Photoluminescence study of Si/Ge quantum dots

Surface Science 532–535 (2003) 832–836 www.elsevier.com/locate/susc Photoluminescence study of Si/Ge quantum dots M. Larsson *, A. Elfving, P.O. Holtz, G.V. Hansson, W.-X. Ni Department of Physics, Link€oping University, SE-581 83 Link€oping, Sweden Abstract Ge quantum dots embedded in Si are studied by means of photoluminescence (PL). The temperature dependent PL measurements show two different types of recombination processes related to the quantum dots. We ascribe a peak near 0.80 eV to the spatially indirect recombination in the type-II band lineup where the electron is located in the surrounding Si close to the interface and the hole in the Ge dot. Furthermore, a peak near 0.85 eV is attributed to the spatially direct recombination. We observe a transition from the spatially indirect to the spatially direct recombination as the temperature is increased. The measurements also show an up-shift of the Ge quantum dot emission energy with increasing excitation power density. The blueshift is primarily ascribed to an enhanced confinement of the electron associated with the increased band bending at the type-II Si/Ge interface at high carrier densities. Comparison is made with results, derived from measurements on uncapped samples. For these uncapped samples, no energy shifts due to excitation power or temperatures are observed in contrast to the capped samples. Ó 2003 Elsevier Science B.V. All rights reserved. Keywords: Photoluminescence; Self-assembly; Silicon–germanium 1. Introduction Strain induced self-assembled Ge quantum dots embedded in Si have attracted a large interest during the last years due to the possibility to realize Si based optoelectronics. For example, detector applications using Si/Ge quantum dot structures in the active region have been suggested. Ge dots will in this case serve as the active material due to the smaller band gap. In spite of the fact that Si/Ge interdiffusion will lead to some alloying of the dots we will refer to them as Ge dots. Epitaxial growth of the lattice mismatched Si/Ge material system has * Corresponding author. Tel.: +46-13-28-27-56; fax: +46-1328-89-69. E-mail address: [email protected] (M. Larsson). demonstrated Stranski–Krastanov formation of islands under certain conditions. This gives a simple and rather straightforward way to create Ge quantum dots, compatible with the Si technology. Unfortunately, both Si and Ge are indirect band gap materials and phonons are normally required for momentum conservation in optical transitions. On the contrary, in SiGe alloys and Si/Ge quantum structures, the symmetry of the lattice is broken, which opens the possibility for optical transitions without any phonon interaction. In quantum structures, the spatial confinement of the carriers will spread the wave functions in k-space and thus increase the probability for a direct no-phonon optical transition [1,2]. In this work, the optical properties of Ge quantum dots were investigated by means of photoluminescence (PL) with temperature and excitation power density as variable parameters. We discuss the involved recombination 0039-6028/03/$ - see front matter Ó 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0039-6028(03)00461-8 M. Larsson et al. / Surface Science 532–535 (2003) 832–836 processes and compare the results with measurements on uncapped samples. (a) Si 833 Ge Si CB B 2. Experimental The two samples studied here were single layer structures grown by solid-source molecular beam epitaxy (MBE). At a growth temperature of 700 °C, eight monolayers Ge were deposited on Si(1 0 0) substrates. The Ge quantum dots were formed from this Ge layer via Stranski–Krastanov growth mode. One of the samples was covered with Si, forming a 160 nm capping layer. In both samples, the average dot diameter was about 200 nm and the typical height was 20–25 nm as determined by atomic force microscopy (AFM) studies. PL measurements were performed in a variable temperature He-flow cryostat, and as excitation source the 514 nm line of an Ar ion laser was used. The PL signals were analyzed with a double-grating monochromator, together with a liquid nitrogen cooled Ge detector, using standard lock-in technique. 