Technical Insight
Waveguide integration boosts laser diode yield
With many optoelectronics manufacturers set up during the telecom boom now looking to target industrial and military applications of lasers, John Marsh describes a technique that improves the manufacturing yield of high-power diode laser arrays.
The advantages of monolithic integration of semiconductor optoelectronic components include high yields, high performance and new functionality. Over the past 25 years, the main research driver for such integration has come from the telecom sector, but significant deployment is now taking place in other market sectors. These sectors range from high-volume industrial markets to lower-volume avionics and defense markets, with specific applications in printing systems, microwave photonic systems and optical signal processing. There is considerable cross-over from telecom requirements into these other market sectors.
In the industrial market, there are strong commercial drivers for integrating high-power single-mode lasers into multi-element laser arrays. In many industrial systems, light from individually packaged lasers is coupled (usually by fiber) into an optical head (figure 1). In the assembly process, semiconductor lasers are cleaved and packaged individually, and then precision alignment of optical fibers is required in both the package and the head. However, if the lasers are integrated into a linear array, the output of the array can be used directly and the module effectively becomes the optical head. In this approach only one package is required, thereby reducing form-factor and cost and eliminating fiber pigtailing. Furthermore, the pitch of the emitters is fixed by lithography, rather than by mechanical fiber placement.
Manufacturing yieldsFor laser arrays to be viable, the manufacturing process must have a high yield, and every element must perform within a tight electrical and optical tolerance. Although fiber coupling has been eliminated, the specification on beam quality may be at least as demanding as that required for coupling to single-mode fiber. Meeting these requirements is a major challenge, as the application might require several hundred milliwatts of power in a single transverse mode from each element. The overall performance is similar to that of a state-of-the-art single 980 nm erbium-doped fiber amplifier pump laser of only a few years ago, and because the wavelength range of particular interest is 650-1000 nm, the array faces the same reliability and yield issues as 980 nm pump lasers. However, the challenge is magnified by the fact that there may be up to 100 elements in the array. In addition, the arrays demand more real estate than single devices, reinforcing the criticality of the yield figure.
Laser arrays are a good example of parallel optoelectronic integration, in that many identical elements are placed side by side. However, serial integration at the chip level is critical in meeting the performance and cost requirements. The inclusion of passive waveguides adjacent to the facets of the laser cavities leads to good electro-optical performance, and also relaxes mechanical tolerances. These passive waveguides improve the yield in three different ways: the catastrophic optical damage (COD) threshold is raised, while the cleaving tolerance is relaxed, as is the packaging alignment tolerance.
Electro-optic performanceCOD is the main failure mode of high-power semiconductor lasers operating in the 650-1000 nm wavelength range. Incorporating passive waveguides in the facet regions to form "non-absorbing mirrors" (NAMs) or "window lasers" is an established technique for raising the COD threshold. The use of NAMs leads to a direct increase in the COD threshold. The transparent passive waveguides in the facet region reduce the absorption of light at the facet, preventing the runaway process responsible for COD. Quantum well intermixing (QWI) is the most common technique for fabricating the passive waveguides (Compound Semiconductor September 2001). In one QWI approach, point defects are created during sputter deposition of SiO2. For simple uncoated devices, NAM lasers have a greater COD level than a standard laser by a factor of 2.6.
The passive waveguides also relax the cleaving and packaging tolerances. The facets of semiconductor lasers are formed by breaking the semiconductor crystal along a cleavage plane. This gives an extremely flat mirror, but the typical precision to which cleaving can be carried out is around ±5 µm. For conventional lasers, the cleave must be located less than 10 µm from the end of the gain contact, but in a laser with low-loss passive waveguides the cleaving position only needs to fall somewhere within the passive region. Angular variations in the cleaving direction have an even larger impact, especially for arrays. Using normal photolithography, the tolerance with which the laser waveguides can be aligned to the crystallographic planes is around ±0.05°. Over a 1 cm facet length this introduces a further ±9 µm of "run-out" in the cleave position. Finally, the passive waveguides relax the packaging alignment tolerances. Because little heat is generated in the passive section, the laser chip can overhang the heat sink.
Array performanceIn The performance and manufacturing yield of laser arrays is therefore increased substantially by integrating passive waveguides within the laser cavity; indeed, such integration is a prerequisite for achieving good yields. Laser arrays with extremely uniform characteristics have been demonstrated. Figure 2 shows the kink power of lasers in an array, in which 86 consecutive elements have a kink power of more than 220 mW - again comparable to the power of leading-edge 980 nm pump lasers. Kink power is a measure of how much laser power is emitted in the lowest-order transverse mode. It is important, because second-order output modes have a detrimental effect on the laser s far-field pattern and, consequently, the effectiveness of the tool for industrial applications. A correspondingly high level of uniformity is seen in parameters such as threshold current and slope efficiency.
