PERGAMON

Solid State Communications 121 (2002) 407±410

www.elsevier.com/locate/ssc

Excitons and multi-excitons in single CdTe quantum dots probed by near-®eld spectroscopy M. Brun, S. Huant 1,*, J.C. Woehl, J.-F. Motte, L. Marsal, H. Mariette Laboratoire de SpectromeÂtrie Physique, Universite Joseph Fourier de Grenoble et CNRS, P.O. Box 87, 38402 Saint Martin d'HeÁres, France Received 1 January 2002; accepted 1 January 2002 by B. Jusserand

Abstract A near-®eld optical spectroscopy study of a single CdTe/ZnTe quantum dot at low temperatures is presented. While the photoluminescence spectrum at low excitation power reveals only one single sharp peak due to the radiative recombination of excitons (X) in the single dot, several additional sharp peaks appear with increasing excitation density. The dominant features are ascribed to exciton complexes and charged exciton complexes such as negatively charged excitons (X 2), neutral (2X and 3X) and negative (2X 2 and 3X 2) biexcitons and triexcitons. Exciton charging arises due to ef®cient hole trapping by residual acceptors in the barrier material. This partly inhibits the formation of biexcitons and triexcitons. A spectral feature appearing close to the X 2 peak is tentatively assigned to X 22 negative excitons. This feature is found to shift to the red with increasing power: two possible explanations for this unexpected behaviour are proposed. q 2002 Elsevier Science Ltd. All rights reserved. PACS: 07.79.Fc; 78.55.Et; 78.66. 2 w Keywords: A. Nanostructures; D. Optical properties; E. Luminescence

Single quantum dot (SQD) optical spectroscopy has received considerable attention recently because the atomlike properties of the dot are best revealed on a single object. The great majority of the optical studies of SQDs have been made with the use of micro-photoluminescence (m-PL). The high spatial resolution required to isolate one object has often been reached by diffracting light through subwavelength holes in a metallic screen deposited on the sample surface [1,2] or by studying small mesa structures fabricated by lithography techniques [3]. Here we apply an alternative approach, namely, near-®eld scanning optical microscopy (spectroscopy)ÐNSOMÐto resolve CdTe SQDs spatially and to study their emission properties. NSOMÐthough dif®cult to implement at low temperature …T ˆ 4:2 K†Ðoffers some advantages over m-PL. Among others, it has the potential for better spatial resolution (see Ref. [4] for the best resolution that confocal microscopy can offer), it can be applied to any surface without further strain-inducing process, it is able to image any * Corresponding author. Fax: 133-476-51-45-44. E-mail address: [email protected] (S. Huant). 1 http://nsom.online.fr.

part of the sample by scanning the optical tip, and the excitation can be launched wherever desired with a non diffraction-limited resolution. Here we take advantage of some of these peculiarities of NSOM to reveal new features in the emission properties of a CdTe SQD. We used a sample grown in a molecular-beam-epitaxy (MBE) chamber on ZnTe substrates. The growth sequence includes a 400 nm thick ZnTe buffer layer, a thin CdTe active layer grown by atomic layer epitaxy (ALE) [5] and a 60 nm thick ZnTe cap layer. The thickness of the CdTe layer is 6.5 monolayers (ML) (2.1 nm). The large lattice mismatch of 6.2% between CdTe and ZnTe is similar to other systems such as InAs/GaAs [6] in which the formation of self-assembled QDs has been demonstrated. Extensive m-PL measurements and structural studies of ALE-grown CdTe/ZnTe structures con®rm the formation of Cd-rich dot-like islands which con®ne carriers to zero dimension [7,8] in samples similar to ours, in particular for a 6.5 ML CdTe thickness. The CdTe/ZnTe system has two interesting properties with respect to the InAs/GaAs system, which makes it a good candidate for NSOM studies. First, there is so far no trace of a wetting layer whatsoever in PL measurements. The absence of a wetting layer should

0038-1098/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved. PII: S 0038-109 8(02)00027-3

