Journal of Microscopy, Vol. 202, Pt 1, April 2001, pp. 202±208. Received 28 August 2000; accepted 11 October 2000

Low-temperature near-field spectroscopy of CdTe quantum dots M. BRUN*, S. HUANT*, J. C. WOEHL*, J.-F. MOTTE*, L. MARSAL² & H. MARIETTE² *Laboratoire de SpectromeÂtrie Physique (LSP), CNRS UMR 5588, Universite Joseph Fourier Grenoble I, BP 87, 38402 Saint Martin d'HeÁres, France ²CEA-CNRS Joint Group `Nanophysique et Semiconducteurs', DeÂpartement de Recherche Fondamentale sur la MatieÁre CondenseÂe, SPMM, 17 Rue des Martyrs, 38054 Grenoble Cedex 9, France

Key words. II-VI semiconductors, near-field optics, photoluminescence, quantum dots, spectroscopy.

Summary A near-field optical microscope has been developed for operation at low temperature. This microscope is used to study the photoluminescence of CdTe-based quantum dots. Spectra collected upon approaching the optical tip into the near-field region of the sample reveal the evolution from a broad far-field luminescence band 2 that is typical for a large ensemble of dots 2 to a near-field structure made up of a few sharp peaks originating from individual dots. Experiments carried out in the excitation-collection mode through the optical tip allow study of the effect of an increase in excitation power on the near-field spectra. It is found that upon increasing the excitation by two orders of magnitude, a spatially resolved spectrum progressively transforms back into a broad `far-field-like' spectrum. Photoluminescence images taken by scanning the sample under the tip are used to discriminate various contributions coming from individual dots.

Introduction In a pioneering work, Hess et al. (1994) have demonstrated the ability of near-field optical spectroscopy to resolve the various quantum constituents of a luminescent semiconductor structure. Consequently, novel effects of fundamental interest can be optically investigated in semiconductors on a length scale which is not diffraction limited (Eytan et al., 1998). Self-assembled quantum dots (QDs) are of special current interest in the semiconductor community both for fundamental research and potential applications, e.g. in laser diodes and optical memories (LundstroÈm et al., 1999). Such dots are obtained by growing a thin layer of a Correspondence to: S. Huant. Tel: 1 33 476 51 47 24; fax: 1 33 476 51 45 44; e-mail: [email protected]

202

semiconductor material, e.g. InAs or CdTe, on another related material with a large lattice mismatch, e.g. GaAs or ZnTe, respectively. Beyond a certain layer thickness and depending on the growth procedure and materials, the large elastic energy accumulated in the thin strained layer relaxes by forming three-dimensional islands that are typically 10± 20 nm in lateral size and 6±8 nm high. In such QDs, the electronic motion is fully quantized which gives them an atomic-like energy spectrum. As a matter of fact, QDs are often considered as prototypical `artificial atoms' (Bayer et al., 2000; Warburton et al., 2000). The remarkable homogeneity in size and composition of self-assembled QDs over macroscopic scales has sparked numerous transport (Drexler et al., 1994) and optical measurements (Warburton et al., 1997) which, although averaged over millions of dots, do not mask completely the properties of the individual artificial atoms. However, the most spectacular phenomena have been revealed in optical measurements carried out on a limited number of dots, possibly on individual dots. For instance, the extremely abrupt changes in the emission energy of the artificial atom induced by adding electrons one-by-one into it 2 a typical atom-like property revealing the shell structure of the QD 2 have been demonstrated on a single dot (Warburton et al., 2000). Two main different instrumental approaches to reach the required high spatial resolution can be found in the literature. Either apertures are produced in an opaque metallic mask processed on the sample, or an aperture nearfield scanning optical microscope (NSOM) can be employed. In the first case, resolution as high as 100 nm can be obtained with a sizable light throughput (see, e.g. Kulakovskii et al., 1999). This proves sufficient to gain information on individual dots provided that the density is not too high (typically 1010 cm22) and/or individuals can be discriminated from each other through their spectroscopic q 2001 The Royal Microscopical Society

