Si nanowires

Journal of Luminescence 155 (2014) 293–297

Contents lists available at ScienceDirect

Journal of Luminescence journal homepage: www.elsevier.com/locate/jlumin

Dynamics of stacking faults luminescence in GaN/Si nanowires K.P. Korona a,n, A. Reszka b, M. Sobanska b, P.S. Perkowska a, A. Wysmołek a, K. Klosek b, Z.R. Zytkiewicz b a b

Faculty of Physics, University of Warsaw, ul. Hoza 69, 00-681 Warsaw, Poland Institute of Physics, Polish Academy of Sciences, al. Lotnikow 32/46, 02-668 Warsaw, Poland

art ic l e i nf o

a b s t r a c t

Article history: Received 7 April 2014 Received in revised form 26 June 2014 Accepted 30 June 2014 Available online 8 July 2014

Evidences are shown that excitons at stacking faults (SFs) in GaN nanowires (NWs) behave like 2dimentional particles in quantum wells. The SFs were studied in samples of coalescent GaN nanowires grown on silicon substrate. They emit strong luminescence at room temperature what promises possible applications of SF for light emitting devices. At 4 K, the NWs exhibit sharp near-bandgap luminescence with the strongest signal from donor-bound excitons (DX) at 3.471 eV and from stacking faults (SF) at 3.42 eV. Observations of the SF dynamics show spectral diffusion of energy in time (shift up to ΔEd ¼8 meV) and S-shaped energy–temperature dependence. Both these features are often reported as characteristic for quantum wells. Moreover we report on two dynamical effects, which confirm two-dimensional character of excitons in SFs. First, decrease of radiative recombination rate at higher temperature rR ¼1/(αT þt0), shows that excitons have momentum. Second, the SF luminescence peak has an exponential slope, i.e. is similar to thermal distribution, which points to kinetic energy of the excitons. The effective temperature calculated from the shape of the SF peak was in the range of few tens of Kelvins. It decreased in time with the cooling time τC ¼0.37 7 0.05 ns. & 2014 Elsevier B.V. All rights reserved.

Keywords: GaN/Si nanowires Time-resolved luminescence 2D excitons Exciton cooling

1. Introduction This paper is on the crossroad of two fields of current interest: semiconductor nanowires (NWs) that lately developed to the application level and stacking faults (SF) physics. We show that SF in NWs have properties similar to quantum wells (QWs) not only in energy level (what was postulated from many years [1], [2]) but also in dynamical properties of carriers located in the SF. We present strong evidence that they behave like quasi twodimensional particles just like in the QWs. It is worth to notice that the SFs have energy level below GaN band absorption, so it is in GaN transparency range. Their PL intensity is high. Moreover they emit light even at room temperature, so they can be used in optoelectronic applications. From a few years a new interest and growing number of publications [3], [4–6] on GaN nanowires are observed that is connected with development of growth techniques resulting in higher quality of NWs obtained. For example, achievements on a coherent light source, also termed a polariton laser, have been lately reported [6].

n

Corresponding author. Tel.: þ 48 225532209; fax: þ48 226219712. E-mail address: [email protected] (K.P. Korona).

http://dx.doi.org/10.1016/j.jlumin.2014.06.061 0022-2313/& 2014 Elsevier B.V. All rights reserved.

The stacking faults are commonly observed by electron microscopy. High-resolution cathodoluminescence revealed that the excitons bound to basal plane stacking faults of I1 type [2] emit at 3.42 eV [3,7,8]. Theoretical description of stacking faults in GaN was developed in last seventeen years [1,2,9,10]. The theory is based on assumption that SF is a kind of quantum well made of cubic GaN inserted between barriers of hexagonal GaN. This gives exactly the energy levels and excitonic properties of SFs. Some similarities between SFs and QWs were also observed experimentally, for example S-shaped temperature dependence of the peak positions [11] that is characteristic for QWs. In this work we present dynamic optical properties of stacking faults studied at various-temperatures by time-resolved photoluminescence on coalescent GaN nanowires. Such samples are chosen since they are expected to contain significant number of SF created as the result of coalescence of densely arranged NWs [12].

