Hybrid Epitaxial−Colloidal Semiconductor Nanostructures - American

subsequent epitaxial growth of a 2D-cap layer, these quantum dots are completely .... the ZnSe cap layer (Figure 2, center and right-hand side). Becau...
0 downloads 0 Views 452KB Size
Download PDF
NANO LETTERS

Hybrid Epitaxial−Colloidal Semiconductor Nanostructures

2005 Vol. 5, No. 3 483-490

U. Woggon,* E. Herz, and O. Scho1 ps Fachbereich Physik, UniVersita¨t Dortmund, Otto-Hahn-Str. 4, 44227 Dortmund, Germany

M. V. Artemyev Institute for Physico-Chemical Problems of Belarussian State UniVersity, Minsk 220080, Belarus

Ch. Arens, N. Rousseau, D. Schikora, and K. Lischka Department Physik, UniVersita¨t Paderborn, Warburger Str. 100, 33908 Paderborn, Germany

D. Litvinov and D. Gerthsen Laboratorium fu¨r Elektronenstrahlmikroskopie, UniVersita¨t Karlsruhe, Engesser Str. 7, 76128 Karlsruhe, Germany Received November 19, 2004; Revised Manuscript Received January 12, 2005

ABSTRACT We present a growth technique which combines wet-chemical growth and molecular beam epitaxy (MBE) to create complex semiconductor nanostructures with nanocrystals as active optical material. The obtained results show that wet-chemically prepared semiconductor nanocrystals can be incorporated in an epitaxally grown crystalline cap layer. As an exemplary system we chose CdSe nanorods and CdSe(ZnS) core−shell nanocrystals in ZnSe and discuss the two limits of thin (d∼2R) and thick (d>2R) ZnSe cap layers of thickness d for CdSe nanorods and nanodots of radii R between 2 and 4 nm. In contrast to the strain-induced CdSe/ZnSe Stranski−Krastanow growth of a quantum dot layer in a semiconductor heterostructure, the technique proposed here does not rely on strain and thus results in additional degrees of freedom for choosing composition, concentration, shape, and size of the nanocrystals. Transmission electron microscopy and X-ray diffractometry show that the ZnSe cap layer is of high crystalline quality and provides all parameters for a consecutive growth of Bragg structures, waveguides, or diode structures for electrical injection.

Semiconductor nanocrystals exhibit an enormous potential concerning size- and shape-controlled tunability of their emission spectra. Many approaches are being tested to incorporate them in a controlled way into various photonic devices such as photonic crystals, microcavity-based switches, single-photon sources, or light-emitting diodes and lasers.1 Tunable photonic structures are produced by using either epitaxially grown or wet-chemically prepared nanostructures. We propose here an alternative way to create complex monolithic semiconductor structures with nanocrystals as active optical material: an epitaxial growth technique based on molecular beam epitaxy (MBE) where externally wet* Corresponding author. Ulrike Woggon, Experimentelle Physik IIb, Universita¨t Dortmund, Otto-Hahn-Str. 4, D-44227 Dortmund, phone: +49231 755 3767, FAX: +49-231 755 3674), E-mail: [email protected]. 10.1021/nl0480870 CCC: $30.25 Published on Web 01/28/2005

© 2005 American Chemical Society

chemically prepared, colloidal nanocrystals are fully integrated into monolithic epitaxial heterostructures. In standard molecular beam epitaxy (MBE), quantum dots grow when a semiconductor material is deposited on a semiconductor substrate with a sufficiently different crystal lattice constant. The coherently strained nanostructure grows three-dimensionally on a wetting layer (Stranski-Krastanow (SK) growth) and forms a quantum dot with threedimensional electronic confinement of excitons.2 After subsequent epitaxial growth of a 2D-cap layer, these quantum dots are completely embedded in a single-crystal heterostructure. These, epitaxially grown, SK dots can be stacked in crystalline multilayer structures and used for further microstructuring, e.g. by standard microlithography techniques to prepare waveguide structures, Bragg mirrors, and other elements of integrated photonic devices. Semiconductor heterostructures with incorporated layers of SK-quantum dots

Figure 1. Overview of possible applications of hybrid epitaxial-colloidal semiconductor nanostructures. The externally, wet-chemically prepared nanocrystals are epitaxially capped by molecular beam epitaxy (MBE).