N EA R - FI E L D SP EC T RO S C O P Y O F Q UA N T U M D O T S

or other properties. However, such a possibility does not always exist in actual systems or the dot density may be too high. Hence, there is need and place for experiments at higher spatial resolutions. In principle, this can be accomplished with a NSOM since aperture optical tips can be produced with apex diameters below 100 nm. Current developments open a promising route towards hightransmission optical tips which could be used for both exciting and collecting the photoluminescence (PL) through a 50-nm aperture or even smaller. In addition to this potentially better resolution, NSOM offers the unique opportunities to study unprocessed samples, to scan the sample under the tip ± and therefore to probe various regions of the sample at will ± and to image simultaneously the topography and the emission properties of the sample. In order to give a flavour of what kind of realizations can be achieved and/or prospected in QDs with a lowtemperature NSOM, we describe here a set of preliminary spectroscopic studies of CdTe/ZnTe QDs at moderate spatial resolution (200 nm). We will focus more on the instrumental demonstration of such studies than on the physics of QDs which requires further measurements in progress. For an updated review on single dot spectroscopy, we refer the reader to the paper by Zrenner (2000). Our home-made low-temperature NSOM is developed along the lines of the instrument briefly described in a paper by the Munich group of Karrai (ObermuÈller et al., 1999). Our instrument is sketched in Fig. 1. Basically, a coated optical tip coupled either to a He-Ne laser or an Ar1 laser is brought in a focal point of an ellipsoidal mirror, while a 200 mm multimode fibre placed in the second focal point collects light emitted by the sample. The tip is glued to one arm of a piezoelectric tuning fork (Karrai & Grober, 1995)

Fig. 1. A scheme of the microscope head. q 2001 The Royal Microscopical Society, Journal of Microscopy, 202, 202±208

203

as is commonly done today to control the tip position in the near-field region of the sample. At helium temperature, the Q-factor of the tip-tuning fork ensemble is routinely in excess of 5000 with peak values approaching 10 000. The use of a tuning fork for feedback permits measurement simultaneously of the surface topography and the local optical properties. The measurements reported here were systematically taken on topography-free regions to minimize possible artefact effects. The tapered optical tips are either pulled under laser heating (Betzig et al., 1991) using a commercial pipette puller (Sutter P2000) or chemically etched in a HF acid solution. Because of their low transmission, pulled tips can only be used in the illumination mode to excite the sample locally. With such fibre tips, collection is made through the elliptic mirror (Fig. 1). In contrast, chemically etched tips can be used in the dual illumination-collection mode because they have a larger transmission (the fibre core is preserved down to the tip apex). Pulled and etched tips are subsequently coated with a 100-nm aluminium layer. To avoid the use of tips where light leaks through the metallic coating (which would deteriorate the spatial resolution), a selection of `good tips' without such leakages and the determination of their actual apex diameter is made using the far-field optical characterization procedure described by ObermuÈller & Karrai (1995). Special care is taken not to damage the tip during approach or scanning which is a real challenge at low temperature. This is accomplished in particular by using shear forces applied to the tip well below 2  1029 N since larger forces may damage the tip aperture (Karrai & Grober, 1995). In addition, the light intensity collected by the elliptic mirror is systematically measured at the end of experiments to compare with the initial value and to check against possible aperture damages. Coarse as well as fine xyz-positioning of the sample is achieved in situ by using cryogenic piezoelectric inertial motors and piezostack-based scanners. The latter offer a limited xy-scan range of 2.1 mm  2.1 mm at liquid helium temperature. The entire microscope head is made of nonmagnetic materials for future experiments in a high magnetic field and is designed to fit in a 2-inch i.d. stainless steel tube which is immersed in liquid helium. This way the microscope is operated in a low-pressure He exchange gas. While the microscope head and the optical tips are homemade, all the other parts of the set-up are commercial. These include a Topometrix Electronic Control Unit ECU-1 to control the microscope head and a 0.5-m focal length spectrometer (Acton SpectraPro 500i) equipped with a nitrogen-cooled charge-coupled device (CCD) camera (Princeton Instruments) to disperse the PL signal. The spectral resolution is set to 0.07 nm at a detection wavelength of 600 nm (approximately 0.25 meV) and the integration time of the CCD camera is chosen between 5 s for PL