2. Experiment The GaN nanowires were grown at  740 1C by plasma-assisted molecular beam epitaxy on in-situ nitridized Si(1 1 1) substrates. No catalyst was used for nucleation of NWs. The growth direction

294

K.P. Korona et al. / Journal of Luminescence 155 (2014) 293–297

luminescence at room temperature can be resolved into two peaks at 3.36 eV (emitted by SF) and 3.41 eV (band to band transitions). At low temperature (4–100 K) a strong emission could be observed in the near bandgap region at energies from 3.3 to 3.5 eV (see Fig. 2). In the TRPL one could distinguish bands well known from studies of good quality GaN layers [14,15]: free excitons (FX) at 3.48 eV, excitons bound to donors (DX) at 3.471 eV, excitons bound to acceptors (AX) 3.466 eV, and donor– acceptor transitions (DA) at 3.29 eV (very weak). Widths of bound exciton peaks (DX and AX) were between 1.5 and 3 meV. The positions of these peaks were exactly the same as in unstrained homoepitaxial GaN samples [16]. Moreover, two extra features could be observed in the time-resolved PL spectra. The first one was a relatively narrow peak at 3.45 eV that had a lifetime of about 0.6 ns (longer than DX lifetime of 0.2 ns). This peak was reported in literature as characteristic for GaN nanowires [4] and related to excitons bound to surface defects [5], [17]. The second feature was a peak at 3.42 eV that was related to I1 stacking faults. This identification is already well established in the literature of Refs. [7,8]. The other SF-related peaks (reported for example by Lähnemann et al. [18]) were weak, that was probably due to growth mode of our NWs. For example I2 peak can be probably observed in Fig. 2 at 3.33 eV. The identification is not clear since the 3.33 eV line could be also the LO phonon replica of the I1 SF line (I1SF,LO). The dynamics of excitonic luminescence for the initial 0.2 ns was nonexponential (see Fig. 2B). Such behavior was already reported for NWs [19,20]. Then the dynamics of the DX peak was linear with lifetime of 0.15–0.2 ns. This value is few times shorter than in standard GaN [14,15]. Similar short lifetimes of NWs PL were also reported by others [20,21]. The SF decay was exponential with time constant 0.7–0.8 ns (depending on position on the sample). This time was longer than the DX decay time. It is probably due to electron–hole separation in SF caused by electric field [10]. During a short time after the laser pulse, the SF and the FX peaks are asymmetrically broadened. Their high-energy wings

was [000–1] (N-polarity). The high-resolution electron microscopy reveals that the nanowires are nearly defect free and are well aligned with the c-axis being perpendicular to the substrate [13]. Such nanowires have very high quality, so they are excelled material for research on exciton dynamics. When the wires were about 1 μm long the mode of growth was changed from N-rich to Ga-rich in order to promote lateral growth of GaN and to form a continuous GaN layer. Then the layer was grown to the thickness of about 1 μm. The structure of the sample is presented in Fig. 1. It is well known already that coalescence of NWs leads to formation of significant number of stacking fault (SF) at the joints [12]. Morphological characterization and cathodoluminescence (CL) measurements were performed using Hitachi SU-70 Scanning Electron Microscope (SEM) equipped with a Gatan MonoCL 3 system. The CL spectra were taken at temperature 5 K. The acceleration voltage of 5 kV and 2 nA beam current were applied. The CL spectra contained a few peaks. The strongest ones were related to donor-bound excitons (DX), I1 stacking faults and donor–acceptor (DA) recombination (see Fig. 1). We have found that the SF emission reaches the highest intensity at upper part of the nanowires where the highest strain is expected. We have observed mainly lines related to I1 SFs at 3.42 eV. The time-resolved photoluminescence (TRPL) was measured with the use of a 30 cm spectrograph and a Hamamatsu streak camera (2.5 ps resolution). The photoluminescence (PL) was excited by third harmonic of Ti:sapphire laser (300 nm). Temperature was controlled by helium continuos-flow cryostat at the range 4–300 K.

