SURFACE AND INTERFACE ANALYSIS Surf. Interface Anal. 27, 491È494 (1999) SNOM-induced Photoluminescence of Individual InGaAs Quantum Dots Using Etched Metal-coated Fibre Tips I. Manke,* J. Lorbacher, J. L. Spithoven,¤ F. Heinrichsdor† and M. Da hne-Prietsch Institut fur Festkorperphysik, Technische Universitat Berlin, Hardenbergstr. 36, D-10623 Berlin, Germany Scanning near-Ðeld optical microscopy is used to investigate the photoluminescence of individual In Ga As 0.4 0.6 quantum dots. The high transmission efficiency of etched Ðbre tips allows investigations of quantum dots at room temperature with sufficient intensity, even after coating the tip with metal for background elimination. The pure background-free signal of single quantum dots results in Lorentzian photoluminescence spectra with Ðnite line widths of 10–20 meV at room temperature to Æ1 meV at 4 K. This is in contrast to the expectation of extremely narrow lines in the microelectron-volt region. Copyright ( 1999 John Wiley & Sons, Ltd. KEYWORDS : quantum dots ; semiconductor nanostructures ; SNOM ; NSOM INTRODUCTION Scanning near-Ðeld optical microscopy (SNOM),1,2 with its outstanding spatial resolution, allows the optical investigation of individual semiconductor nanostructures that are very much less than 1 lm apart. Quantum dots are of special interest in this Ðeld because their discrete electronic energy levels are expected to result in extremely narrow photoluminescence lines.3h5 Strained quantum dots can be prepared in a self-organized way by StranskiÈKrastanov growth, resulting in an inhomogeneous size distribution of the dots on top of a wetting layer. Spatially averaging photoluminescence experiments of such a dot ensemble regularly yield inhomogeneously broadened transition lines,6,7 whereas SNOM allows the study of speciÐc characteristics of individual dots.8 In this work, we present SNOM-induced photoluminescence results of In Ga As quantum dots 0.6etched Ðbre tips, the embedded in a GaAs matrix.0.4Using photoluminescence intensity in internal-reÑection geometry is considerably higher than in the case of pulled tips. This enables measurements of single quantum dots with sufficient intensity, even after coating the Ðbre tips with metal for background elimination. In this way, the pure signal of single quantum dots could be observed. A strongly reduced spatial resolution is observed in experiments using far-Ðeld detection of the photoluminescence signal, which is related to extensive di†usion of the excited carriers. The photoluminescence spectra of individual quantum dots are characterized by Lorentzian lines of Ðnite width 10È20 meV at room temperature, decreasing substan* Correspondence to : I. Manke, Institut fur Festkorperphysik, Technische Universitat Berlin, Hardenbergstrasse 36, D-10623 Berlin, Germany. ¤ On leave from : Faculteit Natuur- & Sterrenkunde, Universiteit Utrecht, PO Box 80000, 3508 TA Utrecht, The Netherlands. CCC 0142È2421/99/050491È04 $17.50 Copyright ( 1999 John Wiley & Sons, Ltd. tially with decreasing temperatures to \1 meV at 4 K. This observation is in contrast both to the theoretical prediction of extremely narrow lines3,4 and to experimental results on similar InAs quantum dot samples.5 The broadening can be related to a lifetime e†ect of thermally activated electrons in the dots. EXPERIMENTAL For the experiments, two di†erent home-built SNOM set-ups were used. One design, mounted in a glass dewar, is used for measurements down to liquid nitrogen temperature.8 A second design is built in a continuous-Ñow He cryostat for temperatures down to 4 K.9 Scanning of the sample is performed by a quadrant-sectored tube piezo (EBL 2) with 25 mm length and 3.2 mm diameter, yielding a scan range of D100 lm ] 100 lm at 300 K and D20 lm ] 20 lm at 4 K. The tip height is controlled by non-optical shearforce detection using a tuning fork with the tip glued to one side.8,10 A vertical resolution of D0.5 nm and a lateral resolution of D15 nm are obtained. At constant temperatures, the thermal drift is D0.1 lm h~1. In contrast to our previous work performed with uncoated pulled tips,8,11 the SNOM tips used here were prepared by chemical etching and subsequent metal coating. Multimode Ðbres were etched in 48% hydroÑuoric acid covered with a layer of baby oil12, resulting in an opening angle of D20¡. Metal coating was performed in a high-vacuum chamber at 10~6 mbar by evaporation of 100È200 nm thick Ag or Al Ðlms under an angle of D75¡ from the Ðbre axis while rotating the tip. We did not observe signiÐcant di†erences between Al- and Ag-coated tips, except that the latter were stable for a longer time. Because the apex of these tips was often not transparent, an opening had to be prepared by gently polishing the tip apex at the sample surface using the lateral vibration of the tuning fork. Received 15 September 1998 Accepted 8 December 1998 492 I. MANKE ET AL . The opening of the