Further readingMarsh et al. Improved COD level from laser with non-absorbing mirrors IEEE Photonics Technology Letters 14 (10) 1394.
In the industrial market, there are strong commercial drivers for integrating high-power single-mode lasers into multi-element laser arrays. In many industrial systems, light from individually packaged lasers is coupled (usually by fiber) into an optical head (figure 1). In the assembly process, semiconductor lasers are cleaved and packaged individually, and then precision alignment of optical fibers is required in both the package and the head. However, if the lasers are integrated into a linear array, the output of the array can be used directly and the module effectively becomes the optical head. In this approach only one package is required, thereby reducing form-factor and cost and eliminating fiber pigtailing. Furthermore, the pitch of the emitters is fixed by lithography, rather than by mechanical fiber placement.
Manufacturing yieldsFor laser arrays to be viable, the manufacturing process must have a high yield, and every element must perform within a tight electrical and optical tolerance. Although fiber coupling has been eliminated, the specification on beam quality may be at least as demanding as that required for coupling to single-mode fiber. Meeting these requirements is a major challenge, as the application might require several hundred milliwatts of power in a single transverse mode from each element. The overall performance is similar to that of a state-of-the-art single 980 nm erbium-doped fiber amplifier pump laser of only a few years ago, and because the wavelength range of particular interest is 650-1000 nm, the array faces the same reliability and yield issues as 980 nm pump lasers. However, the challenge is magnified by the fact that there may be up to 100 elements in the array. In addition, the arrays demand more real estate than single devices, reinforcing the criticality of the yield figure.
Laser arrays are a good example of parallel optoelectronic integration, in that many identical elements are placed side by side. However, serial integration at the chip level is critical in meeting the performance and cost requirements. The inclusion of passive waveguides adjacent to the facets of the laser cavities leads to good electro-optical performance, and also relaxes mechanical tolerances. These passive waveguides improve the yield in three different ways: the catastrophic optical damage (COD) threshold is raised, while the cleaving tolerance is relaxed, as is the packaging alignment tolerance.
Electro-optic performanceCOD is the main failure mode of high-power semiconductor lasers operating in the 650-1000 nm wavelength range. Incorporating passive waveguides in the facet regions to form "non-absorbing mirrors" (NAMs) or "window lasers" is an established technique for raising the COD threshold. The use of NAMs leads to a direct increase in the COD threshold. The transparent passive waveguides in the facet region reduce the absorption of light at the facet, preventing the runaway process responsible for COD. Quantum well intermixing (QWI) is the most common technique for fabricating the passive waveguides (Compound Semiconductor September 2001). In one QWI approach, point defects are created during sputter deposition of SiO2. For simple uncoated devices, NAM lasers have a greater COD level than a standard laser by a factor of 2.6.
The passive waveguides also relax the cleaving and packaging tolerances. The facets of semiconductor lasers are formed by breaking the semiconductor crystal along a cleavage plane. This gives an extremely flat mirror, but the typical precision to which cleaving can be carried out is around ±5 µm. For conventional lasers, the cleave must be located less than 10 µm from the end of the gain contact, but in a laser with low-loss passive waveguides the cleaving position only needs to fall somewhere within the passive region. Angular variations in the cleaving direction have an even larger impact, especially for arrays. Using normal photolithography, the tolerance with which the laser waveguides can be aligned to the crystallographic planes is around ±0.05°. Over a 1 cm facet length this introduces a further ±9 µm of "run-out" in the cleave position. Finally, the passive waveguides relax the packaging alignment tolerances. Because little heat is generated in the passive section, the laser chip can overhang the heat sink.
Array performanceIn The performance and manufacturing yield of laser arrays is therefore increased substantially by integrating passive waveguides within the laser cavity; indeed, such integration is a prerequisite for achieving good yields. Laser arrays with extremely uniform characteristics have been demonstrated. Figure 2 shows the kink power of lasers in an array, in which 86 consecutive elements have a kink power of more than 220 mW - again comparable to the power of leading-edge 980 nm pump lasers. Kink power is a measure of how much laser power is emitted in the lowest-order transverse mode. It is important, because second-order output modes have a detrimental effect on the laser s far-field pattern and, consequently, the effectiveness of the tool for industrial applications. A correspondingly high level of uniformity is seen in parameters such as threshold current and slope efficiency.
Further readingMarsh et al. Improved COD level from laser with non-absorbing mirrors IEEE Photonics Technology Letters 14 (10) 1394.
