InAs/InP Quantum Dash devices at 1.55µm
InAs/InP Quantum Dash devices.at 1.55µm
Quantum dot (QD) materials have been long hoped to offer real and significant device performance improvement on the previous generation of strained and unstrained quantum Well materials. These properties are in particular the prospect of a lower and temperature insensitive threshold current compared with quantum well lasers. Previously successful 1.3µm wavelength GaAs-based QD materials are difficult to extend to the longer 1.55µm wavelength. Given a wealth of existing commercial experience, InAs quantum dots grown on InP substrates offer a cost effective alternative. Growth on these commercially favoured (001) oriented InP substrates favours elongated quantum dot structures called quantum dashes. These InAs/InP quantum dash lasers have demonstrated many impressive characteristics such as high modal gain per dash layer and high characteristic temperature, T0. As part of the EU FP6 funded ZODIAC ) project we have theoretically and experimentally investigated the polarisation sensitivity and gain and loss mechanisms of these devices, providing important information to the growth partner for the next generation of device and generating a real device improvement (see loss mechanisms DW1 and DW2).
Fig. 1. TEM Cross sectional view of quantum dash layers (courtesy of nanoplus).
Band-structure and Density of States
- Poorly Confined Electrons and deeply confined Holes
Two types of active region were grown at Thales IIIV Lab for this project. They included layers of elongated dots enclosed by pure 1.17µm barrier material (Dash -in-a-barrier or DBAR) and a second growth scheme where each layer of dots was surrounded by a 1.45 µm InGaAsP quantum well. We performed 8 band k.p calculations for both schemes to determine the electronic structure and it was shown that the electron states are poorly confined with a large tunnelling probability into the bulk. For the DWELL materials it was found that the electron states lay in the QW layer.
Fig. 2 Schematic Diagrams indicating band offsets (eV) and main transitions for DWELL (left) and DBAR(right). Also shown to the right of each figure is the full DOS between 1 eV and -0.04 eV (indicated by arrow) for each active region type. The wire like DOS due to the long (1-10) direction and the HH excited states in the (110) direction are clearly visible. The full scale of LH and QW like states DOS for the DWELL is not shown due to space constraints.
Examining the DOS of both materials it was clear that in contrast to QW devices the density of states of both conduction band states and valence band were of the same order. The implications of this for the grower is that it is unlikely that the usual p-doping (conventionally used to redress the balance between conduction band levels and dense VB levels) are unlikely to improve the device performance. In fact our experimental work has also shown that to the contrary the losses introduced by p doping greatly outweigh the benefits.
Anisotropic Polarisation Properties- Theory and Experiment
Theory. Calculating momentum matrix elements and absorption for each type of active region has shown that the DWELL and DBAR samples both exhibit not only an anisotropy between TE/TM states, but also an anisotropy between TE(110) and TE(1-10) states. The DWELL devices have a calculated ratio, r=|M(1-10)|2/|M(110)|2 of approximately 1.4 for the lowest area density of dashes while the DBAR device has a sizably higher ratio of 1.6 for the same density. This ratio, r, reduces with an increase in the number of dashes per unit area and increases with the cross sectional height of the dash. Anisotropy can only be related to an anisotropic final potential energy surface in the two in-plane directions.
We can calculate the absorption in the different polarisation directions according to the following equation and compare to experiment.
Fig. 3. Calculated Absorption curves (top) multiplied by filling factor ζ of DWELL (left) and DBAR (right). Also shown (bottom) are the corresponding anisotropy ratios, r=|M(1-10)|2/|M(110)|2 for the DWELL (left) and DBAR (right).
Experiment. We measure the spontaneous emission (SE) through a window in the top contact, using a polarizer to discriminate between TE(110) and TE(1-10) emission. There is a clear enhancement of TE(1-10) over TE(110) emission in both DW1 and DB1. Fig. 4 shows the ratio, r, of TE(1-10) to TE(110) polarised SE for DW1 (left) and DB1 (right) as a function of bias current. It can be seen that the ratio does not depend on current, indicating that it is an intrinsic property of the material.
The value of r~1.5 in DW1 for 0.75<0.85 eV. As the energy increases, the ratio decreases to unity, and then drops just below unity for hν>0.95 eV. In the region 0.75<0.85 eV,the SE emission is due entirely to transitions involving dash valence states, and r is therefore a direct measure of the ratio of the squared matrix elements |Mn|2 used in  to describe TE(1-10) and TE(110) transitions. For hν > 0.95 eV, the SE spectra also include contributions from electron-hole pairs confined in the well and, so r → 1, because matrix elements associated with quantum well holes tend to have equal magnitude for recombination polarised along the (110) and (1-10) directions. DB1 shows similar characteristics, with r∼ 1.8 for transitions originating in the dashes near to the band edges. Comparing the experimental plots of r in Fig. 4 to the calculated plots in Fig. 3 we find good agreement.
Fig. 4. SE spectra for DW1 (left panel) and DB1 (right panel) for drive currents 3.3 mA (black dotted), 5 mA (red dashed), 7 mA (blue solid).
Understanding and Reducing the Device Loss Mechanisms
The initial samples which were grown at Thales IIIV lab and analysed here at Tyndall we have denoted DW1 and DB1. Here at Tyndall we performed experimental characterisation of the loss processes and found significant evidence for substantial Auger recombination processes as in other 1.55µm lasers. We also found that the first DWELL sample, DW1(left panel) exhibited a strong leakage path (see zth>3) which coincided with a strong reduction in differential external efficiency at temperatures greater than 250K. Wavelength analysis of the spontaneous emission indeed showed a significant presence at the barrier wavelength of 1.17µm and which did not clamp at threshold, consistent with a leakage path.. This leakage path was subsequently greatly reduced in the second generation DWELL DW2 shown in the right hand panel of Fig.5
Fig. 7. Top panels: Plot of total threshold current, Ithtotal ,(filled red squares) and its radiative component at threshold, Ithrad ,(filled blue triangles) as a function of normalised to its maximum temperature for DW1(left panel), DB1(middle panel) and DW2(right panel). Bottom panels: Plot of zth (open green squares) and ηextd