PHYSICAL REVIEW B 74, 155310 ͑2006͒ Optical study of excitation and deexcitation of Tm in GaN quantum dots Thomas Andreev,1,* Nguyen Quang Liem,1,2 Yuji Hori,1,3 Mitsuhiro Tanaka,3 Osamu Oda,3 Daniel Le Si Dang,4 Bruno Daudin,1 and Bruno Gayral1,† 1CEA/CNRS/UJF Research Group Nanophysique et Semiconducteurs, DRFMC/SP2M/PSC CEA-Grenoble, 17 rue des Martyrs, 38054-Grenoble cedex 9, France 2Institute of Materials Science (IMS), Vietnamese Academy of Science and Technology, 18 Hoang Quoc Viet-Cau Giay-Hanoi, Vietnam and College of Technology, Hanoi National University, 144 Xuan Thuy-Cau Giay-Hanoi, Vietnam 3NGK Insulators, LTD 2-24 Sudacho, Mizuhoku, Nagoya, Japan CEA/CNRS/UJF Research Group Nanophysique et Semiconducteurs, Laboratoire Spectrométrie Physique (CNRS UMR5588), Université J Fourier, Bte Postale 87, 38402 Saint Martin d’Hères, France ͑Received 16 May 2006; published 13 October 2006͒ We report on the optical properties of molecular beam epitaxy grown GaN quantum dots ͑QDs͒ doped with Tm Under optical excitation above the fundamental energy of GaN QDs, the fundamental transition emission from the GaN QD host was not observed while bright emission from Tm3+ manifolds demonstrated the efficient energy transfer from the host to Tm3+ ions The photoluminescence as a function of temperature was fast quenched for transition from the high-lying manifolds state but very stable for the blue one resulting from the 1D2 → 3F4 transition Mechanisms for excitation to and deexcitation from Tm3+ ions are discussed DOI: 10.1103/PhysRevB.74.155310 PACS number͑s͒: 78.55.Cr, 68.55.Ln, 78.67.Hc, 85.60.Jb I INTRODUCTION Rare-earth trivalent ions ͑RE ͒ exhibit well shielded intra 4f-transitions extending from the near infrared to the ultraviolet, which are nearly independent from host materials For semiconductor hosts the band gap has to be wide enough to allow a good energy transfer probability from the semiconductor host to the rare earth dopants in order to achieve light emission in the whole visible range, for example red ͑Eu, Pr͒, green ͑Tb, Ho, and Er͒ and blue ͑Tm͒ Along this view the combination of the RE luminescence with wide band-gap nitride semiconductors has been found to be a promising solution for light emitting devices.1–4 For strong blue Tm light emission from the 1D2 → 3F4 the band-gap of the host has to be even wider than that of GaN, which can be realized by either adding Al, e.g., in AlGaN: Tm layers,5,6 or by using GaN: Tm QDs.7 In the case of Tm doping of GaN QDs during MBE growth we have shown by atomic force microscopy measurements that QDs are typically small, with heights of about nm and diameters of about 20 nm.7 For such sizes, although QDs undergo a quantum confined Stark effect due to a polarization induced internal field, the fundamental transition of the QDs peaks around eV, at much higher energy than the band-gap of GaN.7 GaN QDs have also other advantages compared to thin ͑Al͒GaN films: they are defect free regions and they act as efficient carrier confinement boxes Actually, we have demonstrated Eu doping of ͑In͒GaN QDs for red emission,8,9 Tb for green10 and Tm for blue emission For the main emission lines photoluminescence ͑PL͒ has been found for all GaN:RE QD systems to be thermally stable from liquid helium to room temperature For understanding the energy transfer mechanism a considerable effort has been made in the case of infrared emission of InP: Yb ͑Refs 11–14͒ and Si: Er.15–18 It is proposed that after band to band excitation of the semiconductor host 3+ 1098-0121/2006/74͑15͒/155310͑6͒ the generated free carriers can be captured by RE-related traps located near-band edge The electron-hole recombination energy is then used to excite the 4f electrons in RE3+ ion from the ground state to the excited states in a so called Auger process.19,20 Then excited RE3+ ions can relax by emitting light and/or transfer their energy back to the host material Comparatively, the literature concerning RE-doped GaN thin films is rather scarce.21 Moreover, the energy transfer mechanism in rare earth doped GaN QDs has not been studied yet although this