1. Introduction

GaN-based diode lasers are on track to become the major source of photonic power in the blue-green spectral region [1]. Therefore, it is desirable to gain comprehensive knowledge of all mechanisms of their operation and effects associated with their functionality. Of course, this involves all residual emissions in addition to the desired laser output. Such secondary emissions are potential sources of error on the one hand, but have the potential to be used for analytical purposes, on the other hand. They can occur as spontaneous edge emissions of materials that are part of the epitaxial architecture, such as barriers or waveguides [2]. Transitions via states generated by defects being located in the heterostructure [3,4], wave­guide [5], or the substrate [6] are further potential sources. In an earlier study, we investigated the microscopic origin of such transitions of GaN-based devices in the range of 900-1650 nm [7]. However, it is obvious to also consider the much more intense yellow emission of these devices, which can also be easily imaged with standard CCD-cameras.

The spontaneous emission of diode lasers in operation has been used for many years as a tool for the analysis of their properties. These include the temperature behavior, carrier heating, inter-valence-band absorption, the quality of optical waveguides, nonlinear gain, and the electronic confinement in quantum wells [8]. However, there is a problem especially above the laser threshold. Since the laser emission emerges from the spontaneous emission, they are naturally very close to each other, so that their spectral separation is difficult or impossible. Furthermore, the spontaneous emission, which is actually (almost) independent of the excitation level with respect to its lineshape, mutates into amplified spontaneous emission with serious effects on the lineshape, especially for low-dimensional gain media. Therefore, it would be highly desirable to find an alternative experimental access to the information contained in the spontaneous emission.

In this report, we show that the yellow emission from operating 450-nm emitting (In,Ga,Al)N-based diode lasers that is peaking at 580 nm contains a large part of the information contained in the spontaneous emission. The principal origin of this emission is also the epitaxial layer system. Spectral separation from primary emission, including spontaneous primary emission, is simple because there is no overlap. Thus, the yellow emission band has the potential to be used for analytical purposes, e.g. for the determination of refractive indices or visualization of non-equilibrium carrier profiles in devices, even without extra preparation, such as windows in the contact stripes.

2. Experimental

The 450-nm-emitting PL TB450B devices are in a TO56 package [9]. The device architecture is based on an (In,Ga,Al)N epitaxial layer sequence that consists of an asymmetric multiple quantum-well (MQW) and a GaN-based waveguide with Si- and Mg-doped n-and p-claddings, respectively [10]. The 15-µm wide emitter stripe is located off-centered on top of the n-side-down packaged chips. The cavity length is 1.2 mm. The threshold current is 200 mA. For further details see the PL TB450B data sheet [11]. A total of 16 devices were included in the study. Emission spectra were recorded with a Horiba optical grating-based spectrometer and a Bruker IFS66v Fourier-transform spectrometer. Emission images were taken with a C11440-22CU ORCA-Flash4.0 V2 digital CMOS camera with 82% peak quantum efficiency mounted to an OlympusBX51WI optical microscope. Images were taken in the 530-625 nm channels determined by precision long- and short-pass filters. In all experiments, the primary 450 nm emission is blocked by two long-pass filters (530 nm, optical density 4). Both, emission spectra and images were taken from the front side of the devices. All measurements were done at 25°C heat sink temperature, which was stabilized by water-cooled Peltier stages.

3. Results

3.1 Emission spectra and emission maps

Figure 1 shows emission spectra for different operation currents taken from the front facet of a device. Note that the laser threshold is at ∼200 mA. In addition to the yellow emission peaking at 580 nm, there is a weak shoulder starting at ∼825 nm. This is the onset of infrared emissions, which were subject of an earlier publication [7]. The lineshape of the emission is completely independent on the operation current, even when surpassing the laser threshold. The full width at half maximum (FWHM) amounts to ∼290 meV, but is cut by the long-pass filters. The signal-to-noise ratio is relatively high, 60-90. The photoluminescence (PL) spectrum of GaN was taken from the substrate of this particular device [12].

 

Fig. 1. Emission spectra taken from the front facet of a device at different operation currents. (a) Primary emission. (b) Yellow and infrared emissions. The vertical line marks the edge of the long-pass filter. The PL spectrum is taken from the GaN substrate; dashed curve.

Download Full Size | PDF

Figure 2 (a) shows the spatial distribution of the yellow emission. The scheme in (b) shows the front view of the device and the dashed frame corresponds exactly to the one in the image of the original data in (a). Subfigure (c) shows the normalized current dependences of the yellow emission for 3 different locations [red, green and blue in (a) and (e)], which differ in intensity by 3 orders of magnitude. After normalizing the profiles shown under (d) and (e), we found no deviation from each other, so that the shape of the profiles does not depend on the operation current.

 

Fig. 2. (a) Map of the yellow emission (b/w, original data) taken from the front facet. (b) Scheme illustrating the geometry of the device and the coordinate system that is used in the entire manuscript. (c) Normalized characteristics ‘yellow emission-versus-operation current’ taken at different locations; see symbols in (a) and (e). (d, e) Emission profiles extracted from maps.