204

M . B RU N E T A L .

imaging and 90 s for some low-noise local reference spectra. The spectra presented here have been excited with the lecx ˆ 514.5nm line of the Ar1 laser. We used two samples grown in a molecular-beam-epitaxy (MBE) chamber on ZnTe substrates. The growth sequence includes a ZnTe buffer layer (thickness of 400 nm and 800 nm in sample 1 and 2, respectively), a thin CdTe active layer grown by atomic layer epitaxy (ALE) and a 60-nm thick ZnTe cap layer. In comparison with more conventional MBE growth, ALE consists of sending cations (here Cd) and anions (Te) alternatively onto the surface (Hartmann et al., 1996). This enhances the mobility of growing species on the surface while simultaneously keeping the growth temperature low enough (here 280 8C) to prevent undesirable interdiffusion at the interfaces. The thickness of the CdTe layer is 6.5 monolayers (2.1 nm) and 5 monolayers in sample 1 and 2, respectively. The large lattice mismatch of 6.2% between CdTe and ZnTe is similar to other systems such as InAs/GaAs (Marzin et al., 1994) in which the formation of self-assembled QDs has been demonstrated. Strictly speaking, evidence for a transition from a twodimensional to a three-dimensional growth mode in the present CdTe/ZnTe system, i.e. for the formation of selfassembled QDs, is not as strong as in the well-established InAs/GaAs tandem. However, there exist circumstantial arguments to sustain such a formation. For instance, it has been shown that after the deposition of five monolayers by ALE, the CdTe lattice parameter relaxes without the formation of misfit dislocations (Besombes et al. 2000), by contrast with MBE grown structures (Cibert et al., 1990). Please note that the sample studied here lies on the low border of this transition. Before turning to the experimental results, it is worth commenting on the spatial resolution in our measurements. The exciting light lexc ˆ 514.5 nm is diffracted by the 200 nm aperture tip into waves which are either progressive or evanescent along the tip axis direction, depending on their in-plane spatial frequency. In-plane spatial frequencies above 1/lexc are evanescent in vacuum, but are converted into progressive waves in ZnTe provided that they are below the cut-off frequency in ZnTe: vcut-off ˆ (n/lexc), where n ˆ 3.2 is the optical index of ZnTe. This cut-off frequency is (160 nm)21. This calls for an adapted aperture of 160 nm for `optimal' excitation of the CdTe active layer because such an aperture diffracts the excitation light into waves that are all progressive in ZnTe. We have used tips in this range which determines the spatial resolution of our measurements (in some cases, this resolution may be further reduced due to diffusion of photo-excited carriers as explained below). The use of tips with smaller apertures, e.g. 50 nm, would improve the resolution but at the expense of a poorer transmission of the excitation to the CdTe QD layer (embedded at 60 nm underneath the surface) not only because such apertures have a smaller transmission (scaling