produced in a closed-cycle MBE process (e.g., InAs/GaAs, CdSe/ZnSe) are widely used in quantum dot lasers, planar microcavities or microdisk and micropillar cavities, and semiconductor optical amplifiers. They are considered to be prospective candidates for single-photon sources or qubits in quantum information processing under the assumption that the quantum dots can be grown with low area density and controlled positions allowing for the spatial addressing of single quantum dots, e.g., by single cavity photon modes. Colloidal nanodots and nanorods, the wet-chemical realization of quantum confined structures, are commonly synthesized in high-temperature reactions of organometallic precursors in a highly coordinating medium.3,4 Such nanocrystals possess a high photoluminescence (PL) quantum yield, exceeding 70% at room temperature. The optical properties of nanocrystals can be widely tuned by variation of size and shape which in turn can be easily controlled during the wet-chemical synthesis.5 In a Stranski-Krastanow growth process, in contrast, the size and size distribution of the 3D islands is determined by the strain parameters and cannot be varied much within a given material system. It does not allow for the simultaneous deposition of two or more different types of nanocrystals within one layer. The quantum dot density in SK structures is difficult to tune below a critical value of about 1 × 108 cm-2. Wet-chemically prepared core and core-shell nanocrystals can be deposited with very low concentrations, however, they show a relatively fast (with respect to SK dots) photodegradation and photoluminescence intermittency. Blinking phenomena are well-known and a drawback for applications on the singledot/single-photon level. 484

In this work we propose to combine the wet-chemical and MBE growth mode to integrate colloidal nanocrystals in epitaxially grown heterostructures. The Stranski-Krastanow growth mode has been the primary means of epitaxial (MBE) growth of quantum dots. However, it relies on strain due to the lattice parameter mismatch between specific material combinations for substrate and epitaxial deposits. Our approach for the growth of a complete heterostructure containing quantum dots or nanorods does not rely on strain which yields additional degrees of freedom in choosing composition, concentration, shape, and size of the nanocrystals. A combination of wet-chemical and MBE growth is promising, e.g., for realizing efficient carrier injection into nanocrystals, to enhance photostability, to suppress blinking of nanocrystals, and to control concentration, size, shape, and position of the nanoemitters (for an overview about possible applications see Figure 1). The hybrid growth of colloidal-epitaxial structures has a huge potential for future technologies and device applications. It allows for the use of colloidal nanocrystals as a chemical carrier for doping with light-emitting ions (e.g., Mn-, Co-, or Eu-doped nanocrystals), the overgrowth of nanocrystals of different material on patterned substrates, or diode concepts based on efficient electrical carrier injection via an epitaxially grown pin structure. When hybrid epitaxial-colloidal growth is applied, the most crucial problems that have to be solved experimentally are the realization of high quality epitaxy after growth interruption and wetting the surface with those organic solvents in which the colloidal nanocrystals are kept. The colloidal nanocrystals have to be stable during the MBE growth Nano Lett., Vol. 5, No. 3, 2005

Figure 2. SEM images of colloidal CdSe nanorods (NRs) deposited on ZnSe substrate and overgrown with a ZnSe cap layer. (left) CdSe NRs capped with 6 nm ZnSe, high NR density; (middle) CdSe NRs capped with 6 nm ZnSe, low NR density, (right) CdSe NRs capped with 12 nm ZnSe, low NR density. The bar is a 50 nm scale, the average NR size derived from the SEM picture is 34.7( 0.5 nm in length and 6.7 ( 0.5 nm in diameter.