3. Results and discussion We have found that the GaN nanowires exhibit very strong photoluminescence, which is easily observed even at room temperature. This suggests low number of defects in the NWs. The

T=5K

CL Intensity [arb.u.], position [μm]

1.8

SF DX

DA,LO DA

1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 3.1

3.2

3.3

3.4

3.5

Energy [eV] Fig. 1. (A) Cross-section SEM image of the fully-overgrown GaN NWs sample and (B) its subsequent CL spectra at 5 K. Color dots on the image show approximate positions of CL measurement points of spectra (B).

K.P. Korona et al. / Journal of Luminescence 155 (2014) 293–297

I1SF

T=4K

where ℏω0 þ E is the exciton emission energy (including kinetic energy E), Tq the exciton temperature and A is the amplitude of the line. So, one can estimate Tq by the analysis of the shape of the excitonic peak. The observed shapes of the FX and SF peaks were successfully fitted with Eq. (1). The Tq at short time after excitation was the highest. Then excitons cooled down (see Fig. 2C). The observed cooling can be described by Newton's law:

1000

DX

100

10

PL Intensity [arb.u.]

FX

I2SF or I1SF,LO

T q ðtÞ ¼ T 0 exp ð t=t C Þ;

t = 0.1ns t = 1.65ns T=4K

SFI1SF

10

FX

DX

1

0.0

Temperature [K]

100

0.5

1.0

1.5

FX

100

SF I1SF 50

0

τC=0.37 ns

0.0

0.5

1.0 Time [ns]

1.5

Fig. 2. Time-resolved photoluminescence (PL) of the coalescent GaN/Si nanowires at T ¼ 4 K. (A) PL spectra vs. time, violet and black arrows show positions of the SF maximum at t¼ 0.1 ns and 1.65 ns, (B) PL transients and (C) cooling curves of FX and SF luminescence. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

PL Intensity (arb.u.)

0.1

ð2Þ

where τC is the cooling time. When fitted with a single exponential decay, the obtained cooling times were τC ¼0.1 70.05 ns and τC ¼ 0.37 70.05 ns for FX and SF, respectively. The free excitons cooling time is in agreement with previously published data [23,15,22]. It is important to notice that idea of high excitonic temperature assumes that they can move. Their high temperature represents their high kinetic energy. In the case of 2D object it should be Ek ¼kBT. The cooling time of particles localized in SF is longer than cooling of FX. This is probably due to weaker coupling of the 2D particles to phonons. This observation is in agreement with cooling time reported for GaInN QWs (τC ¼0.3 7 0.1 ns) [24]. The exciton phonon coupling can be estimated by comparison of strength of phonon replica and the zero phonon line. In this case we can observe a peak at 3.33 eV that can be a LO replica of the SF I1 line. This peak is 30 times weaker than the SF I1 line. So we expect that the coupling is weak. However the 3.33 eV peak can be I2 SF line. In this case the coupling would be even weaker. Temperature dependent measurements shown that heating leads to decrease of DX and FX energy parallel to shrinking of the band-gap (see Fig. 3A). The temperature behavior of the SF

1

100

295

SF

Temperature: 4K 10 25 K 35 K 53 K 75 K 1 100 K 125K 150K 0.1 200K 250K SF 300K

0.01 3.30

t = 1ns

DX FX

FX

3.35

3.40

3.45

3.50

have exponential slopes. The maximum of the SF peak moves from hν ¼3.421 eV at t¼0.1 ns down to 3.414 eV at 1.65 ns (see Fig. 2A). This behavior of FX and SF is in contrast to the dynamics of the DX peak, which keeps symmetrical shape and constant energy. The effect can be explained by assuming that hot excitons have additional kinetic energy, so emitted photons have higher energy, which causes broadening of the high-energy wing of the excitonic peak. This explanation is often claimed in the case of the FX [15,22,23]. In the case of the SF such behavior indicates that particles localized on the SF have kinetic energy and they can move freely in the SF plane. The characteristic exponential shape reflecting thermal distribution of excitons can be described by the following equation [15,22,24]: Iðℏω0 þEÞ ¼ AE1=2 exp



 E ; kB T q

ð1Þ

Decay times (ns)

Energy (eV)

data fit α T fit exp(E/kT) sum of fits

1.5

1.0

0.5 0

100

200

300

Temperature (K) Fig. 3. (A) GaN NWs spectra and (B) SF decay times at different temperatures.