approximately like the 4th power of the aperture size) but also because they diffract the excitation light partly into waves that remain evanescent in the semiconductor. This stresses how important it is not to bury the active medium too deep in the structure. Figure 2 shows three PL spectra of sample 1 taken for various altitudes of the pulled fibre tip above the sample surface. The optical probe has a tip aperture diameter of approximately 200 nm. For the largest altitudes ± the far field region ± a weak and broad PL band is measured. However, as the tip approaches the sample surface, the broad PL band evolves into a set of discrete sharp peaks with approximately 15 transitions showing up in the near field. This behaviour is typical for a large ensemble of dots having different sizes, compositions and/or local environments, i.e. different emission properties being progressively resolved as spatial resolution is increased. Assuming a dot density of 1010 cm22, it can be estimated that the number of dots directly excited by the light spot under the tip aperture is approximately 7000 at an altitude of 3.5 mm, 180 at an altitude of 500 nm and only three (on average) at the `nearfield' altitude of 15 nm. Therefore, inhomogeneous broadening is dominant in the PL spectrum of the ensemble but at higher spatial resolution (smaller altitudes) individual dots start being resolved. There are two salient features in Fig. 2 which we wish to comment on briefly. First, a sharp peak at l ˆ 613.6 nm dominates the spectra at low tip altitudes and is still discernable in the far field. This could indicate a nonstatistical, dominant dot size in the investigated part of the sample. Second, the number of peaks observed in the near field is much larger than the estimated number of directly excited dots. Here several explanations can be invoked: (a) the dot density can be higher (a straightforward but unlikely explanation); (b) the illumination power might be too high to probe ground states of single excitons only; and (c) the actual spatial resolution is poorer than expected from the apex diameter of 200 nm. The very fact that the number of PL peaks does not decrease significantly from an altitude of 500 nm to 15 nm (though peaks at 15 nm are sharper and better resolved from the remaining broad PL) is a hint that the actual resolution achieved on Fig. 2 might be in the submm rather than in the 100 nm range. This is because photoexcited electron-hole pairs that are created under the tip diffuse laterally in the CdTe quantum well before recombining in a dot, thereby producing a PL emission which can be collected in the far field by the ellipsoidal mirror. A typical diffusion length in II-VI and III-IV semiconductor compounds is 0.5 to 1 mm (it can be determined with NSOM as shown by ObermuÈller et al., 1999). This resolution limiting effect could in principle be minimized by collecting PL through the same tip as used for excitation as explained further in the paper. To probe the effect of illumination power (explanation b q 2001 The Royal Microscopical Society, Journal of Microscopy, 202, 202±208

N EA R - FI E L D SP EC T RO S C O P Y O F Q UA N T U M D O T S

205

Fig. 2. Three low-temperature PL spectra of sample 1 measured with a pulled optical probe used for excitation only (excitation wavelength ˆ 514.5 nm, apex diameter ˆ 200 nm). From top to bottom, the tip altitude above the sample surface is 3.5 mm, 500 nm and 15 nm (the near-field zone), respectively. Please note the different vertical scales.

above), PL was measured at different discrete locations of the tip for several power densities. An example of the results is given in Fig. 3. The excitation power was estimated from the measured intensity injected into the fibre and assuming a 1025 transmission of the tip (in future work, the actual transmission of our tips will be measured which was not possible in these first experiments). Clearly, an increase of the excitation power by one order of magnitude strongly modifies the spectrum by producing a much larger number of discrete peaks accompanied by a rise of a broad PL background. At low enough excitation, luminescence originates from radiative recombinations of electron-hole pairs in their ground exciton states in the QDs where both

Fig. 3. Low-temperature near-field spectra of sample 1 measured as in Fig. 2 but for two excitation power densities quoted under the curves. The sharp peaks have spectral full width at half maximum Ê which is the (FWHM) of about 0.7 A resolution limit of our set-up in the present study. q 2001 The Royal Microscopical Society, Journal of Microscopy, 202, 202±208

the electron and the hole are in their conduction and valence ground states, respectively. At high excitation, luminescence can also take place between excited states of excitons (ObermuÈller et al., 1999), multiexciton states, e.g. biexcitons (Brunner et al., 1994; Kulakovskii et al., 1999), or photo-excited carriers can diffuse farther to more remote dots when recombination paths close to the tip are saturated. The spectra shown in Fig. 2 were taken in the low excitation limit of our set up. To investigate whether we were dealing with the one-exciton limit would have required a further reduction of the excitation power, leading to prohibitively long measurement times. This is left for future work. It would also be of interest to map the PL intensity for

206

M . B RU N E T A L .

Fig. 4. Low-temperature near-field spectra of sample 2 measured with a chemically etched fibre tip used for both excitation (excitation wavelength ˆ 514.5 nm) and PL collection. The apex diameter is 250 nm, the tip altitude is 15 nm. Spectra are taken for four different excitation densities. Arrows in the low-density spectrum point out two sharp peaks discussed in the text (see also Fig. 5).