process, which usually requires elevated temperatures above 300 °C. For application in complex photonic devices, a hybrid epitaxial-colloidal semiconductor nanostructure has to fulfill such requirements as: no or negligible shape and composition change of the nanocrystals, direct connection of the crystal lattices of the guest nanocrystal and the host semiconductor layer, and high crystal quality of the cap layer to facilitate compatibility with the MBE growth that follows and any post-growth processing. The optical activity should be maintained, i.e., the emission should show high quantum efficiency, even at room temperature, and the structure should have a pronounced absorption band. The embedding of CdSe nanocrystals into a polycrystalline ZnSe layer was already demonstrated by MOCVD in 1994.6 In this work we present first results for the incorporation of colloidal CdSe and CdSe(ZnS) nanocrystals in a crystalline ZnSe cap layer, epitaxially grown on a ZnSe buffer layer without formation of a ZnCdSe wetting layer as observed, e.g., in standard, closed-cycle MBE growth of quantum dots.7 We analyze the structural and optical properties of ZnSe/CdSe-NC/ZnSe hybrid epitaxial-colloidal semiconductor structures consisting of CdSe(ZnS) core-shell nanodots (NDs) or CdSe nanorods (NRs) on top of a ZnSe buffer layer and overgrown with a ZnSe cap layer of thickness d between 2.5 and 22 nm. The obtained results clearly show the compatibility of wetchemical growth and MBE growth. We will focus in our work on the two important limits of thin (d∼2R) and thick (d>2R) cap layer thickness: (i) a thin cap layer (d∼2R) is essential for positioning and fixing of colloidal nanocrystals, (ii) a thick cap layer (d>2R) of high crystalline quality is a prerequisite for consecutive integration in monolithic photonic devices. Sample Growth. We use colloidal solutions of CdSe nanorods and CdSe(ZnS) core-shell nanodots that have been synthesized by the standard high-temperature reaction of organometallic precursors in the strongly coordinating solvents, like trioctylphosphine oxide (TOPO) and hexadecylamine (HDA) (see for example ref 8 and refs therein). Since high-quality MBE growth is very sensitive to the presence of organic molecules (i.e., TOPO and HDA) on the surface of the ZnSe buffer layer, we have extensively purified the colloidal solution of nanocrystals by repeatedly using the Nano Lett., Vol. 5, No. 3, 2005

deposition/redispersion technique.3 The nanocrystals were dissolved in high-purity pyridine and aged 24 h prior to MBE growth. Pyridine has been selected as the optimum solvent for MBE since it produces stable colloidal nanocrystal solutions with little aggregation because of its high complexation with surface metal atoms. Likewise, pyridine is capable of completely replacing TOPO and HDA molecules on the nanocrystal surface.9 Additionally, being a highly volatile solvent, pyridine will easily be removed either from the surface of the ZnSe buffer layer or from the nanocrystal surface at high vacuum and moderate heat pretreatment inside the MBE chamber. The externally, wet-chemically prepared CdSe NRs have a radius R of 3.5 nm and a length of 35 nm, the nanocrystals are CdSe(ZnS) core-shell nanodots with a core radius R of 2.5 nm and a thin ZnS shell of 1 to 2 monolayers.10 In the MBE process, the hybrid growth samples were produced in three steps. First, the aforementioned ZnSe buffer layer with a thickness of 45 nm was grown on a (001)-GaAs substrate by MBE at 280 °C. The growth was controlled by reflection high energy electron diffraction (RHEED), and the structural properties were examined ex-situ by high resolution X-ray diffraction (HRXRD). Reciprocal space maps (RSM) around the (224) reflex indicate a coherently strained ZnSe buffer layer grown on GaAs (see Supporting Information). In a second step, the sample was removed from the MBE chamber and a few microliters of highly diluted colloidal solution of CdSe nanocrystals in pyridine were deposited under clean-room conditions on the ZnSe buffer. In a third step, the sample was transferred back into the MBE chamber, heated to 240 °C and overgrown by migration-enhanced epitaxy (MEE), a special MBE technique which allows an enhanced movement of adsorbed atoms on the substrate surface. The insitu RHEED pattern taken during the ZnSe cap layer growth after nanocrystal deposition shows a transition from a dominant 3D growth below a nominal cap layer thickness of 5 nm toward a clear 2D cap layer growth above this critical thickness (see Supporting Information). In Figure 2 we show a few examples of scanning electron microscopy (SEM) images for CdSe nanorods covered by ZnSe as described above. The colloidal CdSe nanorods (NRs) are still visible after overgrowth with a thin ZnSe cap layer 485

Figure 3. Comparison of the photoluminescence spectra of colloidal CdSe nanorods deposited on quartz (dashed curve) and on a ZnSe buffer layer after MBE overgrowth with a 6 nm ZnSe cap layer at 240°C (solid curve) measured at T ) 15 K (note the logarithmic scale).