296

K.P. Korona et al. / Journal of Luminescence 155 (2014) 293–297

energy was more complicated. The SF energy initially increased with temperature and started to decrease above 50 K. The intensity decreased with heating but the SF peak was visible even in the room temperature. The dependence of the SF decay times was not monotonic (see Fig. 3B). At temperatures below 50 K, the decay time increased with temperature. It reached maximum of 1.4 ns at about 60 K. Above 80 K, the decay times shortened with rise of temperature down to 0.5 ns at 300 K. Such dependence can be explained by taking into account two recombination processes: radiative (recombination rate rR), and nonradiative, rNR. Resulting recombination rate is reff ¼rR þrNR. Since photons have very low momentum, emission of light by moving particle is suppressed by momentum conservation rule. The dependence of an average momentum and a radiative lifetime for n-dimensional objects should be proportional to Tn/2 [25]. Assuming that the stacking faults are 2D objects, we propose τR ¼ αT þt0. The constant t0 reflects the fact that even for exciton with momentum equal to photon momentum, the recombination process would take some time. The nonradiative recombination rate (rNR ¼1/τNR) is usually thermally activated, so τNR ¼ β exp (EA/kBT). Since lifetime τeff ¼1/reff, we obtain the following equations for the lifetime: 1

τeff

¼

1

1

þ

ð3Þ

αT þ t 0 β exp ð  EA =kB TÞ

The fit of Eq. (3) to the experimental data (see Fig. 3B) gave parameters: α ¼ 0.015 ns/K, t0 ¼ 0.6 ns, β ¼ 0.25 ns and activation energy EA ¼ 22 meV. The good quality of the fit supports our

Center of mass energy (eV)

3.425

4. Conclusions

Temperature: 3.420

75 K 53 K 100 K 35 K 25 K 16 K

3.415

4K 125 K 3.410

0.0

0.5

1.0

1.5

2.0

Time (ns) 3.44

t = 0 ns 3.42

Energy (eV)

assumption about 2D nature of the particles localized at SFs. However, we must mention that the curve for n ¼3 (3-dimensional), τR ¼ αT3/2 þt0, gives also a good fit. As it was shown in Fig. 2 the SF line was asymmetrically broadened and the maximum of the peak was shifted with time to lower energies. The effect can be interpreted in terms of higher temperature of excitons, as given by Eq. (1). However, the simplest way to analyze this effect is to follow the evolution of the center of mass energy of the SF peak. Such dependencies for different temperatures have been plotted in Fig. 4A for temperatures 4– 125 K. At temperatures 150 K and above the energy did not change in time. In the low temperature range, we can see dramatic effects that are most probably due to cooling of excitons. At 4 K the difference between initial and final value is ΔEd ¼ 8 meV. This shift can be partially explained by changes of excitonic temperature with time but it may also be caused by flow of excitons to places with lower energy. These energy minima could be due to potential fluctuations along the SF plane. Since energy changes in time, we get different energy vs. temperature dependencies ESF(T) for different times of measurement. Fig. 4B present ESF(T) for t¼ 0 (just after laser pulse) and for t¼1 ns – after significant cooling of excitons. The ESF,0ns(T) exhibit monotonical dependence similar like band-gap dependence for GaN [26]. On the other hand, in the ESF,1ns(T) case the energy increases with increase of temperature up to 70 K, and then decreases. Difference between low temperature value and maximum ΔES ¼4.8 meV. Such an S-shaped is characteristic for quantum wells. This effect was already reported in the case of continuous excitation PL measurement of the SF [11].