408

M. Brun et al. / Solid State Communications 121 (2002) 407±410

minimise inter-dot carrier diffusion [9], and therefore, help in maintaining the initial high spatial resolution of NSOM. Second, the valence band-offset is extremely small in such a way that the hole ground state is not the quantum-con®ned state in the dot, but rather the acceptor state in the barriers. As a consequence, we expect part of the photo-excited holes to be trapped by these acceptors to give rise to a steady-state population of excess electrons in the dots. This should favour the formation of negative excitons [10]. Our low-temperature NSOM has been described previously [11]. For the purpose of this communication, let us recall that we make use of a coated tapered optical tip for excitation. The optical tip used here has an aperture diameter of approximately 100 nm. The emission pro®le of this tip has been measured to prevent using tips with light leaks in the metal coating, which would deteriorate the spatial resolution [12,13]. The tip is brought into the optical near ®eld of the sample by means of cryogenic nano-motors and scanners. It is maintained in near ®eld by means of a tuning fork acting as a shear-force detector [14]. PL light is collected to far ®eld through an elliptic mirror and spectrally analysed by means of a 0.5 m spectrometer. Fig. 1 compares a typical far-®eld spectrum taken with a conventional PL set-up, i.e. without spatial resolution, with a near-®eld spectrum at 100 nm spatial resolution. The advantage of using NSOM is clearly evident. A broad band with a 40 nm full width at half maximum is observed in conventional experiments. This is typical for a large ensemble of dots having different sizes, compositions and/ or local environments, i.e. different emission properties. This inhomogenous broadening is suppressed with NSOM, which reveals a very limited number of sharp peaks that have line-widths limited to the spectral resolution, namely 0.1 nm in Fig. 1. This is because the high spatial resolution of NSOM allows for excitation of a few quantum dots only. Actually, a single quantum dot is seen with NSOM as con®rmed below. Fig. 2 shows PL spectra taken for increasing excitation powers with the optical tip placed above the same particular `point' of the sample as the one chosen in Fig. 1. At the lowest achievable power, only one peak is seen due to the radiative recombination of a single X exciton. However, an increasing number of additional PL peaks appear with increasing excitation density. Since we do not have the possibility of tuning the charge of the dot as done in Refs. [1,15], a de®nitive assignment of all of these peaks is a complicated task. However, the analysis of their power dependence gives converging evidence for the assignment that we proposed below. Peaks labelled 1 and 2 have linear and (almost) quadratic power dependences, respectively (peak 1 rapidly saturates beyond a power of 11.5P0). Their assignment is straightforward: they are due to radiative recombination of excitons (X) and biexcitons (2X) in the investigated SQD. Their energy separation (4.79 nm or 15.8 meV) is close to previously reported values [8,11] and agrees well with a

A

Fig. 1. A typical far-®eld spectrum of a CdTe QD sample compared with a near-®eld spectrum. The excitation wavelength is 514.5 nm. The temperature is 4.2 K.

CdTe/ZnTe QDs T=4.2 K, λexc=514.5 nm

30 P0 8

25 P0 2

1

7,8 5

4

3 6 23.5 P0

20 P0 1

2 4

7

8

3 6 18.2 P0

4,7

15 P0

5 1 2 4

5

3 6 11.5 P0

1

P0

x 10

608

610

612

614

616

618

620

622

Wavelength (nm) Fig. 2. Evolution of the PL spectra of a single quantum dot with excitation power. The low excitation power P0 corresponds to an injected power into the optical ®bre of 40 mW. The transmission of a 100 nm optical tip is typically 10 24. Other powers are expressed in units of P0. The different PL features are labelled as in the text. Please note that the low power spectrum has been magni®ed by a factor of 10. Additional spectra have been collected for various powers ranging from P0 to 11.5P0 (not shown).

