes with a separation ($d$) of ${0.257}a$ (where $a$ is the lattice constant), which is suitable to realize large-area emission with a high beam quality. Then, a 1-µm-thick $p \text{-} {{\rm{Al}}_{0.4}}{{\rm{Ga}}_{0.6}}{\rm{As}}$ layer was regrown on the photonic-crystal air holes (second growth). We were able to embed the air holes under appropriate regrowth conditions [17,18]. Subsequently, a phase-matching spacer layer of $p \text{-} {\rm{GaAs}}$, 14 pairs of AlGaAs-DBR layers, and a $p +$-GaAs capping layer were grown. The center wavelength of the reflectivity of the DBR was set to be matched with the lasing wavelength. A cross-sectional SEM image after regrowth is also shown in Fig. 1(b). Note that this cross section shows the elliptical air holes cleaved in the horizontal direction. Finally, $n$- and $p$-electrodes were deposited. The $p$-electrodes were circular with diameters of either 800 µm or 1000 µm.

3. THERMAL MANAGEMENT

For the purpose of heat dissipation in the PCSELs under CW operation, we assembled the PCSELs in a water-cooling package using a highly thermally conductive (${\sim}{{1500}}\;{{\rm{Wm}}^{- 1}}\;{{\rm{K}}^{- 1}}$) sub-mount in the epi-down configuration. The water-cooling package consisted of a 1-mm-thick copper baseplate on whose top surface was bonded the 300-µm-thick sub-mount. The bottom surface of the copper baseplate was cooled by flowing water. Efficient heat dissipation was realized by attaching a fin-like structure to the bottom surface of the copper baseplate to increase the surface area over which heat is exchanged. We calculated the operating temperature of the assembled PCSEL by solving the three-dimensional heat equation using a finite-element-method. In our calculations, we considered the detailed PCSEL structure (including all of its layers), the sub-mount, and the solder that is used to bond the sub-mount to the copper baseplate. We assumed that the bottom surface of the copper baseplate is fixed to the temperature of the water (20°C). This assumption is valid provided that a sufficient amount of water is circulated for cooling during experiments. A heat source was placed in the MQW layer. The heat source had a circular shape with a diameter of 850 µm. This diameter incorporates our findings that the current spreads laterally inside the MQW by ${\sim}{{25}}\;\unicode{x00B5}{\rm m}$ on all sides. We first considered that the heat source has a power of 30 W (or a power density of ${{53}}\;{\rm{W}}\;{{\rm{mm}}^{- 2}}$). The calculated 2D cross-sectional temperature profile at this heating power is shown in Fig. 2(a). In addition, the line-profile of the temperature in the middle of the structure [line AA’ in Fig. 2(a)] is shown in Fig. 2(b). It can be seen in Fig. 2(b) that, using our cooling package, the temperature of the MQW layer can be limited to 52.9°C even when considering a high heating power density of ${{53}}\;{\rm{W}}\;{{\rm{mm}}^{- 2}}$. Note that the temperature of the photonic-crystal layer is almost the same as that inside the MQW layer. The temperature of the PCSEL surface was calculated to be 51.2°C, which is 1.7°C lower than the MQW temperature. The maximum rise in temperature across the copper baseplate, the solder, the sub-mount, and the $p$-side of the PCSEL are found to be 11.5°C, 5.0°C, 5.8°C, and 10.6°C, respectively. Thus, the total rise in temperature in the MQW layer with respect to the bottom of the copper baseplate is 32.9°C. The 2D in-plane temperature distribution in the plane of the MQW layer for a heating power of 30 W is shown in Fig. 2(c). The 1D temperature profile along a line bisecting the circular heating area of the MQW layer [line BB’ in Fig. 2(c)] is shown in Fig. 2(d). Temperature profiles at additional heating powers of 10, 20, and 40 W are also shown in the same figure. It is clearly seen that the increase in temperature is greatest within the circular heating area. The temperature distribution across this area is flattened, owing to the highly thermally conductive sub-mount used for heat dissipation. Even when the heating power is increased to 40 W (or the power density is increased to ${{70}}\;{\rm{W}}\;{{\rm{mm}}^{- 2}}$), the temperature of the MQW is kept under 64°C. From these calculations, we found that the heat resistance of this configuration is ${1.1}^\circ {{\rm{CW}}^{- 1}}$.

 

Fig. 2. (a) Calculated cross-sectional temperature profile of the PCSEL, bonded to the copper baseplate via the sub-mount, for heating power of 30 W. (b) Cross-sectional temperature profile along a line corresponding to AA’. Distance is measured from the bottom of the copper baseplate. (c) In-plane temperature distribution in the plane of the MQW layer. (d) Temperature profile along a line corresponding to BB’ using heating powers of 10, 20, 30, and 40 W.

