Optical and magneto-optical properties of erbium doped InGaN and GaN epilayers

Using combined excitation–emission spectroscopy we have studied the erbium incorporation into GaN and InGaN for in situ doped MOCVD-grown layers and compared them to samples grown by MBE. A smaller up-conversion efficiency for the main site is observed compared to minority sites in the same sample as well as versus all sites from MBE grown samples. Furthermore, we show that the 1.54 lm emission efficiency is reduced with increasing In-content both under excitation above the bandgap in the UV as well as under resonant excitation at around 980 nm. This indicates that non-radiative decay channels for the Er ion are the largest contributing factor for this behavior. Finally, the Zeeman splitting of the excitation and emission lines of Er:GaN under application of magnetic fields up to 6.6 T with B||c-axis is shown. 2010 Elsevier B.V. All rights reserved. Following the commercial availability of high power III-nitride light emitting diodes in the 370–540 nm range, a strong desire for efficient Er-doped InGaN epilayers has developed. Recently Er-doped InGaN epilayers have been grown by MOCVD with good crystal quality. Nevertheless they show a relatively weak 1.54 lm emission efficiency when compared to Er-doped GaN [1]. Aside from application in novel light sources, there has been an increasing interest for some time in the magnetic properties of the dilute Er-dopants, since ferromagnetism of Er:GaN has been reported [2]. To contribute to the clarification of the underlying coupling mechanism, we have investigated the effects of applied magnetic fields on Er-doped GaN epilayers as grown by MOCVD in order to determine the effective g-values of the ions excited and ground state. In this paper, we want to particularly address the following questions: (1) are the multitude of excitation and emission lines related to a single site as shown in Ref. [3] or do additional incorporation sites appear as found for MBE grown samples in Ref. [4]. (2) What is the cause of the strong reduction in emission efficiency that has been observed for increased In-content under UV excitation [1]. Is this caused by a less efficient excitation channel or by a more pronounced non-radiative decay channel from excited Er ions. (3) What are the Zeeman splittings of the ground and excited levels and how do they behave for magnetic fields for which the Zeeman splitting is comparable to the crystal field splittings. Samples used in these studies were MOCVD-grown Er:GaN epilayers that were grown on sapphire and in situ doped with Er up to 10 cm . Details of these samples are described in Ref. [1]. ll rights reserved. : +1 610 758 5730. d). As a first step, we show in Fig. 1 the combined excitation– emission spectroscopy (CEES) data (for details of the technique see e.g. Ref. [3]) for the technologically important erbium transition at around 1.54 lm. We excite this transition using a tunable semiconductor laser at around 980 nm. At first sight a large number of peaks appear in the image plot even in the rather small energetic window that is depicted. This may suggest a large number of sites. However, thermally activated levels, electron–phonon coupling, and energy transfer will increase the number of transitions excepted in the ideal case of a single site at zero temperature without coupling to lattice vibrations. In the assignments, we follow the following guidelines: The emission spectra of a single site must be identical in spectral position and relative heights for all its excitation transitions and accordingly for excitation spectra. If this is not the case, we are dealing with multiple centers. Excitation starting from thermally excited levels (such as A2) of the ground state must be reduced in energy (b) by the same amount than seen in the emission (a). As a consequence, we will see transitions with lower energy, in the excitation spectra. The inverse is true when we consider emission starting from thermally excited states (such as B2). Applying these rules to our Er:GaN system, it turns out that all transitions can be explained with the inclusion of thermally excited transitions. The measurement temperature is about 10 K and hence not only the lowest states (i.e. A1, B1, as indicated in Fig. 2) of the ground and excited state multiplets are populated. We can explain all transition energies when we consider excitation Fig. 1. Combined emission excitation spectral image plot of Er:GaN. The horizontal lines represent transitions from the ground level to the I11/2 state, while the vertical lines represent emission peaks from the I13/2 state to the I15/2 ground level. Fig. 2. Complete assignment of the emission peaks due to the I11/2 to I15/2 transition at around 980 nm observed under excitation in the 1.54 lm spectral region. 1060 N. Woodward et al. / Optical Materials 33 (2011) 1059–1062 transition starting from the 2nd level (A2) of the ground state multiplet aswell as emission transitions from the 2nd (B2) and 3rd level (B3)of the I13/2 excited state. Inboth cases, thiswill lead togroupsof transitions that are shifted relative to each other by an identical amount (i.e. the energyof the thermally excited state). In Fig. 1, these groups are indicated by lines of the same color/grayscale, with black lines indicating transitions from the respective lowest level, and the white and grey lines correspond to the groups from the 2nd and 3rd lowest levels. In the presented case, the original and final level is identical such that we can test the assignments for consistency. The separation indicated as a and b both represent the splitting of the two levels of the ground state and must hence be equal. This consistency condition is fulfilled with great accuracy, such that we are confident that themajority of lines can be assigned to amajority site. Using conventional excitation–emission spectroscopy, similar conclusions have been drawn independently for the same type of samples by Makarova et al. [4]. A strength of the CEES technique compared to traditional excitation–emission spectroscopy lies in the ability to identify weaker spectral features by exploiting the dynamic range of the CCD array. By overexposing the main site, a minority site becomes visible which exhibits an emission intensity that is weaker by a factor of about 40. These additional sites becomemore apparent when we consider up-conversion excitation. In the simplest case, we use a fiberamplified tunable 1.54 lm laser to excite the ions in two steps to the I9/2 state and observe the emission from the I11/2 state at around 980 nm. This case is essentially the inverse to the direct excitation and hence identification of the sites is easiest. We show emission spectra under these conditions in Fig. 2 in which the