a) Er³⁺ PL intensity and spectra

Figure 1 shows the room-temperature, infrared PL spectra of Er doped GaN on Al₂O₃ prepared by MOMBE. No significant spectral differences were observed for Er doped GaN on Sapphire or Si substrates. For comparison, the Er³⁺ PL spectra of AlN: Er (MOMBE) and Er implanted GaN are also shown in figure 1.Reference Wilson, Schwartz, Abernathy, Pearton, Newman, Rubin, Fu and Zavada ^{9 ,} Reference MacKenzie, Abernathy, Pearton, Hömmerich, Wu, Schwartz, Wilson and Zavada ¹⁴ All samples were excited with the 363-366.8nm UV output of an Argon-Ion laser. UV excitation was chosen to simulate carrier-mediated Er³⁺ excitation as in GaN: Er electroluminescence devices. All samples showed characteristic Er³⁺ PL centered at 1.54 μm, which can be assigned to the intra-4f Er³⁺ transition Reference Coffa, Polman and Schwartz ⁴ I_13/2--->Reference Coffa, Polman and Schwartz ⁴ I_15/2. The Er³⁺ PL spectra from in-situ Er doped GaN and AlN are very similar. They exhibit nearly featureless spectra (25-35nm @FWHM) suggesting a homogenous distribution of Er sites as typically observed from Er doped glasses.Reference Miniscalco ^{22 ,} Reference Ainslie ²³ Er peaks located at 1.517 μm, 1.540 μm, 1.548 μm, and 1.555 μm are either due to Stark splittings and/or multiple Er sites. The overall spectral width of the Er³⁺ PL from Er implanted GaN was significantly larger (100nm@FWHM) than that observed from Er doped GaN and AlN. Moreover, Er implanted GaN exhibits a more sharply structured Er³⁺ PL than the in-situ Er doped III-Nitride samples. Based on previously published data 9,10,11,12,13, the more complex and broad Er³⁺ PL seems to be characteristic for Er implanted GaN samples and is due to a combination of different Er sites and Stark splittings of Er ions in low symmetry sites. For example, Kim et al. identified up to 4 Er sites in Er implanted GaN grown by MOCVD.Reference Kim, Rhee, Turnbull, Reuter, Li, Coleman and Bishop ¹² The authors demonstrated that it was possible to selectively excite these four sites by three different below-gap absorption processes and through direct intra-4f excitation. Important for future device development was the observation that only a small fraction of incorporated Er³⁺ ions are excitable through carrier-mediated processes.Reference Rhee, Kim, Li, Coleman and Bishop ²⁴ Time-resolved PL studies of in-situ Er doped AlN (MOMBE) have indicated the existence of at least two different classes of Er³⁺ sites with distinct lifetime and excitation schemes.Reference Wu, Hömmerich, MacKenzie, Abernathy, Pearton, Schwartz, Wilson and Zavada ²⁵ These results underline that it is crucial to identify the incorporation and excitation mechanism of Er³⁺ in different sites in order to optimize the material preparation for device applications.

Figure 1: Room-temperature Er³⁺ PL spectra of Er implanted GaN and in-situ Er doped GaN and AlN.

Figure 2 depicts a comparison of the absolute Er³⁺ PL intensities of GaN: Er (MOMBE) for below and above-gap excitation at room-temperature. Under below-gap excitation, GaN: Er (TEGa) with high oxygen and carbon background ([O]∼10²⁰cm⁻³, [C]∼10²¹cm⁻³) showed an Er³⁺ PL peak intensity two orders of magnitude larger than GaN: Er (Ga) with low O and C backgrounds (<10¹⁹cm⁻³). It is interesting to note that also GaN: Er/Si showed strong 1.54μm PL at room temperature, which makes this material attractive for integration with Si based optoelectronics.²¹ Drastic improvements of the absolute Er³⁺ PL intensity have been observed for Er and oxygen codoped Si and GaAs. The enhanced 1.54μm PL was attributed to an increased concentration of optically active Er³⁺ ions and a more efficient Er³⁺ PL excitation.^{26,27,28,29,30} Our results for below-gap Er³⁺ excitation in GaN suggest that the incorporation of O and C introduces beneficial mid-gap states that provide efficient energy transfer pathways for Er³⁺ in GaN. Support for this idea was obtained from photoluminescence excitation (PLE) studies shown in figure 3. In PLE the effectiveness of stimulating 1.54 μm Er³⁺ PL is measured as a function of excitation wavelength. The PLE spectra of GaN: Er/Si (TEGa) reveal that the incorporation of high O and C backgrounds leads to a broad PLE band extending over the entire visible region (∼400-800 nm). The GaN: Er (Ga) sample with low C and O content did not show this broad PLE band, but direct intra-4f Er transitions (see e.g. ∼525nm). Therefore, the visible output from an Argon-Ion laser leads to a significantly more efficient excitation of Er³⁺ in GaN:Er/Si (TEGa) than in GaN: Er/Si (Ga).