3. Results and discussion 3.1. Band alignment Since the growth mode is strain induced and the dot formation is a result of elastic relaxation, the sandwiching Si above and below the islands exhibits tensional strain [3]. It is known that in tensile strained Si, the D(2) valleys of the conduction band is downshifted, which results in a type-II band alignment at the interface between the Ge dots and the surrounding Si (see Fig. 1) [3,4]. During the formation of the dot, the Ge rich wetting layer surrounds the base of the dot and prevents it to expand laterally, while the upper part of the Ge dot is expected to be more relaxed than the base. This will cause an asymmetric strain through the structure. Consequently, the Si layer above the Ge dot will exhibit a higher strain than in the Si material below the dot. As a result, the band offset in the conduction band at the top interface will be larger than at the bottom interface (Fig. 1). In Fig. 1a, the two A C VB (b) CB VB Low carrier density Growth direction High carrier density Fig. 1. (a) A schematic illustration of the band edge alignment along the growth direction of a Ge quantum dot. Spatially indirect transitions are marked A and C, while the spatially direct transition is marked B. (b) An illustration of the band bending effects in the structure at two different carrier concentrations. transitions observed in the PL measurements are labeled A and B for the spatially indirect and direct transition, respectively. The spatially indirect transition is expected to be rather weak due to the spatial separation of the carriers. The probability for this indirect transition is related to the magnitude of the overlap of the wave functions leaking into the potential barriers. For the spatially direct transition, when both the electrons and the holes are located in the quantum dot, the alloy disorder and quantum confinement effects may relax the kconservation condition enough to increase the probability for optical recombination. The third possible transition (labeled C in Fig. 1a) is close in energy to the B transition and can not be separated from the spatially direct transition in PL measurements on capped structures. 3.2. Excitation power dependence The PL dependence of the excitation power density was examined in order to investigate the mechanisms of the carrier recombination. Fig. 2a and b show the PL spectra for varied excitation 834 M. Larsson et al. / Surface Science 532–535 (2003) 832–836 T=10K T=30K 150mW 100mW 50mW 100mW 10mW 50mW 3mW 10mW 1mW 3mW 0.7 0.8 0.9 1.0 1.1 1.2 (a) 0.7 0.8 0.9 1.0 1.1 1.2 (b) Fig. 2. Normalized PL spectra of Ge dots. The excitation power dependence is shown for two different temperatures (a) 10 K and (b) 30 K. The quantum dot related emission is below 0.95 eV, while the emissions above this energy are contributions from the substrate and the wetting layer. power at two different temperatures, 10 and 30 K, respectively. Typical emissions from the Si substrate and the thin Ge wetting layer are observed above 0.95 eV, while the quantum dot related emission occurs below 0.95 eV [2,5–7]. At low temperature (10 K), a significant blueshift of 25 meV of the Ge dot related emission is observed as the excitation power is increased from 3 to 100 mW. Other groups have earlier shown a similar excitation power dependence of the Ge quantum dot emission [2,5,6]. The shift can be explained in terms of a type-II band lineup, where the electrons are located in the Si, while the holes are trapped in the Ge dot. The Coulomb interaction between these spatially separated electrons and holes will bend the energy bands at the interface to form a Hartree potential on each side of the interface [5,8]. At high carrier concentrations, the band bending will shift the electron and hole levels to higher energies due to an increased confinement, resulting in higher transition energies (Fig. 2a). State filling could also cause an effective blueshift of the PL peak at sufficiently high carrier densities [6]. When the temperature is increased to 30 K, the emission spectra look quite different for all excitation powers (Fig. 2b). At low excitation power, the dot luminescence is divided into two branches, centered at approximately 0.80 and 0.85 eV, respectively. When the excitation power is increased from 1 to 150 mW, the intensity of the high energy peak increases faster than the low energy peak, to totally dominate the spectrum at high excitation power. It should be noted that the energy position of this peak (at 30 K) is always higher than the blueshifted low energy peak observed at high excitation power at 10 K, strongly indicating that these two contributions have different origins. It should also be pointed out, that the high energy peak does not shift with increasing excitation power, implying that band bending is not affecting this transition. 