Tunnel injection structures are interesting for applications in diode lasers since they offer additional degrees of freedom for the design of the gain region. This is given by the spatial separation of carrier source (injector) and light emitter. In case of quantum well (QW) and quantum dot (QD) devices, a selective population of the lasing ground state allows for reducing the electron heating within the emitter by direct injection of cold electrons from the separated carrier source into the ground state by tunneling. Potentially this results in reduced gain suppression, increased quantum efficiency, decreased diffusion capacitance, and lower low-frequency roll-off and high-frequency chirp. Furthermore, the expected preferential population of QDs of a particular size eventually leads to reduced inhomogeneous broadening of the gain (emission) spectrum.

Fig. 1

InGaAs tunnel injection structures are grown by solid-source molecular beam epitaxy with the following sequence of layers: InAs QDs (2 monolayers), GaAs spacer), and a single 11 nm thick In_0.15Ga_0.85As QW; all embedded into a GaAs matrix. This layer sequence is rather unusual and is expected to promote the formation of nano-bridges. Reference samples containing either exclusively QDs or QWs are grown as well. Structural information is complemented by transmission electro microscopy (TEM) analysis in the diffraction contrast mode; see Figs. 1 (a-c). The In_0.6Ga_0.4As QDs have a diameter of 18 nm, a height of 4 nm and an array density of ~5x10¹⁰ cm^-2. The standard deviation of these parameters is found to be on the order of 10 percent. Using this information, calculations within the framework of the effective mass approximation are performed, providing a band structure scheme as depicted as Fig. 1 (d). Fig. 1 (e) presents PL and PLE spectra from a sample with 6.5 nm barrier thickness. Tentative assignments of the optical transitions are given, with the subscripts 0 and 1 standing for ground and excited state transitions. These attributions are confirmed by both the calculations and the evolution of the PL intensities when increasing the population density by switching from cw to impulsive excitation.

Fig. 2

Transient PL measurements are made with a setup that is shown in Fig. 2. They are performed at 10 K with sub-100 fs excitation pulses using a mode-locked Ti:Sapphire laser operating at 82 MHz. The photon energies and excitation power density are 1.33-1.60 eV and ~5x10¹¹ photons per pulse and per cm², respectively. The PL is dispersed in an imaging monochromator and detected by a synchro-scan streak camera equipped with an infrared-enhanced cathode. The time resolution of the total system is at best ~10 ps.

Fig. 3

PL maps as detected by the streak camera system are shown in Figs. 3 (a,b). The only difference between both measurements is an increase of the excitation photon energy by 90 meV from 1.33 to 1.42 eV. For an excitation energy below the QW absorption edge, the QDs get predominantly populated by their own ground and first excited state absorption, resulting mainly in 'hot' luminescence via the QD1 state with a decay time of about 150 ps; see Fig. 3 (a). For an only slightly increased excitation photon energy that exceeds the QW0 ground-state transition energy (and thus allows for efficient QW absorption), we observe a substantial increase of the QD0 luminescence (decay time ~ 750 ps); see Fig. 3 (b). The increased QD0 PL is likely to be caused by the efficient population of the electronic QD ground states by tunnel injection from the QWs. Thus we show that in this particular structure pumping at increased photon energies results in 'colder' luminescence. The term 'colder' points here to the fact, that the ratio of the populations of ground and excited states is altered in favor of the ground state population. Basically, the result shown in Fig. 3 (a, b) is a clear demonstration of the functionality of a tunnel injection structure.

The catastrophic optical mirror damage (COMD) effect is analyzed for 808 nm emitting diode lasers in single-pulse operation. This approach allows for better separating facet degradation from subsequent degradation processes in the bulk. The setup is depicted in Fig. 4.

Fig. 4

During each single pulse, both nearfield of the laser emission and thermal image of the laser facet are monitored with cameras being sensitive in the respective spectral regions. A temporal resolution of better than 7 µs is achieved.

Fig. 5

Figure 5 shows results obtained for 2-μs-long 35 A current pulses for 2 devices. Each left and right columns correspond to data from one device. Fig. 5 (a,b) show the thermal images, whereas (c,d) correspond to lateral cuts through the three lines with maximal signal, with the homogeneous thermal background subtracted. Fig. 5 (e,f) give the normalized lateral nearfield profiles. Profiles for the first current pulse (thick black lines) are scaled by 0.05 relative to the succeeding pulses (thin colored lines). Profiles at 2 A, scaled by 0.05 as well, are given as dotted lines for comparison. The positions of the peaks in the thermal data and deep minima in the nearfield data agree (marked by dotted vertical lines and capital letters). The active stripe is located between 100 and 300 μm. The COMD is unambiguously related to the occurrence of a 'thermal flash' detected by thermal imaging.

Fig. 6

Figure 6 shows secondary electron micrographs acquired from the front facet (p-side up) of a device. The epitaxial structure is indicated. Extrusions are found at exactly those positions, which are marked in Fig. 5 by capital letters. Partial darkening of the n-type waveguide at these positions is visible as well. The inset shows the amplitude of the 'thermal flash' versus the area of the COMD damage sites for both devices. Thus the COMD is unambiguously related to the occurrence of a 'thermal flash' detected by thermal imaging. A one-by-one correlation between emission nearfield, 'thermal flash', thermal runaway, and structural damage is observed. As a consequence of the single-pulse-excitation technique, the propagation of 'dark bands' (as observed in photo- or cathodoluminescence maps in the plane of the active region) from the front facet is halted after the first pulse. Because of the rapidness of the thermal runaway, we propose the described technique for the test of the facet stability and the intentional preparation of early stages of COMD; even for diode lasers that regularly fail by mechanisms other than COMD.