Δ E = 4.8 meV

t = 1 ns 3.40 3.38 3.36 0

50

100

150

200

250

300

Optical characterization of the coalesced NWs reveals bright emission, sharp peaks (2 meV) and long lifetimes of bound excitons. The DX peak is at the same position, hν ¼ 3.471 eV, as in homoepitaxial GaN, what confirms good quality of the material. However, the coalescence of nanowires leads to formation of significant number of stacking faults. We show that the SF peak (3.42 eV at T¼ 4 K) exhibits some features characteristic for quantum wells. First, its energy vs. temperature dependence has S-shape with difference between low temperature value and the maximum of ΔES ¼ 4.8 meV. Second, the SF line is asymmetrically broadened and the maximum of the peak shifts with time to lower energies. Both these effects suggest flow of excitons to energy minima. Two other findings also show that excitons in SF can move like free particles: (i) their kinetic energy revealed by analysis of the shape of the SF luminescence and (ii) the decrease of radiative recombination rate following 1/T relation what suggest that excitons have momentum magnitude of which depends on temperature. The decay time of excitons confined in SF was relatively long 0.757 0.05 ns (that was longer than DX decay time, τ ¼ 0.2 ns) suggesting electron–hole separation. Dynamics of temperature of excitons was determined from the shape of the SF peak. The cooling time was τC ¼0.37 70.05 ns. This means that the excitons were cooled faster than they recombined. Finally, high intensity of luminescence observed even at room temperature promises possible applications of SF for light emitting devices.

Acknowledgments

Temperature (K) Fig. 4. (A) Center of mass transients at different temperatures. (B) Temperature dependence of the SF peak energy at t ¼0 ns (circles) and t¼ 1 ns (crosses).

This work was partially supported within European Regional Development Fund, through Grant Innovative Economy (POIG.

K.P. Korona et al. / Journal of Luminescence 155 (2014) 293–297

01.01.02-00-008/08 NanoBiom), by the JU ENIAC MERCURE Project no. 120220 and by the Polish National Science Centre funds granted with the decision DEC-2012/07/B/ST5/02484. The work of P.S. Perkowska was supported by the Foundation for Polish Science International Ph.D. Projects Programme co-financed by the EU European Regional Development Fund. References [1] Y.T. Rebane, Y.G. Shreter, M. Albrecht, Phys. Status Solidi A 164 (1997) 141. [2] C. Stampfl, C.G. Van de Walle, Phys. Rev. B 57 (1998) R15052. [3] Gilles Nogues, Thomas Auzelle, Martien Den Hertog, Bruno Gayral, Bruno Daudin, Appl. Phys. Lett. 104 (2014) 102102, http://dx.doi.org/10.1063/ 1.4868131. [4] L. Geelhaar, C. Chèze, B. Jenichen, O. Brandt, C. Pfüller, S. Münch, R. Rothemund, et al., IEEE J. Sel. Top. Quantum Electron. 17 (2011) 878, http: //dx.doi.org/10.1109/JSTQE.2010.2098396. [5] O. Brandt, C. Pfüller, C. Chèze, L. Geelhaar, H. Riechert, Phys. Rev. B 81 (2010) 045302. [6] Ayan Das, Pallab Bhattacharya, Animesh Banerjee, Marc Jankowski, Phys. Rev. B 85 (2012) 195321. [7] R. Liu, A. Bell, F.A. Ponce, C.Q. Chen, J.W. Yang, M.A. Khan, Appl. Phys. Lett. 86 (2005) 021908, http://dx.doi.org/10.1063/1.1852085. [8] S. Khromov, B. Monemar, V. Avrutin, H. Morkoc, L. Hultman, G. Pozina, Appl. Phys. Lett. 103 (2013) 192101, http://dx.doi.org/10.1063/1.4828820. [9] M. Suffczynski, R. Stepniewski, J.M. Baranowski, Acta Phys. Polonica 92 (1997) 993. [10] Pierre Corfdir, Pierre Lefebvre, J. Appl. Phys. 112 (2012) 053512, http://dx.doi. org/10.1063/1.4749789. [11] P.P. Paskov, R. Schifano, B. Monemar, T. Paskova, S. Figge, D. Hommel, J. Appl. Phys. 98 (2005) 093519. [12] V. Consonni, M. Knelangen, U. Jahn, A. Trampert, L. Geelhaar, H. Riechert, Appl. Phys. Lett. 95 (2009) 241910, http://dx.doi.org/10.1063/1.3275793.