Figure 2: Comparison of Er³⁺ PL intensity from Er doped GaN (MOMBE) for a) below-gap and (b) above-gap excitation (300K).

Figure 3: PLE of GaN: Er/Si samples with varying oxygen and carbon content. The incorporation of high oxygen and carbon backgrounds (upper graph) leads to a broad Er³⁺ excitation band covering the entire visible region.

Under above-gap excitation (figure 2b) all GaN: Er (MOMBE) samples showed a greatly reduced Er³⁺ PL intensity relative to below-gap excitation, independent of carbon and oxygen concentration. A similar PL reduction was reported for Er implanted GaN.³¹ It is somewhat surprising that under above-gap excitation the Er³⁺ PL intensities of samples with high oxygen and carbon backgrounds is only roughly twice as strong as the Er³⁺ PL observed from low O and C contents. These observations suggest that either only a fraction of optically active Er³⁺ ions are excited and/or the excitation efficiency is reduced under above-excitation. Moreover, it also indicates that high oxygen and carbon concentrations do not necessary lead to an enhanced luminescence for carrier-mediated Er³⁺ excitation. More work is necessary to address this issue in more detail.

b) Er PL quenching and lifetime

Figure 4 shows the integrated Er³⁺ PL intensity of Er doped GaN samples for the temperature range from 15-500K. The Er³⁺ PL was excited either with the broad-band visible (Fig. 4a) or uv (Fig. 4b) outputs of an Argon ion laser corresponding to below and above-gap Er³⁺ excitation. Under below-gap excitation a significantly different Er³⁺ PL quenching behavior was observed for samples with varying O and C contents (see Figure 4a). As was previously observed for Er and oxygen codoped Si and GaAs, GaN: Er samples with high O and C concentrations showed a significantly reduced Er³⁺ PL quenching compared to samples with low C and O backgrounds.^{26,27,28,29,30} Between 15K and room-temperature samples with high O and C backgrounds showed Er³⁺ PL quenching of only ∼20%, whereas the 1.54μm luminescence from low O and C concentration samples decreased by ∼90% over the same temperature range. At higher temperatures also the O and C rich samples showed an enhanced Er³⁺ PL quenching and the luminescence decreased to roughly 10% of their low temperature value at 500K. Under above-gap excitation the situation was different. As shown in Fig. 4(b) hardly any difference in Er³⁺ PL quenching behavior was observed for samples with varying oxygen/carbon content. As mentioned above, the oxygen and carbon concentration does not greatly effect the carrier-mediated Er³⁺ excitation. Up to room temperature the Er³⁺ PL decreased for all four samples by less than 50% relative to its low temperature value. When heating the samples above room temperature, we observed a steady decrease of the Er³⁺ PL with increasing temperature. At 550 K, the Er³⁺ PL from all samples had decreased by nearly 90% of its low temperature value.

Figure 4: Temperature dependence of the integrated PL intensity from in-situ Er doped GaN for above and below gap excitation.

To gain more insight in the observed Er³⁺ PL quenching, we performed initial luminescence lifetime studies on the two oxygen and carbon rich samples, GaN: Er/Si (TEGa) and GaN:Er/Sapphire (TEGa). These samples showed the strongest Er³⁺ PL signal, which allowed us to carry out lifetime studies for the entire range 15-500 K. It was observed that the Er³⁺ PL decay curves were non-exponential which provides further support for the existence of a wide distribution of Er³⁺ sites. To describe the luminescence decay an average lifetime <τ> was used:

where I(t) is the luminescence decay curve and t is the time.^32,33 The low temperature lifetime of both samples was found to be ∼130μs and decreased to ∼100μs at room temperature. At 500K the lifetime had shortened to ∼50μs. These lifetimes are significantly shorter than previously reported lifetimes for Er implanted GaN, which were in the order of 1-2ms.^10,16 It is possible that the short lifetimes are due to a higher oscillator strength of Er³⁺ transitions in MOMBE grown samples compared to Er implanted GaN. In order to confirm this explanation direct absorption measurements of intra-4 Er transitions are necessary. It is also conceivable that the Er³⁺ lifetimes are shortened due intra-4f non-radiative decay or cooperative processes such as upconversion or energy migration.³⁴ Power dependent lifetime studies, however, did not show any change in lifetime with increasing excitation density suggesting that the later processes are not dominant in GaN:Er.