of 6 nm by MEE at 240 °C with a growth rate of ∼0,255 ML/s (Figure 2 left-hand side and center). In the case of thin cap layer overgrowth (d∼2R), the SEM images indicate that the overgrown NRs have almost the same shape and size as those in solution and do not transform into a thin ZnCdSe alloy layer on top of the ZnSe buffer layer. The lower the nanocrystal density the smoother the surface of the ZnSe cap layer (Figure 2, center and right-hand side). Because the information we can obtain from such SEM pictures is limited to surface properties, we will analyze in the following section the structural and optical properties in more detail. Thin ZnSe Cap Layer. In Figure 3, we compare the photoluminescence spectra of colloidal CdSe nanorods deposited on a quartz substrate (dashed line) and on top of the ZnSe buffer layer after overgrowth with a ZnSe cap layer of d ) 6 nm (solid line), which is close to the NR diameter. The NRs remain optically active with a slightly blue-shifted emission, most probably due to a change in confinement conditions (diameter, barrier height, Cd-Zn gradient, etc.) or in a change of surface polarizations caused by the NR embedding in ZnSe. While the line shape of the NR emission spectrum does not change during the MBE overgrowth with a ZnSe cap layer, the PL efficiency decreases by about 1 order of magnitude with respect to the spectrum of the NRs on quartz. Both at low and room temperature, the spectrum of the hybrid epitaxial-colloidal CdSe NR sample does not show any defect-related bands within the dynamic range in detector sensitivity (3 orders of magnitude, see Figure 3). Figure 4 shows a cross-section high-resolution transmission electron microscope (HRTEM) image of a single colloidal CdSe nanorod overgrown by a thin layer of about 1 nm of ZnSe by migration-enhanced epitaxy (MEE) at 240 °C. The ZnSe buffer and cap layers are viewed along the [1-10] zone axis. In contrast to the cubic zinc blende 486

Figure 4. Cross-section HRTEM image of a single colloidal CdSe nanorod overgrown with a thin layer of ZnSe by migration-enhanced epitaxy (MEE) at 240 °C.

structure of the ZnSe, the CdSe NR occurs in the hexagonal wurtzite structure. Due to the different crystal structures, the nanorod can be clearly distinguished from the ZnSe. The nanorod with a diameter of 7 nm is oriented with its longitudinal axis parallel to the electron beam. The image pattern corresponds to a [0001]-zone axis pattern, indicating that the longitudinal axis is aligned along the [0001] direction. This information indicates that after heating to 240 °C and a short MEE growth period, the CdSe NR still exhibits its original size and crystal structure. The original surface of the ZnSe buffer layer indicated by the solid line in Figure 4 has been extrapolated from the interface position between the ZnSe buffer and the ZnSe cap layer from an overview TEM image of a larger region around this particular nanorod. No gap between the lower part of the nanocrystal and the ZnSe buffer layer is observed. The interface position with respect to the position of the nanorod demonstrates that the lower part of the NC is embedded in the ZnSe cap layer. The overgrowth also fixes the NR position on the ZnSe surface. The nanocrystal does not “float” on the ZnSe during MEE- ZnSe deposition since the original ZnSe buffer layer surface lies about 1 nm below the surface of the ZnSe cap layer. Thick ZnSe Cap Layer. With Figure 5 we analyze the change in optical properties during the different stages of the hybrid epitaxial-colloidal growth process up to ZnSe layer thicknesses of 22 nm (growth temperature 240 °C). The colloidal nanocrystals used here are small, spherical CdSe(ZnS) core-shell nanocrystals (core radius R∼2 nm, shell 1 to 2 ML ZnS10) with a room-temperature emission at 520 nm before the ZnSe cap layer growth (black curve in Figure 5a). The purified colloidal solution of NCs in pyridine is deposited onto the 45 nm ZnSe buffer layer and annealed for 2 min at 240 °C, the future growth temperature, under ultrahigh vacuum conditions in the MBE chamber. The exposure of the colloidal nanocrystals to high-temperature did not affect the spectrum of the nanocrystals, but again decreases the PL efficiency by about 1 order of magnitude. Nano Lett., Vol. 5, No. 3, 2005

Figure 5. (a) Comparison of the photoluminescence spectra of colloidal, spherical CdSe/ZnS core-shell nanocrystals before (black, T ) 300 K) and after growth of a 22 nm ZnSe cap layer on top of the NCs on the ZnSe buffer (blue, T ) 10 K). The red and green curves show the PL spectrum of the NCs on ZnSe buffer after annealing at 240 °C for 2 min measured at T ) 10 K and 300 K. (b) Change in PL emission intensity as a function of laser illumination time (HeCd laser, 354 nm, 50 W/cm2) for CdSe/ZnS core-shell NCs (R∼2 nm) after annealing at 240 °C for 2 min (left axis) and after cap layer growth of a d ) 7 nm ZnSe cap on top of the NCs on the ZnSe buffer layer (right axis), T ) 300 K.