is important for understanding of further device operations In previous work7 we have studied morphology of Tm doped GaN QDs embedded in AlN spacing layers and have primarily shown that GaN QDs can be efficiently doped with Tm during molecular beam epitaxy ͑MBE͒ growth The aim of this paper is to address in detail the excitation and deexcitation mechanism of Tm3+ ions in GaN QDs embedded in AlN spacing layers The related processes for high-lying excited states of Tm3+ ion, i.e., 1I6 and P1, have been addressed with a special care because the transition energies from these levels are similar and rather close to the fundamental energy of GaN QDs that mediate possible excitation and deexcitation processes Furthermore, a particular attention has been paid to the 1D2 → 3F4 transition which is the most thermally stable transition emitting blue light, and is therefore particularly important for applications Our discussion is based on the experimental data obtained from temperature dependent, excitation-wavelength dependent, and time-resolved PL measurements II EXPERIMENTAL All samples were grown on 1-m-thick AlN pseudosubstrates deposited by metal organic chemical vapor deposition 155310-1 ©2006 The American Physical Society PHYSICAL REVIEW B 74, 155310 ͑2006͒ ANDREEV et al on c sapphire.22 After a standard chemical degreasing procedure and acid cleaning, they were fixed onto a molybdenum sample holder, and introduced in a molecular beam epitaxy ͑MBE͒ chamber equipped with Al, Ga, and Tm effusion cells and a radio-frequency plasma cell to produce monoatomic nitrogen The GaN QDs were grown at a substrate temperature of 720 ° C following the Stranski-Krastanow growth mode, i.e., the QDs appear after the deposition of about GaN monolayers.23,24 The growth conditions were controlled with reflection high-energy electron diffraction ͑RHEED͒, which allows in situ monitoring of the QD formation During the growth of GaN, the Tm shutter was opened in order to dope the material Next, the QDs were capped by about 12 nm of undoped AlN This process was repeated to achieve a superlattice of 100 QD planes From the chosen growth conditions the Tm content inside the sample was estimated to be around 3% PL measurements were carried out with a frequency doubled Ar-ion laser emitting at 244 nm PL as a function of excitation wavelength was carried out with a tunable excitation source consisting of a 500 W high-pressure Xe lamp equipped with a Jobin-Yvon high-resolution double-grating monochromator ͑Gemini 180͒; the PL was analyzed by another Jobin-Yvon grating monochromator ͑Triax 550͒ and detected by either a CCD camera operating at liquid nitrogen temperature or a photomultiplier tube operating in the photon counting mode ͑Hamamatsu H8259͒ The excitation power density was about 200 W / cm2 at 350 nm The sample was mounted on the cold finger of a microcryostat which enabled us to record PL and PLE spectra at various temperatures between K and 300 K Time-resolved PL measurements were carried out using the frequency tripled output of a pulsed Ti:sapphire laser emitting at 250 nm with an average power of 0.6 mW The base repetition rate was 54 MHz, which was then divided using a cavity dumper, yielding a repetition frequency of 140 kHz Finally time-resolved PL was analyzed with a Hamamatsu Streak camera in single sweep mode FIG PL spectra from GaN: Tm QDs measured at various temperature between K and 300 K as indicated on the curves The spectra are normalized by the excitation power density and the integration time for detection The excitation source was a frequency doubled Ar-ion laser emitting at 244 nm Excitation power density was ϳ0.8 W / cm2 The three most intense transitions at K are indicated in the figure The ͑unstrained͒ GaN band-gap energy is indicated are stable between liquid helium and room temperature An intermediate quenching is observed for the blue 1G4 → 3H6 and infrared 3H4 → 3H6 transitions Detailed thermal behavior for the blue spectral range is displayed in the high resolution PL spectra in Fig This spectral range is of particular interest as it involves transitions