1. Introduction

Laser annealing (LA) of semiconductor photonic and electronic devices [1] has decades of history and has attracted research and industrial interests due to its unique advantages, such as its selectivity [2] and its application to post device fabrication annealing [3]. Selective area LA could avoid unintentional modification to the photonic device active region such as in the case of dielectric capped intermixing processes [2,4–8] in which the whole device structure is subject to a high temperature rapid thermal annealing process with a risk of compromising the quality of the photonic structures [4–8]. However, the methods used previously for LA have never being able to provide in situ monitoring and control of the annealing process; in particular over the effectiveness and degree of atomic intermixing in semiconductors during LA. This often requires extensive experiments to be carried out to estimate the degree of intermixing one would wish to obtain [9]. Also, unavoidable issues with repeatability occur due to experimental variations in the optical alignment and setup parameters, affecting the consistency of the results. Real time monitoring of the intermixing in semiconductors is desirable at a targeted area of the device which requires precise control in the LA process for optimum bandgap modifications. In this letter, we propose and demonstrate a simple and effective apparatus with in situ monitor and control capability by utilizing a 1x2 fiber splitter. The apparatus and method are then demonstrated by forming a non-absorbing mirror on a quantum dot (QD) laser.

2. Apparatus and methods

The schematic diagram of the apparatus is presented in Fig. 1 . A 50:50 1x2 fiber splitter was used to link-up the device under test (DUT), the optical spectrum analyzer (OSA) and the high power laser (972 nm) annealing light source, as depicted in Fig. 1. For demonstration the DUT chosen was a 3 mm long QD ridge waveguide diode. The DUT was electrically probed on a sample stage and under cw operation and injected using a constant current (c.c.) source. The single end of the 1x2 fiber splitter was mounted on a xyz stage in front of the DUT, whereas the x2 ends of the fiber splitter were connected separately to the OSA and the 972 nm high power light source. This 1x2 fiber splitter setup allows easy and precise alignment of fiber at QD laser aperture for annealing. The fiber tip from the single fiber end was first aligned perpendicular to the optical axis of the DUT using the xyz stage by optimizing the optical output collected from the DUT, with helps from an optical microscope situated directly above the DUT and using the OSA as a power meter. A small gap of < 1 mm between the fiber tip and the DUT is allowed to prevent possible contamination or damage to the DUT facet. Using this setup design, changes in peak optical power and emission spectrum of the DUT electroluminescence during annealing are monitored in real time on OSA. The OSA was used to display and monitor the emission spectra in real time under repeat scan mode to continuously scan within the range of interest. This technique provides real time control and a better insight into the LA process, since it has the dual purpose of delivering high energy photons to anneal the DUT and simultaneously measures its emission spectrum.

 

Fig. 1 (Colour online) Schematic diagram of the experimental setup used for laser annealing with in situ control and monitor capacities.

Download Full Size | PDF

3. Applications and results

In our demonstration, during LA, the power of the 972 nm light source during LA was set to 150 mW and the DUT was operated under cw at a constant injection current of 200 mA. Owing to the use of the 1x2 fiber splitter, signals from the 972 nm and DUT can be monitored simultaneously on the OSA in real time during LA. A weak second harmonic signal from the light-matter interaction between 972 nm light and the InAs/GaAs semiconductor is also observed at 486 nm on the OSA. A cw laser light source was used for the annealing since it is known that high energy laser pulses can cause excessive damage to the sample [10]. The damage mechanism is thought to be the formation of point defects due to bond breaking. Such lattice disruption created by laser pulses need a subsequent RTA step to induce interdiffusion and to recover the quality of the material [11].

In our experiment, the LA process was carried out on post-fabricated QD laser diodes using a 972 nm cw laser and is not observed to cause any degradation to the laser active layer [12]. The wavelength selection of the annealing light source (in this case, 972 nm) needs to be shorter than that of the DUT (in this case, 1300 nm QD laser) for efficient absorption in the DUT [3]. Figure 2 depicts the emission spectra from the 972 nm annealing light source, the QD laser and the second harmonic signal at 486 nm which arises from the light-matter interaction between 972 nm light source and InAs/GaAs semiconductor.

 

Fig. 2 Optical spectra from the 972 nm light source, QD laser and the second harmonic signal captured on OSA.

Download Full Size | PDF

During LA, a progressive blueshift in emission wavelength of the electroluminescence (EL) from the DUT was clearly observed on the OSA whilst at the same time high energy photons are being delivered to anneal the device facet. This enables in situ control on the degree of intermixing. An example of the spectra recorded during annealing is presented in Fig. 3 .