The spectral position and PL line shape are unchanged (red curve in Figure 5a), confirming the previous result from SEM, i.e., at the applied MEE growth temperature the NCs do not dissolve or transform into a homogeneous 2DCdZnSSe mixed crystal layer. The green and blue curves in Figure 5a show the PL spectrum of the thermally annealed nanocrystals on the ZnSe buffer layer before and after growth of a 22 nm ZnSe cap layer measured at T ) 10 K. In general and for reasons still under investigation, with increasing ZnSe cap layer thickness growing on the ZnSe buffer layer, the emission efficiency of the NCs decreases. Additionally, we observe in some samples a ZnSe-related trap emission centered around 550 nm with 80 nm fwhm (full width at half maximum). Because of the low PL efficiency at room temperature, samples with ZnSe cap layer thickness d >10 nm were examined at cryogenic temperatures. For the example shown in Figure 5a, the low-temperature spectra of the annealed CdSe nanocrystals before and after ZnSe cap layer growth have been found to be identical, both in energy position and spectral shape. In the overall set of samples investigated, Nano Lett., Vol. 5, No. 3, 2005

however, we sometimes observe a slight variation in the PL maximum before and after cap layer growth in a range of (10 nm. Unlike Stranski-Krastanow grown II-VI quantum dots, wet-chemically synthesized II-VI nanocrystals may show a relatively fast photodegradation during prolonged laser illumination. In hybrid epitaxial-colloidal quantum dot structures, the grown ZnSe cap layer might protect the embedded NDs and NRs against photochemical reactions. Figure 5b shows the change in emission intensity as a function of laser illumination time for CdSe(ZnS) core-shell nanocrystals exposed 2 min to a temperature of 240 °C before and after growth of a ZnSe cap layer (layer of thickness d>2R). We interpret the obvious degradation of the PL intensity of the thermally annealed nanocrystals as a consequence of partial thermal degradation of the passivating thin ZnS shell. After overgrowth, however, the PL intensity again increases with illumination time. Although the overgrowth with a thicker ZnSe cap layer results in a lower PL efficiency and deep trap formation, obviously, in a postgrowth process the PL efficiency can be improved by further laser annealing. We thus conclude that for the hybrid epitaxial-colloidal growth, post-growth annealing may be as important as the epitaxial growth itself. In Figure 6, the crystalline quality of a 22 nm ZnSe cap is analyzed by high-resolution X-ray diffraction. It shows rocking curves (ω-scan around the 004-ZnSe reflex) for ZnSe cap layers exposed to air and different solutions. The crystalline quality of the MBE grown 2D-ZnSe cap layer does not necessarily decrease when the overgrown material is not a layer of self-assembled quantum islands or a quantum well but, instead, consists of colloidal nanocrystals. We first determine in Figure 6 (left image) the crystalline quality of a reference sample of the ZnSe cap layer without growth interruption or exposure to other media and obtain a value for the fwhm of 6.8 arcsec. Exposure to air, a necessary process step to deposit the nanocrystal solution, only slightly influences the fwhm of the rocking curve, while the wetting with pyridine (without nanocrystals) increases the fwhm by a factor of 3.5. The most important result, however, is the analysis of the cap layer crystallinity after the overgrowth of nanocrystals. It turns out that neither shape nor size of the nanocrystals determines the cap layer crystallinity, but rather the nanocrystal surface concentration. For low nanocrystal concentrations, the value of the rocking curve fwhm is close to that of the pyridine-wetted ZnSe buffer. Thus, a cap layer quality can be achieved that is sufficiently high to continue the epitaxial process to obtain more complex structures if the nanocrystal concentration is small enough. With the HRTEM images in Figure 7 we give an overview of different, representative growth scenarios we have found for the growth of a ZnSe cap layer of thickness d>2R on top of a ZnSe buffer layer which is covered with CdSe nanorods at low density. A 12 nm layer of ZnSe is grown by migration-enhanced epitaxy (MEE) at 240 °C. Figure 7a shows a single CdSe nanorod, Figure 7b a nanocrystal-free region. In the nanocrystal-free region of the ZnSe buffer layer we find a perfect growth continuation of the ZnSe crystal 487