aperture was then detected simply by the onset of the photoluminescence signal. The sample was excited optically by a HeÈNe laser with hl \ 1.96 eV, and the signal was detected using a monochromator with a resolution of D1 meV and a Peltier-cooled InGaAs photodiode with a noise level of D8 ] 10~17 W JHz ~1. For spectrally integrated photoluminescence imaging, the monochromator was replaced by a 1.12 eV low-pass Ðlter in order to suppress the radiation from the laser and from recombination outside the dots. Measurements in internalreÑection geometry were performed using a 2 ] 1 Ðbre coupler for feeding the laser radiation through the tip onto the sample and collecting the photoluminescence light by the tip for detection. For far-Ðeld excitation or detection, a cleaved Ðbre was used in close position to the tip apex. The In Ga As quantum dot sample was prepared 0.4 0.6 using metal-organic chemical vapour deposition by growing 0.94 nm In Ga As on an undoped 0.6 GaAs(001) substrate and0.4 subsequent capping with a 20 nm thick GaAs and AlGaAs layer.6,7 They were characterized by transmission electron microscopy (TEM), yielding an area dot density of D10 lm~2, and by conventional spatially averaging photoluminescence.7 In the latter experiment, transitions at hl \ 0.93, 1.02 and 1.10 eV were observed at room temperature, which were inhomogeneously broadened by D40 meV. These three lines were explained on the basis of a term scheme with one electronic quantum state for the electrons and at least three states for the holes,13 resulting in the socalled ground-state transition and two or more transitions involving excited hole states. RESULTS AND DISCUSSION Plate 1 shows a bird view of the spectrally integrated photoluminescence intensity of the quantum dots taken with an etched, metal-coated tip at room temperature in internal-reÑection geometry. Most peaks represent the emission from individual dots, demonstrating a lateral resolution (FWHM) of D300 nm, whereas some broader and brighter peaks can be assigned to groups consisting of a few dots. The emission intensities of the quantum dots are usually very similar. At tip positions between the dots the signal reduces almost to zero, so that there is only a negligible background signal from neighbouring dots, which is in strong contrast to our previous data obtained using uncoated pulled Ðbre tips.8,11 In addition, the photoluminescence intensity is now about two orders of magnitude higher. Thus it is clearly demonstrated that etched coated Ðbre tips operated in internal-reÑection geometry are very useful for resolving the pure signal from individual quantum dots with sufficient intensity. In Plate 2 we present alternative imaging results using far-Ðeld excitation or detection. Far-Ðeld excitation results are shown in Plate 2(a), where the quantum dots are observed with a lateral resolution similar to that in internal-reÑection geometry. When reversing the light direction, the quantum dots are excited in the near Ðeld and the photoluminescence is collected in the far Ðeld. In this case, di†usion e†ects of the carriers will Surf. Interface Anal. 27, 491È494 (1999) have a strong inÑuence on the lateral resolution. Indeed, such a behaviour is observed in Plate 2(b), where the quantum dots are no longer resolved. Only small variations in intensity are found when going from the lower left of the image to the upper right, which is related to a higher dot density in the lower left region. Plate 2(c) shows contour lines through the two images. Because the observed lateral variations in far-Ðeld detection extend over [1 lm, a di†usion length of the excited carriers far above 1 lm is derived. With the exception of di†usion length studies, the internal-reÑection geometry is much more useful for studying quantum dots with high spatial resolution and sufficient intensity, because of more easy handling and the lack of tip-shadowing e†ects, yielding more reproducible results. Even in the case of small amounts of light penetrating through the tip Ñanks, e.g. due to incomplete coverage or holes in the metal coating, only a small, laterally constant background was observed in internal-reÑection geometry. Therefore, we preferred this mode in all further experiments. In Plate 3, spectrally resolved photoluminescence images in internal reÑection mode are shown at three di†erent detection energies. Some dots appear at only one photon energy, e.g. the dot marked 1, but others can be seen at two or even at all three di†erent energies, marked 2 and 3, respectively. This observation indicates a Ðnite spectral width of the dot emission in the millielectron-volt range, in contrast to the expectation of extremely narrow lines of the order of several microelectron-volts.3h5 In order to analyse this behaviour in more detail, photoluminescence spectra of single dots at di†erent positions on the sample were recorded, as shown in Fig. 1. In Fig. 1(a) the