Download Full Size | PDF

4. CW CHARACTERISTICS OF THE PCSEL

In our experiments, we evaluated the CW characteristics of a 800-µm-diameter double-lattice PCSEL, which had a slope efficiency of ${\sim}{0.45}\;{\rm{W/A}}$. The slope efficiency was limited by a small radiation constant and an unoptimized phase-matching spacer layer. (Discussions on improving the slope efficiency to ${\gt}{0.8}\;{\rm{W/A}}$ will be provided elsewhere in the near future.) As discussed above, for heat dissipation under CW operation, we assembled the PCSEL in a water-cooling package using a highly thermally conductive sub-mount. This package consisted of a 1-mm-thick copper baseplate, upon which a sub-mount can be mounted, and a fin-like structure for improving the rate of heat exchange between the package and the water, as explained above. A photograph of the assembled PCSEL and package is shown in Fig. 3(a), where the inset is an enlarged photograph around the PCSEL. This enlarged photograph shows multiple contact wires bonded to the $n$-side window electrode. The back-side ($p$-side) contact electrode was taken via the sub-mount. We note that care was required to bond the PCSEL to the sub-mount without forming any voids, which would weaken the thermal conductivity and create a non-uniform temperature distribution. To avoid the formation of voids, we ensured that the PCSEL chip and sub-mount were parallel to each other during their bonding, and we scrubbed the sub-mount on the solder, which was melted on the package. To evaluate the temperature of the PCSEL after assembly under CW operation, we observed the surface temperature using a thermographic camera, as illustrated in Fig. 3(b). Thermal emission emitted from the surface of the device within the wavelength range of 8–14 µm was measured. A germanium filter was used to block the near-infrared laser beam emitted from the device while allowing the mid-infrared thermal emission to pass through. The thermographic temperature was calibrated by observing a PCSEL placed at a heater, where the temperature of the PCSEL was precisely controlled and measured by a thermo-couple. A thermographic image at an injection current of 10 A (equivalent to a heating power of 20.5 W) during cooling by water at a temperature of 20°C is shown Fig. 3(c). The color bar indicates the calibrated temperature. It is seen that the temperature increased within an area corresponding to the current injection area of the PCSEL. We measured temperatures at different injection levels and plotted these temperatures with respect to the heating power as red circles in Fig. 3(d). Note that the heating power in this plot was calculated by subtracting the output light power from the input electric power. The calculated surface temperatures of the PCSEL for several heating powers are also shown as connected green circles in the same figure. These calculations were in good agreement with the experimental results. The difference between calculation and experiment is found to be ${\sim}{1.5}^\circ {\rm{C}}$. Thus, we conclude that we were able to assemble the PCSEL without forming any voids. These results strongly indicate that the surface temperature of the PCSEL can be kept under ${\sim}{{62}}^\circ {\rm{C}}$ even at a high heating power of 40 W.

 

Fig. 3. (a) Photograph of the PCSEL assembled in the water-cooling package via a sub-mount. Several wires are bonded to the $n$-side ring electrode, while the $p$-side electrode is bonded to the sub-mount. (b) Schematic of the setup used to measure the device surface temperature during CW operation. The temperature was measured using a thermographic camera. (c) Thermographic image of the device operating at a CW current of 10 A. (d) Measured and calculated device surface temperatures as a function of heating power.

Download Full Size | PDF

The light–current (LI) curves and the current–voltage (IV) curves of the assembled PCSEL under CW operation and cooling by water at temperatures of 20°C and 5°C are shown in Fig. 4(a). The LI curves indicate that the threshold current was 3 A (current density ${0.6}\;{\rm{kA}}\;{{\rm{cm}}^{- 2}}$) and the slope efficiency was 0.45 W/A when the temperature of the water was 20°C, and also that the threshold current fell to 2.8 A and the slope efficiency rose to 0.5 W/A (up to an injection current of 9 A) when the temperature of the water was decreased to 5°C. A slight roll-over is observed in the LI curve for 20°C water at high injection currents exceeding 14 A, while no such roll-over is observed in the LI curve for 5°C water, even at higher injection currents. Owing to this improvement, we were able to achieve a high output power of ${\sim}{{7}}\;{\rm{W}}$ from a single chip with 800-µm-diameter resonator when cooled with 5°C water. Very recently, we have also achieved ${\sim}{{8}}\;{\rm{W}}$ of CW output power from a 1000-µm-diameter single-chip PCSEL (not shown). These are the highest CW output powers ever reported using single-chip PCSELs. We note that a slight kink was observed at ${\sim}{{9}}$ A in the LI curve. This kink was particularly visible when the water temperature was 5°C. From the LI and IV curves in Fig. 4(a), we calculated that the heating power at an injection current of 16 A (the maximum injection current for 20°C water) was 39.4 W. Following our previous discussion, the MQW/photonic crystal temperature at this heating power for a water temperature is 20°C is estimated to be ${\sim}{{63}}^\circ {\rm{C}}$. From this estimation, it is apparent that the roll-over commenced when the MQW temperature exceeded ${\sim}{{63}}^\circ {\rm{C}}$. In Fig. 4(b), we show the lasing spectra at different injection currents for a water temperature of 20°C. For these measurements, we used a spectrometer with a liquid nitrogen-cooled Si-CCD detector. The wavelength resolution of our setup was ${\sim}{{8}}\;{\rm{pm}}$. It is seen in the spectra of Fig. 4(b) that, at injection currents of 8–10 A, a single emission peak with a resolution-limited linewidth is obtained. At other injection currents, additional emission peaks were observed. These results indicate that a CW output power of ${\sim}{{4}}\;{\rm{W}}$ from the oscillation of a single laser mode was achieved. The loss of single-mode oscillation is considered to be due to a temperature distribution in the photonic-crystal cavity [as calculated in Fig. 2(d)], which induces a refractive index distribution (wherein the refractive index is higher in the middle of the cavity than the edges). Such a refractive index distribution would cause the photonic bandgap to become spatially dependent, which would affect the properties of lateral confinement of the lasing mode. The kink in the LI curve in Fig. 4(a) is suspected to be due to mode-switching caused by such a refractive index distribution.