several excitation powers in order to reveal possible spatial correlations among various PL features. In view of the behaviour revealed in Fig. 3 it is tempting to further increase the excitation power. However, we are facing here a real problem with a pulled fibre tip in that it might be irreversibly damaged by attempting to inject too much light into it. To circumvent this problem, we use a special optical probe obtained by chemical etching. Its apex diameter is 250 nm (experiments with thinner tips are in progress). As already mentioned, such fibre tips have large transmission and can be used for collection as well. It turns out that etched tips can also support a much larger injected

power. Figure 4 shows four spectra of sample two taken for four excitation powers covering almost two orders of magnitude. The power density was estimated by assuming a 1023 transmission of the tip aperture, which is reasonable in view of the far-field characterization carried out on the optical probe as compared to a pulled fibre tip. The spectra of Fig. 4 were collected through the tip itself, not through the ellipsoidal mirror. While at low density, the spectrum contains approximately 15 sharp peaks on the low energy side of a broad band centred at 578 nm, the latter progressively gains intensity with increasing power and eventually dominates the spectrum at large power. As a

Fig. 5. Three PL intensity coloured images of sample 2 shown for three PL wavelengths at which sharp peaks are observed. Here, PL collection is made through the ellipse as in Figs 2 and 3. The upper left spectrum shown for reference exhibits two of the imaged peaks at 581.23 nm and 587.31 nm. The third peak imaged at 589.65 nm does not show up in this reference spectrum. q 2001 The Royal Microscopical Society, Journal of Microscopy, 202, 202±208

N EA R - FI E L D SP EC T RO S C O P Y O F Q UA N T U M D O T S

matter of fact, the large-excitation spectrum resembles a far field spectrum with a 8-nm (, 30 meV) FWHM. This behaviour is probably due to the contribution of dots located further apart from the tip because recombination paths under the tip or in its vicinity saturate for increasing excitation. The higher the excitation, the larger the contribution of remote dots. Here again, additional (and somehow related) effects can make a contribution, such as recombination through excited excitonic states and/or multiexciton states or even laser heating of the illuminated area. The following question now arises: how can light emitted by remote dots still be collected by the tip? We believe that this happens through the following mechanism. The contribution of remote dots is primarily evanescent in the tip axis direction due to internal total reflection at the ZnTe/ vacuum interface. This evanescent field can propagate on the sample surface to reach the tip area where it can be diffracted partly into progressive waves in the tip material by the tip itself. However, such a mechanism should become less and less likely in the excitation-collection mode as emitting dots are located further away from the tip, by contrast with the collection mode through the ellipsoidal mirror (Figs 2 and 3) which has no spatial resolution (its focal `sphere' is estimated at 10 mm from approach curves). In the last case, there is only need for the PL light to be emitted out of the sample below the critical incidence for total internal reflection uc ˆ arcsin(1/n) to reach the collection optics. Therefore, we expect the spatial resolution to improve with collection through the tip. In order to check this promising scenario, we intend to compare both collection modes in situ with smaller aperture tips and at low excitation (the one-exciton limit). NSOM offers the unique possibility of mapping PL spectra and/or intensity. As an example, Fig. 5 shows three 2.1 mm  2.1 mm images of the PL intensity for three different reference wavelengths at which strong and sharp peaks are observed. Here, PL was collected in the far field by the ellipsoidal mirror. It can be seen that the peaks at 581.23 and 587.31 nm visible in Fig. 5 and in Fig. 4 originate from the same locations of the sample, i.e. from the same dot(s), by contrast with the peak at 589.65 nm (not visible in the spectra of Figs 4 and 5). Here, one or several other dots are involved. A tempting explanation for the peaks at 581.23 and 587.31 nm (energy separation of 22 meV) is to assign them to an exciton and a biexciton, respectively, as PL features due to these species are separated by 24 meV in CdSe (a relative of CdTe) QDs (Kulakovskii et al., 1999). However, our experimental results are presently not detailed enough to reach any definitive conclusion at this point although the peak intensities seem to follow the expected trend of a linear and super-linear dependence on the excitation power, respectively. q 2001 The Royal Microscopical Society, Journal of Microscopy, 202, 202±208

207

Acknowledgements We wish to dedicate this work to Y. Merle d'Aubigne who died last spring. We are grateful to him and to M. Vallade for having favoured the development of a low temperature NSOM activity at the L.S.P. This work has benefited from financial support from the Institut de la Physique de la MatieÁre CondenseÂe (IPMC), Grenoble, from the ReÂgion RhoÃne-Alpes and from the C.N.R.S.