room-temperature emission from a single dot is displayed, including the transitions from ground state and Ðrst excited state at 0.935 eV and 1.02 eV, respectively. The spectrum shown in Fig. 1(b) displays the ground-state transitions from two neighbouring dots. All spectra are characterized by relatively broad symmetric lines that can be well described by a Lorentzian lineshape, as revealed from a least-squares Figure 1. (a,b) Photoluminescence spectra of individual quantum dots, taken at 300 K in internal-reflection geometry (data points), together with Lorentzian fit curves (solid lines). (c) Spectrum taken at 4 K. The different line widths (FWHM) are indicated, varying between 10 and 20 meV from dot to dot at room temperature down to below 1 meV at 4 K. Copyright ( 1999 John Wiley & Sons, Ltd. Plate 1. Spectrally-integrated photoluminescence image of In0.4Ga0.6As quantum dots taken at 300 K in internal-reflection geometry. Plate 3. Spectrally-resolved photoluminscence image In0.4Ga0.6As quantum dots perfomed at 300 K in internal-reflection geometry at the same sample with different detection energies. Three positions are marked with numbers, indicating a quantum dot in (1) only one image, (2) two images and (3) all three images. Plate 2. Images of the spectrally-integrated photoluminescence intensity taken at 300 K, using (a) farfield excitation and (b) farfield detection. Both images were taken at the same sample area. (c) Cross sections along the lines in (a) and in (b). Copyright © 1999 John Wiley & Sons, Ltd. Surf. Interface Anal. 27 (1999) SNOM OF InGaAs QUANTUM DOTS Ðt analysis (solid lines in Fig. 1). The analysis of the spectra from several dots results in Lorentzian line widths (FWHM) scattered between 10 and 20 meV at room temperature. It is noted that the Ðrst excited-state transition also shows comparable line widths. In addition, a strong temperature dependence was observed. Figure 1(c) shows the photoluminescence of a few dots taken at 4 K. Around 1.0 eV, i.e. in the spectral range of the ground-state transition at this temperature, it is clearly observed that the line width is now \1 meV, mainly determined by the resolution of the monochromator. In order to investigate this transition in more detail, spectra of the same dots were taken at di†erent temperatures. It is difficult to track the photoluminescence emission from selected quantum dots in the complete temperature range 4È300 K because of the high thermal drift, amounting to D100 lm in our experimental set-up. Therefore, an experiment was performed where an uncoated tip was Ðxed mechanically at a certain position of the sample by gently touching the surface, in this way compensating for drift e†ects and allowing observation of the temperature dependence of the spectra from a certain group of quantum dots. Figure 2 shows the resulting photoluminescence spectra at di†erent temperatures, originating mainly from three dots. In the ground-state emission, the photoluminescence signal of the three dots cannot be resolved very well because of the use of an uncoated Ðbre tip, resulting in a higher background from neigh- Figure 2. Temperature dependence of the photoluminescence of a group of about three quantum dots, taken with an etched uncoated tip positioned at a fixed location of the surface. The line width (FWHM) decreases from Á20 meV at 300 K down to the monochromator resolution of Á4 meV at 5 K. Copyright ( 1999 John Wiley & Sons, Ltd. 493 Figure 3. Arrhenius plot of the line width as a function of temperature. bouring dots.8,11 In the excited-state photoluminescence, in contrast, sufficient excitation intensity is required to observe this transition, so the background is more suppressed. Here, it is clearly observed that the line width is decreasing with decreasing temperature from D20 meV at 300 K to \4 meV at 5 K. It should be noted that the spectral resolution in this experiment was only D4 meV, so the actual line width at low temperatures will be much lower, as has been demonstrated already in Fig. 1(c). Furthermore, the photoluminescence intensity decreases by less than one order of magnitude when going to higher temperatures, and the photoluminescence energies show the expected red shift by D60 meV. The Ðnite temperature-dependent linewidth and the Lorentzian line shape can be explained by thermal excitation of carriers in the dot, reducing the lifetime particularly at higher temperatures. In this case, the thermal excitation probability (and therewith the line width) is given by ! \ ! exp([E /k T ), where E is the activa0 Arrhenius A B plot of theA line width is tion energy. An shown in Fig. 3. Because of the spread in the observed line widths of di†erent dots at a given temperature, the activation energy can only be determined with a rather large error bar to 3 ^ 2 meV. Based on the model that the photoluminescence is described by transitions between one electron state and three hole