3.3. Temperature dependence Fig. 3a and b show the temperature dependence of the luminescence at two different excitation M. Larsson et al. / Surface Science 532–535 (2003) 832–836 P=10mW 835 P=50mW 5K 8K 6K 11K 10K 15K 15K 20K 20K 30K 25K 40K 50K 30K 75K 35K 100K 42K 0.7 0.8 0.9 1.0 1.1 1.2 (a) 150K 0.7 0.8 0.9 1.0 1.1 1.2 (b) Fig. 3. Temperature dependence of the PL spectra from the Ge dots at two different excitation power densities, (a) 10 mW and (b) 50 mW, respectively. The spectra shown in (b) are normalized with respect to the quantum dot emission. powers. At low excitation power densities and low temperatures, the low energy peak at 0.80 eV dominates the spectrum (Fig. 3a). As the temperature is raised, a redistribution of the emission intensities from the low energy peak to the high energy peak at 0.85 eV is observed. When the excitation power is increased to 50 mW (Fig. 3b), the temperature behavior is quite similar as for the lower excitation power, but with one notable difference: The low energy peak is up-shifted due to band bending, as described in the previous section. This behavior is consistent with the high energy peak as being due to the spatially direct recombination, while the low energy peak at 0.80 eV is related to indirect transitions across the dot interface. At low temperatures, the only possible recombination channel is the spatially indirect transition across the interface for which the electron is located in the Si surrounding the Ge dot, while the hole is confined inside the Ge dot. A higher temperature results in an increased probability for the electrons to populate the higher energy level inside the dot, which opens the possibility for an alternative recombination channel in addition to the low energy emission at 0.80 eV. In our case, the probability for the spatially direct transition at 0.85 eV increases. For this transition, the overlap of the wave functions is considerably larger than for the spatially indirect transition, resulting in a more efficient luminescence. 3.4. Uncapped structures The PL from uncapped dots is detectable, but the intensity is significantly reduced due to the strong non-radiative surface recombination. Fig. 4 shows that the shape of the spectra of uncapped dots differ significantly from those measured on the capped dots. The position of the quantum dot related emission band is down shifted to a peak energy at 0.77 eV. It is established that the capping of self-assembled Ge islands will result in increased intermixing of Si and Ge [7]. The intermixing will accordingly up-shift the band gap of the dot. As a result, the emission of the capped dots is expected 836 M. Larsson et al. / Surface Science 532–535 (2003) 832–836 direct (B in Fig. 1a) and indirect transition in uncapped dots (C in Fig. 1a) due to the small band offset described above. 4. Conclusion Fig. 4. Normalized PL spectra from an uncapped dot structure at 30 K with a varied excitation power. to occur at higher energies than in uncapped structures, which is in consistence with our experimental results. Furthermore, no shift of the transition energy is observed as the excitation power is increased. This behavior is expected, since the small band offset at the bottom of the dot is not sufficient to capture enough electrons to give rise to any significant band bending. As mentioned above, the intensity of the emission is lower than that for the capped samples, due to competing recombination via non-radiative surface states. This fact implies that the surface recombination will ensure low carrier densities and consequently no detectable energy shifts due to band bending. At increased temperatures, no second peak at higher energy is observed in contrast to the capped dot sample, in consistence with our band lineup picture. The energy position for the emission should be practically the same for the spatially The present PL study of Ge quantum dots embedded in Si shows two different dot related transitions. Based on the experimental results, we suggest a type-II energy band lineup that gives possibility for one spatially indirect transition, which is blueshifted with increasing excitation power together with one spatially direct transition inside the dots. Temperature dependent measurements show that the direct transition is a more efficient recombination channel, as expected. Measurements on uncapped structures show only one quantum dot related emission band, which has excitation power and temperature dependencies that are consistent with our model proposed on the recombination processes. References [1] K. Eberl, O.G. Schmidt, R. Duschl, O. Kienzle, E. Ernst, Y. Rau, Thin Solid Films 369 (2000) 33. [2] G. Bremond, M. Serpentini, A. Souifi, G. Guillot, B. Jacquier, M. Abdallah, I. Berbezier, B. Joyce, Microelectron. J. 30 (1999) 357. [3] O.G. Schmidt, K. Eberl, Phys. Rev. B 62 (2000) 16715. [4] C.G. Van de Walle, R.M. Martin, Phys. Rev. B 34 (1986) 5621. [5] J. Wan, G.L. Jin, Z.M. Jiang, Y.H. Luo, J.L. Liu, K.L. Wang, Appl. Phys. Lett. 78 (2001) 1763. [6] P. Boucaud, V. Le Thanh, V. Yam, S. Sauvage, N. Meneceur, M. Elkurdi, D. Debarre, D. Bouchier, Mat. Sci. Eng. B 89 (2002) 36. [7] O.G. Schmidt, K. Eberl, Phys. Rev. B 61 (2000) 13721. [8] T. Baier, U. Mantz, K. Thonke, R. Sauer, F. Sch€affler, H.-J. Herzog, Phys. Rev. B 50 (1994) 15191.