stemming from 1D2 and from 1I6, and 1G4 mostly responsible for the blue light emission The 1G4 → 3H6 transition is a main emission peak in bulk GaN doped with Tm,5–7 however for Tm doped GaN QDs, it results only in weak luminescence This is due to the fact that for GaN QDs other transitions involving 1D2 and 1I6 are more efficiently excited than 1G4 → 3H6 ͑for bulk GaN, 1D2, and 1I6 transitions cannot be excited due to the too low bandgap of GaN͒ Therefore the 1G4 → 3H6 transition will not be further discussed in this paper III OPTICAL PROPERTIES OF GaN: Tm QDs The PL spectra of GaN: Tm QDs excited at 244 nm, i.e., well above the absorption band gap of GaN QDs are presented in Fig for various temperatures between K and 300 K The lack of emission from fundamental transition in GaN QD ͑in UV region͒ even at K indicates a high Tm concentration and a high energy transfer probability from the QD host to the Tm3+ ions Along with blue light, Tm3+ ions emit sharp emission lines extending over the whole visible range The identification of transitions resulting in these lines is mainly based on the energetic position and is in agreement with published data.5,6,25,26 In the present study, the transitions from 1I6, 1D2, 1G4, and 3H4 have been found While PL quenching for the different lines has been previously studied,7 we shall study here this behavior more in detail, notably with time resolved PL data Actually PL from each transition exhibits different thermal behavior: PL from the I6 level is fast quenched and hardly observed at temperatures higher than 100 K, whereas those starting from the 1D2 FIG High resolution PL spectra at various temperatures of Tm doped GaN QDs for the blue spectral range The spectra are normalized to the accumulation time Excitation: 250 nm from the Xe lamp Assigned transitions are indicated in the figure The vertical dotted lines are shown as a guide for the eyes 155310-2 PHYSICAL REVIEW B 74, 155310 ͑2006͒ OPTICAL STUDY OF EXCITATION AND… Emissions in the spectral range of 463 nm to 471 nm where both the 1D2 → 3F4 and 1I6 → 3H4 transitions can exist and give strong luminescence were analyzed in more detail It is clearly seen that most of the emission lines in the mentioned range are stable with temperature because the 1D2 level is located at much lower energy than 1I6, in agreement with published data.26 We assign it to 1D2 → 3F4 transition By contrast the line at the longer wavelength side ͑469.5 nm͒ is fast quenched and hardly observed at temperatures higher than 100 K Thus, this emission line is assigned as stemming from the higher-lying 1I6 state The fast thermal quenching of PL originating from 1I6 may result from various processes: ͑i͒ From phonons emission to relax down to lower-lying states of Tm3+ ion This process is unlikely as many phonons are needed to fill the energy gap between the starting and final discrete levels ͑ii͒ By phonon-assisted deexcitation to the host material This is a most reasonable process because transitions from high-lying 1I6 state can release an energy comparable to the energy range of the fundamental energy of GaN QDs Though we cannot determine the exact levels of the RE3+ ion corresponding to the fundamental edge of the semiconductor host, the near resonance energy of the 1I6 → 3H6 transition and fundamental gap of the host GaN QD at least makes more likely Auger transfer with phonon assistance Population of LO-phonons follows Bose-Einstein distribution, increasing fast at temperatures higher than 100 K hence being responsible for thermal quenching of this emission Similar result has been reported for InP: Yb ͑Refs 11–14͒ in which the temperature activated the energy back transfer from the guest RE ions to the host semiconductor Some other processes such as cross relaxation between adjacent Tm atoms for resonant 1I6 → 3H4 and 3H6 → 1G4 ͑this process takes place with increasing Tm concentration͒ or emitting spontaneous IR to lower levels could be considered but they are weakly dependent on temperature.27 Note that the phonon-assisted deexcitation from the 1D2 to the GaN QDs host is much less effective because