Issue



Fabricating buried tunnel junctions for InP-based VCSELs


02/01/2006







Markus Ortsiefer, Vertilas GmbH, Garching, Germany

Long-wavelength (λ≥1.3µm) vertical cavity surface emitting lasers (VCSEL) are highly attractive light sources for applications in optical communication and sensing. Unlike their short-wavelength GaAs-based counterparts, the realization of devices on InP for wavelengths beyond 1.3µm has been hampered by a number of fundamental technological drawbacks, particularly higher temperature sensitivity of active material gain, lower index contrast, and unsatisfying thermal conductivity of ternary or quaternary semiconductor compounds used as mirror layers.

The buried tunnel junction (BTJ) VCSEL concept, which was first demonstrated in 1999 [1], represents a breakthrough for such devices. InP-based BTJ VCSELs have demonstrated excellent device performance with respect to operating temperature (>100°C), output power (1mW single mode @ 80°C), high-speed capability (10Gbit/sec), and wavelength versatility (1.3-2µm) [2].

Design

The BTJ concept circumvents the problems of poor thermal and optical properties of InP-based semiconductor materials by significantly reducing excess device heating and improving heatsinking. The technology enables a significantly reduced series resistance and consequently smaller Joule heating (see figure), which is accomplished by a low resistive tunnel junction that substitutes low resistive n-doped material for high resistive p-doped layers. The specific sheet resistance of the tunnel junction itself is as small as 3×10-6Ωcm2, which is comparable to optimized p contacts.


Device structure of a BTJ-VCSEL.
Click here to enlarge image

With the tunnel junction being restricted to a well-defined diameter, the region outside the BTJ resembles an effective blocking junction. Hence, the current flow is effectively restricted to the ohmic region. Since at least one of the quarter wave layers in an InP-based distributed Bragg reflector (DBR) consists of ternary or quaternary materials, the thermal resistance of such a stack that used a back mirror would deteriorate the laser performance. Instead, a reflectivity beyond 99% is achieved with a short-period dielectric layer stack.

Fabrication

The epitaxial growth process of BTJ VCSELs is accomplished using solid-source molecular beam epitaxy (MBE) and comprises two runs. For BTJ VCSELs, the material system InGaAlAs turns out to be most effective. Layers of InGaAlAs exhibit several advantages as compared to the almost equivalent InGaAsP system that principally covers the same wavelength range. Having only one group-V element (As), the quaternary InGaAlAs grown with solid source MBE enables sharp interfaces and gives the possibility of implementing very high strain in the active region.

Even for active regions emitting with a wavelength of ~2µm, high crystal quality can be combined with a substantial number of quantum wells. Furthermore, the higher conduction band offset in the InGaAlAs system allows improved electron confinement and consequently better temperature stability of active material gain, which is particularly important for high current densities in VCSELs.

In the first epitaxial run, an all-arsenic base structure, including the epitaxial mirror, the active region, and a final tunnel junction, is grown. The junction comprises two heavily doped layers with silicon and carbon concentrations of ~1×1020/cm3, respectively. Due to a small gain length, lasing activity in VCSELs presumes very high mirror reflectivities beyond 99%. To achieve this reflectivity, a typical number of quarter wavelength layer pairs in the DBR is ~30-40 depending on the wavelength and, subsequently, the layer composition and refractive index difference. At 1.55µm wavelength, the refractive index is ~0.28 and is significantly smaller than the refractive index differences known from GaAs VCSELs (~0.5).

After the first growth step, the tunnel junction is laterally structured by means of standard photolithography and chemical dry etching. It is then turned back to MBE and regrown with thermally advantageous InP using a valved phosphorous cracker. While gas phase techniques such as metal-organic chemical vapor deposition (MOCVD) are most often used for regrowth processes, MBE is also found to provide high-quality regrowth layers. The total thickness of all epitaxial layers is ~10µm. Hence, accurate control of the heteroepitaxial material composition and layer thicknesses is essential, primarily because the Bragg wavelength and the emission peak depend on it. Precise layer composition of the thick DBR mirror is crucial to prevent lattice relaxation. When targeting wavelengths for sensing applications, it is mandatory to provide a thickness accuracy well below 1%; for this purpose, standard pyrometry is used.

The remaining processing is done with standard planar technology, including wet and dry etching, metallization, plating, and coating. The bottom mirror is realized by electron beam evaporation of dielectric materials. Although these materials usually exhibit insulating characteristics and cannot be used as current supplying layers, they are highly attractive for their versatile refractive indices. For example, a dielectric layer stack of only 3.5 pairs of CaF2/ZnS (∆n ≈ 0.9) with additional Au coating results in a reflectivity of 99.9% at a wavelength of 1.55µm.

A distinctive feature of BTJ VCSELs is the complete removal of the InP substrate by selective wet etching. To ensure mechanical stability, an electroplated Au layer is formed on the bottom side, which simultaneously serves as an excellent heat sink. The wafer’s ~60µm thickness is comparable to that of a human hair. The structure enables all of the relevant fabrication steps such as dicing, mounting, packaging, etc. With respect to the anticipated high-volume markets for VCSEL applications, cost-effective production is mandatory. The BTJ concept offers the advantage of full-wafer manufacturability even for larger diameters than the commonly used 2-inch size.

Acknowledgments

The author appreciates the collaboration with M.-C. Amann and G. Böhm from the Walter Schottky Institute of the Technical U. of Munich.

References

  1. M. Ortsiefer, M. Lohner, R. Shau, G. Böhm, M.-C. Amann, “Low-resistance InGaAs Tunnel junctions on InP for Long-wavelength VCSELs,” Semiconductor and Integrated Optoelectronics (SIOE), Cardiff, 1999.
  2. R. Shau, M. Ortsiefer, J. Rosskopf, G. Böhm, C. Lauer, M. Maute, M.-C. Amann, “Long-wavelength InP-based VCSELs with Buried Tunnel Junction: Properties and Applications,” Vertical-Cavity Surface-Emitting Lasers VIII, Proceedings of SPIE, 5364, 1-15, 2004.

Markus Ortsiefer is head of R&D and co-founder of VERTILAS GmbH, Lichtenbergstrasse 8, c/o Gate Garching, D-85748, Garching, Germany; ph +49/89-5484-2007, e-mail [email protected].