M. Brun et al. / Solid State Communications 121 (2002) 407±410

large electron±hole exchange interaction in II±VI quantum dots, which leads to large binding energies for 2X [16]. In addition to that, a peak labelled 3 (actually, this peak is accompanied by a close-lying peak 6 whose possible origin is discussed below) is seen around 620 nm. It is already visible at modest excitation (curve 11.5P0). Although its power dependence is dif®cult to extract from Fig. 2, it seems to be superlinear. We propose that peak 3 is due to triexcitons (3X), which are expected to contribute in the s shell [1,3,15]. The ®nal states of a triexciton recombination in the s shell is an excited biexciton state formed by one s-exciton and one p-exciton, with electron spins being either parallelÐspin triplet or anti-parallelÐspin singlet. As shown by Bayer et al. [3], triexcitons recombine primarily in the triplet state. It is worth noting that all of these PL features are observed simultaneously in a single spectrum because of the large measurement time of a few minutes, which is required to gain a suf®cient signal-to-noise ratio. Therefore, a large number of radiative events in a SQD are averaged in our measurements. In addition to neutral excitons and multi-excitons, charged excitons [1] are also expected to show up in our measurements, although dots are not charged with resident electrons [1,15]. This is because photo-excited holes can be ef®ciently trapped by residual acceptor states in the barrier due to the very small valence-band offset in the CdTe/ZnTe system. This results in a steady-state population of photogenerated electrons in the dots that can negatively charge excitons [10]. We propose that peak labelled 4 in Fig. 2 is a signature of X 2 excitons. Its energy separation of 1.58 nm (5.20 meV) is indeed very close to the binding energy of an X 2 exciton in a modulation-doped CdTe QD [8]. This binding energy is larger than in CdTe quantum wells [10], in agreement with the intuitive expectation of enhanced electron correlation in low-dimensional systems. In fact, it can be seen from Fig. 2 (see, e.g. excitation power 11.5P0) that each neutral exciton is accompanied by a low-energy counterpart. For instance, the biexciton peak 2 at 617.85 nm is replicated by peak 5 at 618.46 nm (energy separation of 2.03 meV), which is already observed at modest excitation power. Similarly, the weak feature peak 3 (620.58 nm) appears with a counterpart at 620.81 nm (peak 6, energy separation 0.75 meV). It is natural to assign these weaker replica to the corresponding negatively charged multi-excitons, namely 2X 2 (peak 5) and 3X 2 (peak 6). The energy sequence is compatible with this assignment, namely 2X 2 is more weakly bound than X 2, in agreement with previous observations in InAs [15,17], 3X 2 being even less bound. Similarly, a double-negatively charged exciton X 22 can form by trapping of two photo-excited holes from a 3X triexciton. We believe that signatures of X 22 excitons are present in the spectra of Fig. 2. These negative excitons should contribute to two luminescence lines in the s shell because they have two possible ®nal states after radiative recombination, namely a spin-singlet and a spin-triplet

409

[1,15]. Since peaks 7 (at 614.77 nm) and 8 (at 616.46 nm) have higher than quadratic power dependences, more than two electron±hole pairs must be excited to create the corresponding species. In addition, peak 7 appears near X 2. Therefore, a tempting assignment is that it is due to the X 22 triplet, which has been observed very close to X 2 in InAs dots [15]. Then, peak 8 could be assigned to the X 22 singlet. This assignment is reinforced by the fact that peak 7 has larger integrated intensity than peak 8, in agreement with the expectation and previous observations for X 22 spin-triplet and spin-singlet states [1]. Following Ref. [1], con®rmation of this assignment would allow to infer 2.8 meV to the exchange energy in the investigated CdTe SQD. Peak 7 behaves in a very striking way with increasing power. First, it shifts to the red with increasing power and tends to converge to peak 8. To the best of our knowledge, such a behaviour has not been reported previously, neither in InAs, nor in CdTe. Upon merging with peak 8 at a power of 25P0, the latter blows up and splits into several overlapping bands and peaks. This complex behaviour is observed over a narrow power range. To explain this behaviour, novel selection rules in the near ®eld can hardly be invoked at our resolution of 100 nm [18,19], which is much larger than the exciton Bohr radius in CdTe. One possibility is that this band is due to a light-induced Stark shift of negative excitons. Related observations have been reported previously for single molecules [20]. Another intriguing posibility is that the spin splitting of these many-electron states is suppressed on increasing the excitation power or, in other words, that the electron spin is screened out [21,22]. This somehow recalls to us a related problem encountered with D 2 negative-donors in quantum wells in the presence of a large electron surplus [23] (a D 2 centre is analogous to X 2 with an in®nite hole mass [24]). Since the present experimental results do not allow for a de®nitive assignment of all of the salient spectral features in Fig. 2, it is not possible to speculate further on the origin and behaviour of peaks 7 and 8. For completeness, it is worth mentioning that peaks 7 and 8 are not observed in all of the SQDs that we were able to isolate under the optical tip. It seems that the close neighbouring of the SQD plays a key role in their occurrence. This makes sense since an electron excess i.e. in our case the vicinity of acceptor(s) to the SQD, is needed anyway to populate negative-exciton states. Finally, we wish to comment on one additional observation. At the highest excitations in Fig. 2, most of the peaks broaden or split into several components. This is a consequence of the unstable electric environment of the dot under high excitation, which causes constant charge exchanges between dot excitons and residual impurities in their neighbourhood or interactions between dot carriers and surface charges. These ¯uctuating potentials induce ¯uctuating PL-peak energies [1,8], which are mostly averaged in our long-time measurements. At lower excitation, these ¯uctuating ®elds only cause small peak shifts or broadening,