Figure 6. X-ray diffraction data (rocking curves) demonstrating the crystalline quality of an epitaxially grown 22 nm thick ZnSe cap layer grown on top of a ZnSe buffer layer which exposed before to air, pure pyridine, and colloidal solutions of nanodots and rods. The left picture is a reference structure for direct ZnSe growth without interruption. The numbers indicate the fwhm in arcsec.

lattice up to a final thickness of 12 nm. In the vicinity of a nanocrystal, the growth rate is very sensitive to the atomic structure of the lattice planes at the open nanocrystal surface. The nanorod is overgrown on the top (1) and left-hand side (2). The crystal orientation in the small triangular region on top of the nanorod (1) (Figure 7a) is epitaxial with respect to the ZnSe buffer layer, which demonstrates that CdSe nanocrystals can be covered, at least partially, with epitaxially grown ZnSe. Twinning has occurred on the left-hand side (2) of the nanocrystal, possibly due to the strain. The ZnSe overgrowth is impeded on the right side by a second nanorod that is only vaguely visible because it is not aligned along a low-index zone axis. Again, as in the case of thin ZnSe-cap layer growth, no gap is observed at the interface between the nanorod and the ZnSe buffer layer. A possible mechanism for the epitaxial relationship of the ZnSe on top of the nanocrystal with respect to the buffer layer is lateral epitaxial overgrowth from a neighboring epitaxial ZnSe region. Lateral epitaxial overgrowth of ZnSe on SiO2 patterned GaAs substrates has indeed been shown to occur using chemical vapor deposition.11 Modifications of the Optical Properties in Hybrid Epitaxial-Colloidal Semiconductor Nanostructures. An epitaxially grown matrix around a nanocrystal should effect emission efficiency, photostability, blinking and spectral jitter. Of greater interest, however, is the variation of the quantum confining barrier conditions. The electron and hole wave functions can be engineered in such a way that the optical transition dipole moment and the mechanism of coupling to phonons will change substantially. Since the goal of our studies is to provide a monolithic quantum dot heterostructure which unifies advantages of both MBE grown SK quantum dots and wet-chemically prepared CdSe nanoc488

rystals, for example, for applications in optical microcavities and single photon sources, we compare the photoluminescence dynamics of both types of capped quantum dotstructures. While the SK dots provide the advantage of short radiative lifetimes (i.e., larger optical transition dipole moment because of larger extension of the excitonic wave function) and the compatibility with existing process technology for such things as, quantum dot doped microcavities, these dots grow with high area density of >109 cm-2 which makes single dot coupling to photon modes difficult. With the NC-quantum dots, the area dot density can have arbitrary values, however, for an efficient exciton-photon interaction, the nanosecond lifetime indicates a smaller optical transition dipole moment. Any increase of the optical transition dipole moment in hybrid epitaxial-colloidal nanostructures toward the values known for CdSe/ZnSe SK-dots (and keeping low dot densities) is thus promising for the aforementioned application. Figure 8 is a first example of PL dynamics as modified by integration of the CdSe NRs in a hybrid epitaxial-colloidal structure. The time-resolved PL at low temperatures has been measured for the different types of CdSe/ZnSe semiconductor nanostructures: CdSe NRs of 35 nm length and 7 nm diameter on quartz substrate, hybrid epitaxial-colloidal CdSe/ZnSe nanostructures containing the same CdSe NRs embedded in 6 nm ZnSe, and fully MBEgrown Stranski-Krastanow (SK) CdSe/ZnSe quantum dots (2 nm in height, 10 nm in width). The SK dots are grown on the same ZnSe buffer layer in the same MBE chamber as used for the hybrid growth technique.12 The PL decay times of the colloidal CdSe nanorods of some nanoseconds and the decay times of the SK dots of some hundreds of picoseconds reproduce well the known data sets for these types of nanostructures.5,13 The PL decay Nano Lett., Vol. 5, No. 3, 2005

Figure 7. (a) HRTEM image of colloidal CdSe nanorods overgrown with a 12 nm layer of ZnSe by migration-enhanced epitaxy (MEE) at 240 °C. The atomic structure of the open lattice planes at the (wurtzite) nanorod surface control the growth rate of the ZnSe cap layer around the NR. The small triangular region on top (1) is free of defects while twinning has occurred on the left-hand side (2). The ZnSe overgrowth is impeded on the right side by another nanorod. A good bonding between the nanorod and the ZnSe buffer is observed. (b) Cross-section HRTEM image along the [1-10]zone axis of a 12 nm layer of ZnSe grown on the ZnSe buffer layer by migration-enhanced epitaxy (MEE) at 240 °C. Shown here is the interface region between the nanocrystals where we find an almost perfect pseudomorphic overgrowth of the ZnSe cap layer.