states,7,13 and because the di†erent optical transitions are a†ected by the thermal broadening in a similar way, it can be speculated that the electron state is only weakly bound by the activation energy as determined above. A model of the band structure is shown in Fig. 4. The thermal activation of electrons in the dot, either to excited states within the dot or to continuum states in the two-dimensional wetting layer, can result in the observed broadening. This model requires that there is no signiÐcant optical-transition probability from the excited electron states to the hole states, e.g. due to a strongly reduced wavefunction overlap, otherwise intense additional lines would be observed. An activation energy in this range also agrees with the intensity variation observed in Fig. 2 : at thermal energies exceeding the activation energy, a signiÐcant fraction of the Surf. Interface Anal. 27, 491È494 (1999) 494 I. MANKE ET AL . Boltzmann distribution, i.e. strongly asymmetric, and the line width would be comparable for all dots, this model can be rejected. It should be noted that calculations for InAs quantum dots result in much higher electron binding energies of [100 meV.13 The di†erent values derived here for In Ga As dots can be assigned to di†erent 0.4 0.6 modifying the electronic structure structural properties, as compared with InAs dots. Because of this discrepancy and because of the lack of theoretical investigations on the electronic structure of InGaAs dots, the presentation of a conclusive model of the electronic structure for explaining the observed line width behaviour is presently not advisable. In order to investigate the line shape variation in more detail, more experiments at low temperatures are currently under way. CONCLUSIONS Figure 4. Model of the electronic structure of the InGaAs quantum dots. The electron is only weakly bound with a binding energy of a few millielectron-volts, resulting in a thermal activation at higher temperatures to excited electron states, e.g. in the wetting layer. The line width is thus identical for all three optical transitions. electrons is excited and therefore does not contribute to the photoluminescence signal. It can also be assumed that there are no bound electron states at all in the dot. In this case, the dot hole state would recombine with an electron from the twodimensional wetting layer, also leading to a temperature-dependent line broadening of the emission lines. Because the expected line shape would be the result of the two-dimensional density of state and the In this work we have demonstrated that SNOM experiments with etched metal-coated Ðbre tips provide sufficient spatial resolution and signal intensity to separate photoluminescence features from individual quantum dots. We clearly observe that the spectral line width of the dots is much broader than expected and can be described by a Lorentzian. It is assumed that the broadening is due to a lifetime e†ect, possibly related to a low binding energy of the electron in the dots. Acknowledgements W. Busse, C. K. Kim, T. Kalka and A. Bauer are acknowledged for assistance during the experiment and M. Grundmann and D. Bimberg for discussion of the results. Part of this work was supported by Sonderforschungsbereich 296 of the Deutsche Forschungsgemeinschaft. REFERENCES 1. M. A. Paesler and P. J. Moyer, Near -Field Optics : Theory , Instrumentation , and Applications . Wiley–Interscience, New York (1996). 2. R. Wiesendanger, Scanning Probe Microscopy and Spectroscopy . Cambridge University Press, Cambridge (1994). 3. M. J. Kelly, Low -dimensional Semiconductors . Clarendon Press, Oxford (1995). 4. U. Woggon, Optical Properties of Semiconductor Quantum Dots . Springer, Berlin (1996). 5. M. Grundmann, J. Christen, N. N. Ledentsov, J. Bo hrer, D. Bimberg, S. S. Ruvimov, P. Werner, U. Richter, U. Go sele, J. Heydenreich, V. M. Ustinov, A. Y. Egorov, A. E. Zhukov, P. S. Kop’ev and Z. I. Alferov, Phys . Rev . Lett . 74, 4043 (1995). 6. F. Heinrichsdorff, A. Krost, M. Grundmann, D. Bimberg, A. Kosogov and P. Werner, Appl . Phys . Lett . 68, 3184 (1996). Surf. Interface Anal. 27, 491È494 (1999) 7. F. Heinrichsdorff, A. Krost, M. Grundmann, D. Bimberg, A. Kosogov, P. Werner, F. Bertram and J. Christen, in The Physics of Semiconductors , ed. by M. Scheffler and R. Zimmermann, p. 1321. World Scientific (1996). 8. I. Manke, D. Pahlke, J. Lorbacher, W. Busse, T. Kalka, W. Richter and M. Da hne-Prietsch, Appl . Phys . A 66, S381 (1998). 9. J. Lorbacher, I. Manke, J. L. Spithoven and M. Da hnePrietsch, to be published. 10. K. Karrai and R. D. Grober, Appl . Phys . Lett . 66, 1842 (1995). 11. D. Pahlke, I. Manke, F. Heinrichsdorff, M. Da hne-Prietsch and W. Richter, Appl . Surf . Sci . 123/124, 400 (1998). 12. D. R. Turner, US Patent 4,469,554 (1984). 13. M. Grundmann, O. Stier and D. Bimberg, Phys . Rev . B 52, 11969 (1995). Copyright ( 1999 John Wiley & Sons, Ltd.