any transition from the 1D2 to lower levels gives much less energy than the fundamental energy of the host On the other hand, the 1D2 excited state at lower energy can benefit from repopulation from other higher-lying levels during their quenching with temperature The above mentioned facts are possible reasons for the temperature stability of emission lines from the 1D2 excited state In Fig one can also see several emission lines between 465 nm and 468 nm, whose intensity increase with temperature These are believed to be Stark-split levels which are PL-observable depending on the temperature-dependent population factor A similar feature has been observed and discussed in the AlN: Tm system.26 The intermediate thermal quenching was observed for the blue 1G4 → 3H6 transition and for the infrared 3H4 → 3H6 transition This can be explained assuming that transitions from 1G4 or 3H43 are not in resonance for deexcitation processes and are not likely to benefit from repopulation with temperature so that the excitation mechanism is temperature independent In this case, the thermal quenching is likely due to thermally activated nonradiative deexcitation only Based on the energetic positions of emission lines and their thermal quenching behaviors an energy diagram can be established as shown in Fig 3.25 As the blue and green spec- FIG Energy diagram of Tm3+ ions and observed transitions in GaN host The energy range of the QDs fundamental transition is indicated on the energy scale The indicated wavelengths present means values for transitions tral regions are concerned, two sets of transitions result in emission lines at very similar energies in steady-state PL spectra, but exhibit distinct decay times, namely ϳ0.5 s for the emissions at 466.3 nm, 469.7 nm, 534.5 nm, 537.5 nm ͓Figs 4͑a͒ and 6͑b͔͒ and ϳ1.2 s for the emission at 532 nm ͓Fig 6͑b͔͒ The former emission group has been identified as resulting from the 1I6 level ͑1I6 → 3H4 for the 466.3 nm and FIG ͑a͒ The open circles correspond to the PL from GaN: Tm QDs measured at K with the time resolved setup ͑see experimental part͒ The spectrum has been fitted with a multiple Lorentzian containing peaks Measured decay times and assigned transitions are also indicated in the figure ͑b͒ Time resolved PL signal of the D2-3F4 transition measured at 468 nm at K ͑upper curve͒ and 300 K ͑lower curve͒ Fitting of the measurements with monoexponential decay are plotted bold The inset shows decay times for the D2-3F4 transition at 468 nm and 465 nm as a function of temperatures 155310-3 PHYSICAL REVIEW B 74, 155310 ͑2006͒ ANDREEV et al 469.7 nm lines, and 1I6 → 3F3 for the 531 nm, 534.5 nm, and 537.5 nm lines in agreement with the different thermal quenching of the PL Further information can be gained about the excitation and deexcitation of Tm3+ ion in GaN QD host based on time-resolved PL measurements As the emission lines in the 463–471 nm range are overlapped, before analyzing the decay times the spectra have been deconvoluted with a multiple Lorentzian curve fit ͓Fig 4͑a͔͒ For the two strongest isolated lines ͑at 465.5 nm and 468.6 nm͒, the PL decays could be well fitted with a monoexponential function Figure 4͑b͒ presents the measured and fitted curves for the PL decays of the 468.6 nm emission line at K and 300 K using the typical equation: y = Ae−t/, where is the decay time, t the elapsed time after the laser pulse, and A a fitting parameter For lines that include a contribution from nearby stronger lines ͑for which a monoexponential decay fit could be done͒ the PL decay was fitted as the sum of a known exponential decay ͑stemming from the overlapping stronger line͒ and another exponential ͑that gives the decay time of the line of interest͒ For the two main blue emission lines at 468.6 nm and 465.5 nm and the one at the higher energy side ͑464.5 nm͒ the decay times obtained at K are similar, i.e., respectively 3.2 s, 2.6 s, and 2.5 s This is to be expected since the three mentioned lines are just originating from the Stark splitting in C3v symmetry of the 1D2 → 3F4 transition as presented in the energy diagram ͑Fig 3͒ However, it is worth noticing that even if the three mentioned lines originate from the same 1D2 → 3F4 transition, there is a trend to observe longer decay time for the higherenergy Stark component This feature has also been observed for the Stark-splitting components of the 1I6 → 3H4 transition that result in the 466.2 nm ͑0.56 s͒ and 470 nm ͑0.45 s͒ lines; and the 1I6 → 3F3 transition that result in the 531.6 nm ͑1.2 s͒, 534.5 nm ͑0.6 s͒, and 537.5 nm ͑0.5 s͒ lines This result is difficult to explain Also in tendency the measured decay time of the 1D2 → F4 transition at K has been found to be ϳ2.6 s, which is longer than those for the 1I6 → 3F3 ͑1.2 s͒ This is likely due to the fact that the 1D2 → 3F4 transition is spin forbidden as in the case of the 1I6 → 3F3 transition In addition the 1D2 state is located much below the 1I6 and P1 so that energy released from the 1D2 transitions to any manifolds is far from resonances with fundamental energy of the host material or near-fundamental-edge traps to favor the back-energy transfer which generally contributes to the reduction of the measured decay time Note that a similar trend, i.e., a longer decay time for the 1D2 excited state compared to 1I6 was observed in Tm doped AlGaN.5 Although the longest decay time has been found for emissions resulting from the 1D2 → 3F4 transition, as a result of the reduced energy back transfer to the host material, a reduction of the decay time has been found when increasing temperature, from ϳ2.6 s at K to ϳ1 s at 300 K This reduction in decay time can be attributed to an increase in nonradiative recombination with increasing temperature due to an increased back transfer to the host.11 This increase in the nonradiative recombination rate is however not correlated with a corresponding decrease in luminescence intensity Indeed, from K to 300 K, the luminescence intensity FIG PLE spectra at K and 300 K for the 468 nm emission line of Tm3+ ion in GaN QDs Spectra have been normalized to the excitation intensity and presented in logarithmic scale The horizontally dotted line is a guide to the eye The ͑unstrained͒ GaN bandgap energy is indicated for the 1D2 → 3F4 transition is constant within ±20% As we are in the low excitation regime ͑well below saturation of the transitions͒, this suggests that the excitation efficiency of the D2 level increases with increasing temperature and that this effect compensates for the nonradiative losses In particular, D2 could benefit from depopulation of 1I6 with temperature, as already discussed above The excitation scheme proposed for the 468 nm emission is as follows: ͑i͒ photoexcited and/or generated carriers ͑in host GaN QDs͒ → ͑ii͒ traps ͑isoelectronic Tm3+ ions͒ → ͑iii͒ 4f-electrons of Tm3+ ions ͑by near resonant energy transfer to high-lying states such as 1I6 and P1 and/or nonresonant process to 1D2͒ → ͑iv͒ transitions ͑e.g., 1D2 → 3F2͒ to emit light From the time-resolved spectra taken at K and 300 K ͓Fig 4͑b͔͒, it appears that the rise time is shorter than 50 ns This value can be accounted for by the energy transfer rate from GaN QDs to the 4f electrons of Tm3+ ions through processes ͑i͒ to ͑iii͒ As the energy transfer processes from the QDs to the RE3+ ions take place faster than other nonradiative processes, one can practically expect high efficiency luminescence from the RE-doped GaN QDs This is also consistent with the fact that we did not observe the fundamental level emission for GaN QDs doped with Tm because energy transfer processes to Tm3+ ions are much faster than other possible processes ͑including radiative recombination͒ for electron-hole pairs trapped in QDs As already observed for InGaN QDs doped with Europium,8,9 the photoluminescence excitation spectra for Tm doped GaN QDs at 300 K display a gradual absorption edge related to the QD absorption ͑Fig 5͒ This confirms that the transfer to the Tm states is indeed mediated by the QD states, and that in these samples the QDs have rather high energy levels ͑absorption below 330 nm= 3.75 eV͒ To gain more insight in the excitation mechanism, we checked how the different emissions