of the hybrid structure is strongly nonexponential and characterized by a fast initial decay, which we assign to a capture in ZnSe-related traps, and a second component slightly longer than the decay time of the SK dots. The origin of PL decay modification of CdSe NRs in the hybrid structure can be attributed to various effects including changes in the optical transition dipole moment due to the extension of electron or hole wave functions into the surrounding barrier, reorganization of the nanorod surface, and changes in defect density during the thermal annealing, or a new, combined, decay time consisting of the radiative lifetime of the nanorod and a transfer time between the NR and the substrate. To explore this modification is an interesting future challenge. Summarizing our results, we presented a growth technique that combines wet-chemical growth and molecular beam epitaxy (MBE) to create complex semiconductor nanostrucNano Lett., Vol. 5, No. 3, 2005

Figure 8. Time-resolved photoluminescence of different types of CdSe/ZnSe semiconductor nanostructures: CdSe nanorods (35 nm length, 7 nm diameter) on quartz substrate (triangles), hybrid epitaxial-colloidal CdSe/ZnSe nanostructures containing the same CdSe NRs embedded in 6 nm ZnSe (squares), MBE-grown Stranski-Krastanow CdSe/ZnSe quantum dots (2 nm in height, 10 nm in width) (circles). The experiments have been performed at T ) 15 K. The exciting laser pulse is a 120 fs pulse of a frequencydoubled Ti:Sa laser. Upper panel: PL detection using single-photon counting technique (time-resolution 500 ps). Lower panel: detection using a streak camera (time-resolution 20 ps). The detection energy is tuned in the spectral maximum of the emission spectrum.

tures with nanocrystals as active optical material. The obtained results show that wet-chemically prepared semiconductor nanocrystals can be incorporated in an epitaxally grown crystalline cap layer. As an example we chose colloidal CdSe or CdSe(ZnS) NCs in MBE-grown ZnSe. Transmission electron microscopy and X-ray diffractometry show that the ZnSe cap layer is of high crystalline quality and has adapted the crystalline structure of the ZnSe buffer layer, even though this was previously wetted by the NCs dissolved in pyridine. The best crystalline quality of the ZnSe cap is obtained for low nanocrystal density, while the overgrowth of close-packed NC layers result in partly polycrystalline ZnSe caps. The nanocrystals are stable at the elevated temperatures that are necessary for epitaxial growth in molecular beam epitaxy. They retain their original shape and emission spectrum. With increasing ZnSe cap layer thickness a defect emission around 550 nm develops which might be suppressed by post-annealing processes. In the vicinity of a nanocrystal, the overgrowth and the growth rate is very sensitive to the atomic structure of the lattice planes at the open nanocrystal surface that allow for control of the overgrowth, i.e., to promote or to terminate directional cap 489

layer growth. In contrast to CdSe/ZnSe quantum dot heterostructures obtained in a closed-cycle MBE process based on Stranski-Krastanow growth of CdSe on top of the ZnSe buffer layer, we did not observe the formation of a 2D wetting layer in our samples. Future tasks we plan to undertake are the minimization of quantum dot density to achieve mean nanocrystal separations larger than 1 µm in space and the defining of suited post-growth annealing routines in order to perfectly link the crystal lattices of the NCs to the surrounding matrix. In the presented study we use wurtzite-type nanocrystals (CdSe nanorods) which, however, do not exhibit a crystal lattice that is optimum for a complete epitaxial overgrowth on a cubic ZnSe buffer layer. This choice was determined by the need to prove the existence (and survival) of colloidal nanocrystals in an MBEprocess which requires nanocrystals that can be easily identified by the available structural characterization methods. Now we can also consider growing, for example, hybrid epitaxial-colloidal nanostructures with Mn-doped ZnSe nanocrystals or InAs nanocrystals in GaAs (although, for the III-V epitaxy, the higher growth temperature again raises the problem of thermal stability of the NCs). Nevertheless, the epitaxial overgrowth of colloidal nanocrystals has the great potential of compatibility with the epitaxial growth of Bragg mirrors, waveguide structures, and pin structures, allowing for the integration of flexible colloidal chemistry with traditional semiconductor processing. Acknowledgment. We thank the DFG (Graduiertenkolleg GRK 726), the EU Project HPRN-CT-2002-00298, the German-American Fulbright Commission and INTAS 01/ 2331 for financial support. E.H. would also like to thank Dr. Kenith Meissner and Danielle Irving for their support in the production of CdSe(ZnS) core-shell nanocrystals. Supporting Information Available: Images showing reciprocal space maps around the (224) reflex and the insitu RHEED pattern taken during the ZnSe cap layer growth after nanocrystal deposition. This material is available free of charge via the Internet at http://pubs.acs.org.