It is instructive to compare the temperature dependence of PL intensity and lifetime as shown in figure 5. In the following discussion it is assumed that the Er³⁺ PL intensity is proportional to the number (or concentration) of excited Er ion (N_Er*) and the radiative decay rate (w_rad). Under steady state condition and low pump power intensities the Er³⁺ PL signal can be written as:

where C is a constant, I_p is the pump intensity, λ_p is the pump wavelength, σ_ex is the excitation cross-section defined as the product of pump efficiency and absorption cross-section, N_tot is the total Er concentration, w_rad is the radiative decay rate, and τ is the luminescence lifetime. Under the assumption that the radiative decay rate and total Er concentrations are temperature independent, the Er³⁺ PL quenching at higher temperatures can be explained by either the onset of nonradiative deexcitation processes reducing the Er³⁺ lifetime and/or by a decrease of the excitation efficiency.

Figure 5: Comparison of the temperature dependence of Er³⁺ intergrated PL and lifetime for in-situ Er doped GaN.

The comparison of the temperature dependent Er³⁺ PL intensity and lifetime enables one to separate the above mentioned Er³⁺ PL quenching mechanism. As can be inferred from figure 5, the Er³⁺ lifetime follows the change in integrated PL intensity very closely up to room temperature. This means that between 15 and 300K the small decrease in Er³⁺ PL intensity is due to nonradiative decay. Above room-temperature the lifetime changes less than the integrated Er³⁺ PL intensity. This suggests that Er³⁺ PL quenching occurring above room temperature is not only due nonradiative deexcitation, but also due to a reduction in excitation efficiency. Nearly the same Er³⁺ PL intensity and lifetime behavior was observed for GaN:Er/Si (not shown). Similar to other Er doped semiconductor systems, we assume that the excited Er ion can decay either through emission of a 1.54 μm photon or through non-radiative energy transfer processes to the host (also referred to as “backtransfer” ^28,35). It is further assumed that internal Er³⁺ quantum efficiency does not change with temperature. The change in lifetime is then described by a thermally activated nonradiative decay process, which is controlled by Boltzmann’s law:

where τ_rad is the radiative lifetime, E is the thermal activation or threshold energy for non-radiative decay, and B is a constant. Fitting equation 3 to the lifetime data shown in figure 5 gives the following parameter values: τ_rad= 124.6μs, B=31.6, and an activation energy E=132meV. The fit result suggests the existence of a defect-related level close to Er³⁺, which acts as a particular effective channel for nonradiative energy transfer between Er³⁺ and the host. The deexcitation of Er ions through some mid-gap level has been discussed before for Er doped Si and GaAs. In the backtransfer model as proposed by Takahei et al.²⁸, the existence of an Er-related level provides the possibility of energy backtransfer where excited Er ions transfer an excited carrier back to the defect level followed by a thermalization into the conduction band. This model for the deactivation of excited Er³⁺ ions may also be relevant for GaN:Er, however, no experimental evidence for an Er-related defect level has yet been reported. A more detailed comparison of the Er³⁺ lifetimes of GaN: Er samples with varying oxygen and carbon content is in progress and promises to provide deeper insight in the Er excitation and deexcitation mechanisms.

c) visible PL from GaN: Er

Very recently green luminescence was observed from GaN: Er grown by solid source molecular beam epitaxy under above-gap excitation.³⁶ Narrow lines centered at 537nm and 558nm were identified to be due to the intra-4f Er transitions ²H_11/2→⁴I_15/2 and ⁴S_3/2→⁴I_15/2. Subsequently the same authors reported green electroluminescence from Er doped GaN Schottky barrier diodes.³⁷ In figure 6 we show the visible PL spectra of GaN: Er/Sapphire prepared by MOMBE. We observed no green emission lines from any of the investigated samples under below (488nm) or above-gap (325nm) excitation. Interestingly, however, for below-gap excitation we observed a broad red luminescence band, which showed sharp structure at around 665 nm and 690 nm. This emission was not found for the GaN: Er/Si samples. Preliminary studies indicated that the peaks at 665 nm and 690 nm had lifetimes in the millisecond range. The peak at 665 nm is assigned to the ⁴F_9/2→⁴I_15/2 transition of Er³⁺ ions, whereas the origin of the 690 nm peak is not yet know. The data suggest that the 690 nm line is a pure electronic transition (no-phonon line) accompanied by a vibronic sideband at around 720 nm. A more detailed study is currently being undertaken to identify the origin of the red luminescence from GaN:Er/Sapphire.

Figure 6: Visible PL spectrum of GaN: Er/Sapphire for below-gap excitation at room-temperature.