 

Fig. 4. (a) Light–current (LI) and current–voltage (IV) curves of the PCSEL under CW operation at cooling water temperatures of 20°C and 5°C. A CW output power of ${\sim}{{7}}\;{\rm{W}}$ was achieved from the single-chip PCSEL with a 800 µm diameter. (b) Lasing spectra at different injection currents, measured at a cooling water temperature of 20°C.

Download Full Size | PDF

As seen in the spectra of Fig. 4(b), the emission wavelength was redshifted as the injection current rose. This shift is due to a temperature-induced rise of the refractive index of the photonic crystal layer. Thus, we can estimate the temperature within the photonic-crystal layer from the shift of wavelength. In a separate experiment, we heated the PCSEL using an external heater, then measured the emission wavelength of the PCSEL under pulsed operation as a function of temperature. From this experiment, we found that the shift of the emission wavelength with respect to the temperature was ${0.086}\;{\rm{nm}}\;{{\rm{K}}^{- 1}}$. From this value, we estimated the temperature of the photonic-crystal/MQW as a function of the heating power; our estimates are shown as blue circles in Fig. 5. The connected gray circles in the same figure indicate temperatures at the center of the photonic-crystal/MQW region, calculated using the method discussed in the previous section. Our calculations are in good agreement with the experimental results. The difference between calculation and experiment is found to be ${\sim}{1.1}^\circ {\rm{C}}$. We note that Fig. 3(d) shows the device surface temperature, while Fig. 5 shows the temperature inside the photonic-crystal/MQW region, which is slightly higher than the surface temperature, as discussed above. These experimental results also verify that the temperature of the MQW layer can be kept under ${\sim}{{64}}^\circ {\rm{C}}$ even at a high heating power of 40 W.

 

Fig. 5. Photonic crystal/MQW temperature estimated using the wavelength shift at different heating powers [based on the spectra in Fig. 4(b)]. Calculated temperatures are also shown.

Download Full Size | PDF

 

Fig. 6. (a)–(d) Calculated power consumption of 800-µm-diameter PCSELs as a function of injection current above the threshold, for slope efficiencies ranging from 0.25 to 1 W/A. Red, green, blue, and orange colors represent the laser output power, unwanted cavity losses, threshold power loss, and ohmic loss due to the series resistance of the device, respectively.

Download Full Size | PDF

5. POWER CONSUMPTION AND DISCUSSION

Here, we estimate the power consumed by the PCSELs during CW operation. Power consumption is broadly categorized into four parts: (1) laser output power; (2) power loss due to unwanted cavity losses such as intrinsic absorptive loss and in-plane emission loss at the cavity edges; (3) power loss due the threshold current (i.e., spontaneous emission); and (4) ohmic loss due to the series resistance of the device. The total power consumption, which is equivalent to the input power $P$, is found by evaluating the sum of these four parts,

6. SUMMARY

Thermal management for the CW operation of a large-area PCSEL has been discussed. We have carried out a thermal analysis of a PCSEL that is assembled in a water-cooling package via a sub-mount for heat dissipation. We have also performed experiments with an 800-µm-diameter PCSEL, wherein we measured the operating temperature of the PCSEL, and we have compared these measurements with the results of our analysis. We have found that the temperature of the MQW and photonic crystal layers can be kept under 64°C even at a high heating power of 40 W (or a power density of ${\sim}{{70}}\;{\rm{W}}\;{{\rm{mm}}^{- 2}}$) during CW operation. Owing to this effective thermal management, we were able to achieve ${\sim}{{8}}\;{\rm{W}}$ of CW output power from the single-chip PCSEL.