490

References (1) For an overview see J. LightwaVe Technol. 1999, 17. Special Issue about photonic structures. (2) Bimberg, D.; Grundmann, M.; Ledentsov, N. N.; Quantum Dot Heterostructures; Wiley: New York, 2001. (3) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, 8706. (4) Peng, X.; Manna, L.; Yang, W.; Wickham, J.; Scher, E.; Kadavanich, A.; Alivisatos, A. P. Nature 2000, 404, 59. (5) Landolt-Bo¨rnstein New Series III/34C Optical Properties of Semiconductor Nanostructures; Klingshirn, C., Ed.; Springer-Verlag: Berlin, 2004. Woggon, U.; Gaponenko, S. V.; Chapter 5.5. II-VI Semiconductor Quantum Dots: Nanocrystals, and Chapter 5.6. IIVI Semiconductor Quantum Dots: Self-organized, epitaxially grown nanostructures (6) Danek, M.; Jensen, K. F.; Murray, C. B.; Bawendi, M. G. Appl. Phys. Lett. 1994, 65, 2796. (7) Leonardi, K.; Heinke, H.; Ohkawa, K.; Hommel, D.; Selke, H.; Gindele, F.; Woggon, U. Appl. Phys. Lett. 1997, 71, 1510. Gindele, F.; Woggon, U.; Langbein, W.; Hvam, J. M.; Leonardi, K.; Hommel, D.; Selke, H. Phys. ReV. B 1999, 60, 8773. Schikora, D.; Schwedhelm, S.; As, D. J.; Lischka, K.; Litvinov, D.; Rosenauer, A.; Gerthsen, D.; Strassburg, M.; Hoffmann, A.; Bimberg, D. Appl. Phys. Lett. 2000, 76, 1. Peranio, N.; Rosenauer, A.; Gerthsen, D.; Sorokin, S. V.; Sedova, I. V.; Ivanov, S. V. Phys. ReV. B 2000, 61, 16015. Litvinov, D.; Rosenauer, A.; Gerthsen, D.; Kratzert, P.; Henneberger, K. Appl. Phys. Lett. 2002, 81, 640. (8) Murray, C. B.; Sun, S.; Gaschler, W.; Doyle, H.; Betley, T. A.; Kagan, C. R. IBM J. Res. DeV. 2001, 45, 47. (9) Peng, X.; Schlamp, M. C.; Kadavanich, A. V.; Alivisatos, A. P. J. Am. Chem. Soc. 1997, 119, 7019. (10) The ZnS shell is very thin (1 to 2 monolayers). Since the thickness of the ZnS shell is not larger than the ZnS-lattice constant, the ZnS is present here on a molecular scale. If the ZnS shell plays a role at all in the ZnSe cap process, then it will most likely help to reduce the strain between CdSe and ZnSe and thus to decrease the defect density. In the MBE growth of quantum wells, ZnS is sometimes used for strain compensation between CdSe and ZnSe layers since its lattice constant of 0.5406 nm is smaller than that of both CdSe (0.60813 nm) and ZnSe (0.5668 nm). See, e.g., Engelhardt, R.; Pohl, U. W.; Bimberg, D.; Litvinov, D.; Rosenauer, A.; Gerthsen, D. J. Appl. Phys. 1999, 86, 5578. (11) Mauk, M. G.; Feyock, B. W. J. Cryst. Growth 2000, 211, 73. (12) Schikora, D.; Schwedhelm, S.; As, D. J.; Lischka, K.; Litvinov, D.; Rosenauer, A.; Gerthsen, D.; Strassburg, M.; Hoffmann, A.; Bimberg, D. Appl. Phys. Lett. 2000, 76, 418. (13) Patton, B.; Langbein, W.; Woggon, U. Phys. ReV. B 2003, 68, 12 5316.

NL0480870

Nano Lett., Vol. 5, No. 3, 2005