behave in PL as a function of the excitation wavelength in a range extending from 230 nm to 330 nm to cover the whole band gap energy values corresponding to the QDs size distribution The results are shown in Figs 6͑a͒–6͑c͒ for the most intense emissions In the blue spectral range ͓Fig 6͑a͔͒ the emission lines at 155310-4 PHYSICAL REVIEW B 74, 155310 ͑2006͒ OPTICAL STUDY OF EXCITATION AND… FIG High resolution PL spectra in blue spectral range for Tm doped GaN QDs measured at 300 K with various excitation wavelengths as indicated in the figure The spectra are normalized to the power density and the accumulation time The spectra under 290 nm and 310 nm excitation are multiplied with a factor and 10 for clarity Assigned transitions are also indicated in the figure FIG PL spectra at K for GaN: Tm QDs measured with different excitation wavelengths ͑indicated at the curves͒ in the ͑a͒ blue, ͑b͒ green, and ͑c͒ infrared spectral region The spectra are normalized by the excitation power density and the integration time for detection Spectra with low emission intensity are multiplied by a factor for clarity The arrows indicate Tm3+ emission from the AlN spacing layer ͑see text͒ Assigned transitions are shown in the figure The decay times for the green spectral region are also indicated in ͑b͒ with respect to its measured wavelength position 464.5 nm, 465.5 nm, and 468.6 nm have been assigned to the 1D2 → 3F4 transition Also in agreement with the above discussion of the thermal quenching ͑Fig 2͒ is the 1I6 → 3H4 transition emitting around 466.2 nm and 470 nm, and the 1G4 → 3H6 emission at 479 nm Emissions from the higher excited states ͑ 1I6͒ tend to be weakened when using longer excitation wavelengths As a consequence the corresponding emissions from the 1I6 → 3H4 transition and the I6 → 3F3 transition could not be detected by exciting the sample at 310 nm ͓see Figs 6͑a͒ and 6͑b͔͒ In agreement with this behavior transitions from the 1D2 excited state show very weak intensity for excitation wavelengths longer than 310 nm, while the infrared 3H4 → 3H6 transition could be well detected at an excitation wavelength of 380 nm It is possible that the excitation wavelength dependent effect be assigned to the GaN QDs size distribution because the excitation energy transfer to the high-lying levels of Tm3+ ions needs a relatively high energy from GaN QDs, which corresponds to the smaller sizes dots By exciting the sample with shorter wavelengths ͑240 nm͒, Tm3+ ion luminescence from the AlN spacer is getting visible ͓Figs 6͑a͒ and 6͑b͔͒: in agreement with cathodoluminescence studies in Ref 7, the weak emissions at 462.5 nm and 467 nm are consistent with the presence of Tm3+ ions in the AlN spacing layer that contains some defect-related energy states where carriers can be captured so that they then can transfer their energy to Tm3+ ions One point we would like to mention regarding the temperature-activated emission lines ͑within 466.5 nm to 467.5 nm͒ is that they are observable only at short excitation wavelength ͑230 nm to 270 nm͒ and high temperatures ͑Figs and 7͒ At long excitation wavelengths these lines did not appear This again supports the assumption that they come from the high-lying 1I6 state that is not likely to be excited with long excitation wavelength It is interesting to note that emission from the 1D2 excited state can be detected, using excitation energy ͑380 nm͒ even below its energetic position ͑370 nm͒ This could be explained by two photon processes, however further enlightenment is required IV CONCLUSIONS In conclusion, we have studied in detail the optical properties of GaN QDs doped with Tm by analyzing the PL characteristics as functions of sample temperatures, excitation wavelengths, and measuring the decay times for some transitions from the high-lying Tm3+ manifolds as a function of temperature Tm3+ ions are excited through the excitation of GaN QDs host with shallow traps which are gradually distributed below their fundamental edge and observable in the 155310-5 PHYSICAL REVIEW B 74, 155310 ͑2006͒ ANDREEV et al ACKNOWLEDGMENTS low-temperature PLE spectrum The near resonant Auger process with LO-phonon assistance is proposed for the fast thermal quenching transition stemming from the 1I6 level The 1D2 → 3F4 transition results in stable blue emission with increasing temperature The PL by selective excitation is in good agreement with this thermal behavior as longer wavelength excitation is only possible for lower-lying energy states of Tm3+ ions We acknowledge Marlène Terrier, Yann Genuist, and Yoann Curé for their technical assistance Joël Bleuse is acknowledged for experimental contributions One of the authors ͑N.Q.L.