Download Full Size | PDF

We now address the implications of these experimental results. Since the yellow spectrum, see Fig. 1(b), seems to consist of more than one contribution, we tried a separation by using different spectral channels within the emission band. Nevertheless, in all cases the active region, i.e. the depletion region of the pn-junction that contains quantum wells, barriers, and vicinal waveguides, remains as the one and only source of the emission, and all signals show the same normalized current dependence including the pronounced bend at the laser threshold; see Fig. 2(c). If it is really the case that the yellow emission band is made up of several contributions, these emission bands will probably be excited by the same mechanism. Obviously, this excitation is locked to the non-equilibrium carrier concentration in the active region; see the bend at the threshold at 200 mA in Fig. 2(c). Therefore, the source can only be genuine defect-related electroluminescence from the pn-junction (active region), or defect-related PL excited by spontaneous 450-nm emission generated in the active region. However, the latter seems less likely because of the following arguments:

 

Fig. 3. Comparison between modelled (red) and at 200 mA measured emission profiles. The parameters are given in the text.

Download Full Size | PDF

Thus, we conclude that the yellow emission is very likely a defect-related electro­luminescence from the active region of the device. The spectral shape of this emission, however, looks very similar to the PL from the substrate. This is not necessarily a contradiction, since the active region of these lasers also contains GaN; e.g. in the waveguide. According to Reshchikov [13,14], this yellow band has been the topic of ‘literally hundreds of publications. It is always broad, nearly Gaussian with a FWHM of about 350-450 meV.’ This is consistent with our observations. Reshchikov notes in section A of this review [13] ‘that the yellow band in all n-type GaN samples is related to transitions from the conduction band or from a shallow donor to a deep acceptor.’ He states that ‘the V_Ga-shallow donor complex is a very strong candidate.’ Additionally we have to take into account that doped, undoped, as well as In- and Al-alloyed GaN belong to the active region of our devices. Probably the contributions from several or all these active-region materials combined cause the observed deviation from a single Gaussian.

3.2 Modelling

In order to find out whether our preliminary conclusion that the yellow emission comes almost exclusively from the optical active region is correct, we have modelled the emission profiles shown in Figs. 2 by ray-tracing. For this purpose, we used a home-made software based on python^TM [15]. This happened mainly because the geometric problem is quite simple and the program is also used for the modelling of images taken by thermocameras [16]. Therefore, we checked the consistency of the results obtained with our own code with thermocamera images, which were modelled also by the commercially available ZEMAX software [17]. We found perfect agreement and consider this as the justification of our approach.

In our code, the geometry of the device that is sketched in Fig. 2(b) is approximated by two rectangular parallelepipeds. One stands for the GaN substrate (greyish), the other for the ∼2 µm thick epitaxial sequence with its different layers (red). The active region below the stripe contact, i.e. the 15 µm wide part of the epitaxial layer between the emitter stripe and substrate, is considered as the only source of all yellow emissions. The source is modelled as a volume in which individual light beams are generated at random positions with random propagation direction. Within the media, the rays are spreading straight. At surfaces between two media, according to Fresnel's formulas, refraction and reflection occur, whereby no preferred polarization is assumed, which is why the mean value of vertical and parallel polarization is calculated. Thus the active region, i.e. the epitaxial layer below the emitter stripe, is characterized by the following boundary conditions: The reflectivities at the top of the p-contact and at the rear facet are each 1. At the front facet there is the index jump to the air, while downwards the index step to the substrate (n_s) is effective. On both sides, the significantly weaker index step between the epitaxial layer (n_e) and the emitting active epitaxial layer (n_ae) is effective. At each surface, the path length within a medium is calculated and the absorption is calculated according to Lambert-Beer's law. The intensity of the light beam is reduced accordingly and the absorbed energy is stored with a reference to the volume element. 10⁷ beams were used for the calculations shown in Fig. 3.

Of course, Figs. 3(a) and (b) were modelled with the same parameters, which are (all for 580 nm): n_s=2.415; n_e=2.3644; n_ae=2.3654; absorption coefficient (α) of the substrate is 35 cm⁻¹; α of the emitting epitaxial layer is 10 cm⁻¹. The parameters used for the heavily doped substrate and the epitaxial layer system including the waveguide are in accordance with the literature [18,19].

Considering the simplicity of the model assumptions, the agreement of the ray-tracing modelling with the data in Fig. 3 is surprisingly good. This applies in particular to the y-axis behavior in Fig. 3 (b) in the substrate (∼5-100 µm), which answers the original question about contributions of PL from defects in the substrate pumped by spontaneous emission. We clearly see, that the yellow emission observed from the substrate is fully explained by light from the 15-µm wide emitting stripe.

In order to explain the x-axis behavior of the yellow emission, an index step of 0.001 at the edges of the 15-µm wide emitting stripe is required. Using dn/dT∼ 8.7⁻⁵ K⁻¹ from Ref. [20], we find a temperature rise of ∼11 K at 200 mA. Although this still sounds reasonable, the lack of any current dependence of this step in the entire range up to 400 mA indicates that this index step is most likely not due to a thermal effect, but to the built-in index guide determined by the ridge-waveguide structure; see Ref. [10]. This would also explain why the experimental data at the edges of the emitter stripe look sharper than the modelling prediction; see Fig. 3 (a) upper part.

4. Summary

In summary, we have analyzed secondary yellow emissions from operating 450-nm emitting (In,Ga,Al)N diode lasers. Both spatial and spectral behaviors were analyzed and modelled. The emission is very likely an electroluminescence from the part of the active area under the emitter stripe. With the exception of the spectral information, the yellow emission contains much information about the spontaneous emission, including its spatial intensity distribution. Both emissions are spectrally far apart. Therefore, the yellow emission has the potential to be useful for analytical purposes, e.g. for the determination of refractive indices or the visualization of non-equilibrium carrier profiles along devices, even without extra modifications at the devices, such as windows in the contacts.