 

Fig. 3 (Colour online) Progressive blueshift of the electroluminescence spectrum of the InAs/GaAs QD laser during annealing. From right to left: 0, 1, 3, 7, and 10 min after the 972 nm annealing light source was turned on. The blueshift is temporary and will revert back to the original position after the LA process has ended.

Download Full Size | PDF

The wavelength shift, Δλ, of the QD ridge waveguide diode during LA could correspond to the net result of a combination of several possible effects: (i) Wavelength shortening as a result of a change in the refractive index resulting from a change in carrier density during direct modulation. However as the QD laser is operated in cw mode, the change in carrier density can be neglected in our case. (ii) Wavelength lengthening due to Joule heating, Δλ_Joule, where heating of the lattice can result in a red-shift in the emission wavelength. (iii) Wavelength shortening due to a reduction in the cavity length, L, which is now minus the portion which has undergone intermixing, ΔL. For this case the shift should follow the relationship Δλ_length = 2 • ΔL/n, where n is an integer and ΔL is negative in this case. This shift originates because the section under annealing is undergone a rearrangement of atoms and hence the “effective” mirror of the device during laser annealing is at a shorter position. As the intermixing depth extends deeper into the laser cavity over time, L becomes shorter and hence a progressive blueshift in emission wavelength with time is observed. From these arguments, we believe that the spectral blueshift observed in Fig. 3 is a net result of Δλ_length + Δλ_Joule.

The blueshift in EL spectrum shown in Fig. 3, recorded in real time during LA, is temporary and can be recovered back to the original position in about a couple of minutes after the LA process has ended. This is because atoms in the section which has undergone intermixing have stabilized and have returned back to crystalline lattice. Hence the device cavity length is back to the original length. A gradual decrease in the output intensity during LA is observed, most likely due to imperfections in the non-absorbing mirror during its formation phase in which the rearrangement of the atoms is still dynamic. The QD laser emission wavelength remains the same before and after LA [12], because the cavity length of the device remains the same, even though the length of the gain region has been reduced after the formation of non-absorbing mirrors.

The 972 nm light is estimated to be fully absorbed within a 200 μm depth from the edge of the QD ridge waveguide device. Hence the centre active region is undisturbed. More extensive blueshift is possible by increasing the power of the 972 nm light source, which will provide additional degree of freedom in controlling the annealing duration required. However, too high power density is known to cause physical damage to the semiconductor facets, as it has been demonstrated in ref [10]. using high power density laser pulses. In our case of cw LA, the power density used is relatively low and there was no observable damage to the facet when the annealing duration was increased. The extent of blueshift in bandgap at the annealed facet depends on the type of DUT. As reported in literature, a more extensive bandgap blueshift is expected for a QD structure [2] as compared with quantum well (QW) structure [13].

Using the setup described in Fig. 1, the facets of a InAs/GaAs QD laser diode was annealed using the 972 nm light source with its power set at 150 mW. Figure 4 shows the light current characteristics of the QD laser tested on a thermoelectric cooler at 22 °C. There is a strong improvement in the light output when the facets have been annealed for 5 min cumulative periods. This suggests that our LA setup is very effective for intermixing. The effect is to widen the energy bandgap at the facets and hence reduce the carrier recombination velocity and optical absorption from the facets. The improved light output power and slot efficiency shown in Fig. 4 are due to the formation of non-absorbing mirrors on the semiconductor laser facets. We have previously shown that this has an important role in high power devices, greatly suppressing thermal runaway effects at the facet [12].

 

Fig. 4 (Colour online) Light-current characteristics measured from the same InAs/GaAs QD laser before (solid line) and after facet 1 annealed for 5 min (broken line); facets 1 and 2 annealed for 5 min each (dot-dashed line); and facet 1 annealed for 10 min, facet 2 annealed for 5 min (dot-dot-dashed line).

Download Full Size | PDF

The in situ control and monitor capabilities of the setup enable precise control of annealing conditions and can prevent any possible damages to the device during annealing, which would be difficult to guarantee without the monitoring function. The proposed instrumentation is also ideal for post fabrication broadening of the emission spectra in light emitting devices such as superluminescent diodes [14,15] and broadband lasers [16]. By proper selection of the annealing light source, the technique is applicable to both QD and QW based devices over a wide range of operating wavelengths.

4. Conclusion

In summary, the apparatus and method for in situ monitoring and control of photonic device intermixing are described in detail. The setup enables photons to be delivered to a targeted area of the photonic device and simultaneously measures the emission spectra from the device to monitor the intermixing process in real time. The in situ monitoring of a laser annealing process to modify a semiconductor laser diode facet is demonstrated using the instrumentation. The monitor capabilities of the setup enable precise control of the annealing conditions which can prevent possible damage to the photonic device during annealing, which otherwise cannot be guaranteed without the monitoring function.