297

[13] A. Wierzbicka, Z.R. Zytkiewicz, S. Kret, J. Borysiuk, P. Dluzewski, M. Sobanska, K. Klosek, A. Reszka, G. Tchutchulashvili, A. Cabaj, E. Lusakowska, Nanotechnology 24 (2013) 035703, http://dx.doi.org/10.1088/0957-4484/24/3/035703. [14] B. Monemar, P. Paskov, J. Bergman, G. Pozina, A. Toropov, T. Shubina, T. Malinauskas, A. Usui, Phys. Rev. B 82 (2010) 235202. [15] K.P. Korona, Phys. Rev. B 65 (2002) 235312. [16] R. Stępniewski, K.P. Korona, A. Wysmołek, J.M. Baranowski, K. Pakuła, M. Potemski, G. Martinez, I. Grzegory, S. Porowski, Phys. Rev. B 56 (1997) 15151. [17] D. Sam-Giao, R. Mata, G. Tourbot, J. Renard, A. Wysmolek, B. Daudin, B. Gayral, J. Appl. Phys. 113 (2013) 043102, http://dx.doi.org/10.1063/1.4775492. [18] J. Lähnemann, O. Brandt, U. Jahn, C. Pfüller, C. Roder, P. Dogan, F. Grosse, A. Belabbes, F. Bechstedt, A. Trampert, L. Geelhaar, Phys. Rev. B 86 (2012) 081302, http://dx.doi.org/10.1103/PhysRevB.86.081302. [19] K.P. Korona, Z.R. Zytkiewicz, P. Perkowska, J. Borysiuk, J. Binder, M. Sobanska, K. Klosek, Acta Phys. Polonica A 122 (2012) 1001. [20] Christian Hauswald, Timur Flissikowski, Tobias Gotschke, Raffaella Calarco, Lutz Geelhaar, Holger T. Grahn, Oliver Brandt, Phys. Rev. B 88 (2013) 075312, http://dx.doi.org/10.1103/PhysRevB.88.075312. [21] A. Gorgis, T. Flissikowski, O. Brandt, C. Chèze, L. Geelhaar, H. Riechert, H. Grahn., Phys. Rev. B 86 (2012) 041302(R). [22] D. Hägele, R. Zimmermann, M. Oestreich, M. Hofmann, W. Rühle, B. Meyer, H. Amano, I. Akasaki, Phys. Rev. B 59 (1999) R7797. [23] K.P. Korona, J. Kuhl, J.M. Baranowski, S. Porowski, Phys. Status Solidi B 216 (1999) 85. [24] J. Binder, K.P. Korona, A. Wysmołek, M. Kamińska, K. Köhler, L. Kirste, O. Ambacher, M. Zając, R. Dwiliński, J. Appl. Phys. 114 (2014) 223504, http: //dx.doi.org/10.1063/1.4845715. [25] D. Rosales, T. Bretagnon, B. Gil, A. Kahouli, J. Brault, B. Damilano, J. Massies, M. V. Durnev, A.V. Kavokin., Phys. Rev. B 88 (2013) 125437, http://dx.doi.org/ 10.1103/PhysRevB.88.125437. [26] K.P. Korona, A. Wysmołek, K. Pakuła, R. Stępniewski, J.M. Baranowski, I. Grzegory, B. Łucznik, M. Wróblewski, S. Porowski, Appl. Phys. Lett. 69 (1996) 788.