References Bayer, M., Stern, O., Hawrylak, P., Fafard, S. & Forchel, A. (2000) Hidden symmetries in the energy levels of excitonic `artificial atoms'. Nature, 405, 923±926. Besombes, L., Marsal, L., Kheng, K., Wasiela, A., Bleuse, J. & Mariette, H. (2001) Fine structure of single zero-dimensional excitons in CdTe/ZnTe quantum dots. Phys. Rev. B, in press. Betzig, E., Trautman, J.K., Harris, T.D., Weiner, J.S. & Kostelak, R.L. (1991) Breaking the diffraction barrier: optical microscopy on a nanometric scale. Science, 251, 1468±1470. Brunner, K., Abstreiter, G., BoÈhm, G., TraÈnkle, G. & Weimann, G. (1994) Sharp-line photoluminescence and two-photon absorption of zero-dimensional biexcitons in a GaAs/AlGaAs structure. Phys. Rev. Lett. 73, 1138±1141. Cibert, J., Gobil, Y., Le Si Dang, Tatarenko, S., Feuillet, G., Jouneau, P.H. & Saminadayar, K. (1990) Critical thickness of epitaxial CdTe/ZnTe. Appl. Phys. Lett. 56, 292±294. Drexler, H., Leonard, D., Hansen, W., Kotthaus, J.P. & Petroff, P.M. (1994) Spectroscopy of quantum levels in charge-tunable InGaAs quantum dots. Phys. Rev. Lett. 73, 2252±2255. Eytan, G., Yayon, Y., Rappaport, M., Shtrikman, H. & Bar-Joseph, I. (1998) Near-field spectroscopy of a gated electron gas: a direct evidence for electron localization. Phys. Rev. Lett. 81, 1666± 1669. Hartmann, J.M., Feuillet, G., Charleux, M. & Mariette, H. (1996) Atomic layer epitaxy of CdTe and MnTe. J. Appl. Phys. 79, 3035± 3041. Hess, H.F., Betzig, E., Harris, T.D., Pfeiffer, L.N. & West, K.W. (1994) Near-field spectroscopy of the quantum constituents of a luminescent system. Science, 264, 1740±1745. Karrai, K. & Grober, R.D. (1995) Piezoelectric tip-sample distance control for near field optical microscopes. Appl. Phys. Lett. 66, 1842±1844. Kulakovskii, V.D., Bascher, G., Weigand, R. et al. (1999) Fine structure of biexciton emission in symmetric and asymmetric CdSe/ZnSe single quantum dots. Phys. Rev. Lett. 82, 1780± 1783. LundstroÈm, T., Schoenfeld, W., Lee, H. & Petroff, P.M. (1999) Exciton storage in semiconductor self-assembled quantum dots. Science, 286, 2312±2314. Marzin, J.-Y., GeÂrard, J.-M., IzraeÈl, A., Barrier, D. & Bastard, G. (1994) Photoluminescence of single quantum dots obtained by self-organized growth on GaAs. Phys. Rev. Lett. 73, 716± 719. ObermuÈller, C., Deisenrieder, A., Abstreiter, G. et al. (1999) Mechanical nano-manipulation of single strain-induced semiconductor quantum dots. Appl. Phys. Lett. 75, 358±360.

208

M . B RU N E T A L .

ObermuÈller, C. & Karrai, K. (1995) Far field characterisation of diffracting circular apertures. Appl. Phys. Lett. 67, 3408± 3410. Warburton, R.J., SchaÈflein, C., Haft, D, et al. (2000) Optical emission from a charge-tunable quantum ring. Nature, 405, 926±929.

Warburton, R.J., DuÈrr, C.S., Karrai, K., Kotthaus, J.P., MedeirosRibeiro, G. & Petroff, P.M. (1997) Charged excitons in selfassembled semiconductor quantum dots. Phys. Rev. Lett. 79, 5282±5285. Zrenner, A. (2000). A close look on single quantum dots. J. Chem. Phys. 112, 7790±7798.

q 2001 The Royal Microscopical Society, Journal of Microscopy, 202, 202±208