͒ thanks the National Programme for Basic Research ͑Vietnam͒, CNRS, and CEA ͑France͒ for financial support *Present address: EADS Astrium GmbH, 81663 Munich, Germany 14 M †Corresponding author Electronic address: bruno.gayral@cea.fr D S Lee and A J Steckl, Appl Phys Lett 81, 2331 ͑2002͒ H J Lozykowski, M Jadwisienczak, and I Brown, Appl Phys Lett 74, 1129 ͑1999͒ H Bang, S Morishima, J Sawahata, J Seo, M Takiguchi, M Tsunemi, K Akimoto, and M Nomura, Appl Phys Lett 85, 227 ͑2004͒ K Lorenz, U Wahl, E Alves, S Dalmasso, R W Martin, K P O’Donnell, S Ruffenach, and O Briot, Appl Phys Lett 85, 2712 ͑2004͒ U Hömmerich, Ei Ei Nyein, D S Lee, A J Steckl, and J M Zavada, Appl Phys Lett 83, 4556 ͑2003͒ D S Lee and A J Steckl, Appl Phys Lett 83, 2094 ͑2003͒ T Andreev, Y Hori, X Biquard, E Monroy, D Jalabert, A Farchi, M Tanaka, O Oda, D Le Si Dang, and B Daudin, Phys Rev B 71, 115310 ͑2005͒ T Andreev, E Monroy, B Gayral, B Daudin, N Q Liem, Y Hori, M Tanaka, O Oda, and D Le Si Dang, Appl Phys Lett 87, 021906 ͑2005͒ Y Hori, X Biquard, E Monroy, D Jalabert, F Enjalbert, D Le Si Dang, M Tanaka, O Oda, and B Daudin, Appl Phys Lett 84, 206 ͑2004͒ 10 Y Hori, T Andreev, D Jalabert, E Monroy, D Le Si Dang, B Daudin, M Tanaka, and O Oda, Appl Phys Lett 88, 053102 ͑2006͒ 11 H J Lozykowski, A K Alshawa, and I Brown, J Appl Phys 76, 4836 ͑1994͒ 12 A Taguchi, M Taniguchi, and K Takahei, Appl Phys Lett 60, 965 ͑1992͒ 13 A Taguchi and K Takahei, J Appl Phys 79, 3261 ͑1996͒ A J Klik, T Gregorkiewicz, I V Bradley, and J-P R Wells, Phys Rev Lett 89, 227401 ͑2002͒ 15 H Kühne, G Weiser, E I Terukov, A N Kusnetsov, and V Kh Kudoyarova, J Appl Phys 86, 896 ͑1999͒ 16 I Tsimperidis, T Gregorkiewicz, H H P Th Bekman, and C J G M Langerak, Phys Rev Lett 81, 4748 ͑1998͒ 17 W Fuhs, I Ulber, G Weiser, M S Bresler, O B Gusev, A N Kuznetsov, V Kh Kudoyarova, E I Terukov, and I N Yassievich, Phys Rev B 56, 9545 ͑1997͒ 18 S Coffa, F Priolo, G Franzò, V Bellani, A Carnera, and C Spinella, Phys Rev B 48, 11782 ͑1993͒ 19 S Schmitt-Rink, C M Varma, and A F J Levi, Phys Rev Lett 66, 2782 ͑1991͒ 20 H J Lozykowski, Phys Rev B 48, 17758 ͑1993͒ 21 C W Lee, H O Everitt, D S Lee, A J Steckl, and J M Zavada, J Appl Phys 95, 7717 ͑2004͒ 22 T Shibata, K Asai, T Nagai, S Sumiya, M Tanaka, O Oda, H Miyake, and K Hiramatsu, in GaN and Related Alloys—2001, edited by J E Northrup, J Neugebauer, D C Look, S F Chichibu, and H Riechert, MRS Symposia Proceedings No 693 ͑Materials Research Society, Pittsburgh, 2002͒ 23 C Adelmann, J Simon, G Feuillet, N T Pelekanos, B Daudin, and G Fishman, Appl Phys Lett 76, 1570 ͑2000͒ 24 I N Stranski and L Krastanow, Sitzungsber Akad Wiss Wien, Math.-Naturwiss Kl., Abt 2B 146, 797 ͑1937͒ 25 G H Dieke, Spectra and Energy Levels of Rare Earth Ions in Crystals ͑Wiley, New York, 1968͒ 26 J B Gruber, U Vetter, H Hofsäss, B Zandi, and M F Reid, Phys Rev B 70, 245108 ͑2004͒ 27 M Dejneka, E Snitzer, and R E Riman, J Lumin 65, 227 ͑1995͒ 155310-6 ... nm, 534.5 nm, and 537.5 nm lines in agreement with the different thermal quenching of the PL Further information can be gained about the excitation and deexcitation of Tm3 + ion in GaN QD host based... within ±20% As we are in the low excitation regime ͑well below saturation of the transitions͒, this suggests that the excitation efficiency of the D2 level increases with increasing temperature and. .. in nonradiative recombination with increasing temperature due to an increased back transfer to the host.11 This increase in the nonradiative recombination rate is however not correlated with