410

M. Brun et al. / Solid State Communications 121 (2002) 407±410

of the order of 0.1 nm, i.e. close to the spectral resolution limit. In the above scenario, the assumption has been made that only one SQD is excited throughout the entire power range. This is certainly the case at low excitation since only one peakÐpeak 1Ðis seen at the lowest power. However, this could be questioned at high excitation, where photo-excited carriers could diffuse to adjacent dots. We believe that this assumption is reasonable in view of the fact that no signature of a wetting layer is seen in PL measurements, in contrast with Stranski±Krastanov dots [1,2] for which inter-dot diffusion can take place via the wetting layer. Therefore, we expect inter-dot carrier diffusion [9] to be drastically reduced in our Cd-rich islands. This helps to maintain the initial high spatial resolution of NSOM and to reveal the reach variety of single-dot emission phenomena described in this paper. Acknowledgements The authors wish to acknowledge stimulating discussions with K. Karrai. References [1] R.J. Warburton, C. Schaȯein, D. Haft, F. Bickel, A. Lorke, K. Karrai, J.M. Garcia, W. Schoenfeld, P.M. Petroff, Nature 405 (2000) 926. [2] A. Zrenner, J. Chem. Phys. 112 (2000) 7790. [3] M. Bayer, O. Stern, P. Hawrylak, S. Fafard, A. Forchel, Nature 405 (2000) 923. [4] E. Dekel, D. Gershoni, E. Ehrenfreund, D. Spektor, J.M. Garcia, P.M. Petroff, Phys. Rev. Lett. 80 (1998) 4991. [5] J.M. Hartmann, G. Feuillet, M. Charleux, H. Mariette, J. Appl. Phys. 79 (1996) 3035.

[6] J.-Y. Marzin, J.-M. GeÂrard, A. IzraeÈl, D. Barrier, G. Bastard, Phys. Rev. Lett. 73 (1994) 716. [7] L. Marsal, PhD Thesis, University of Grenoble, March 2001, unpublished. [8] L. Besombes, K. Kheng, L. Marsal, H. Mariette, 2002, in press. [9] C. ObermuÈller, A. Diesenrieder, G. Abstreiter, K. Karrai, S. Grosse, S. Manus, J. Feldmann, H. Lipsanen, M. Sopanen, J. Ahopelto, Appl. Phys. Lett. 74 (1999) 3200. [10] K. Kheng, R.T. Cox, Y. Merle d'AubigneÂ, F. Bassani, K. Saminadayar, S. Tatarenko, Phys. Rev. Lett. 71 (1993) 1752. [11] M. Brun, S. Huant, J.C. Woehl, J.-F. Motte, L. Marsal, H. Mariette, J. Microsc. 202 (2001) 202. [12] C. ObermuÈller, K. Karrai, Appl. Phys. Lett. 67 (1995) 3408. [13] A. Drezet, J.C. Woehl, S. Huant, Europhys. Lett. 54 (2001) 736. [14] K. Karrai, R.D. Grober, Appl. Phys. Lett. 66 (1995) 1842. [15] J.J. Finley, P.W. Fry, A.D. Ashmore, A. LemaõÃtre, A.I. Tartakovskii, R. Oulton, D.J. Mowbray, M.S. Skolnick, M. Hopkinson, P.D. Buckle, P.A. Maksym, Phys. Rev. B 63 (2001) 161305(R). [16] V.D. Kulakovskii, G. Bacher, R. Weigand, T. KuÈmmell, A. Forchel, E. Borovitskaya, K. Leonardi, D. Hommel, Phys. Rev. Lett. 82 (1999) 1780. [17] R.M. Thompson, R.M. Stevenson, A.J. Shields, I. Farrer, C.J. Lobo, D.A. Ritchie, M.L. Leadbeater, M. Pepper, Phys. Rev. B 64 (2001) 201302(R). [18] O. Mauritz, G. Goldoni, F. Rossi, E. Molinari, Phys. Rev. Lett. 82 (1999) 847. [19] C.D. Simserides, U. Hohenester, G. Goldoni, E. Molinari, Phys. Rev. B 62 (2000) 13657. [20] P. Tamarat, F. Jelezko, C. Brunel, A. Maali, B. Lounis, M. Orrit, Chem. Phys. 245 (1999) 121. [21] W. Izumida, O. Sakai, S. Tarucha, Phys. Rev. Lett. 87 (2001) 216803. [22] K. Karrai, private communication. [23] Z.X. Jiang, B.D. McCombe, P. Hawrylak, Phys. Rev. Lett. 81 (1998) 3499. [24] S. Huant, S.P. Najda, B. Etienne, Phys. Rev. Lett. 65 (1990) 1486.