1.2.3.1 Epitaxial QDs
The first epitaxial QDs experimentally observed were naturally formed in the unavoid- able monolayer defects in thin QWs [40]. However, these QDs are not suitable for most of the above-mentioned applications due to their weak confinement energy. The most popular method for fabricating QDs is based on the Stranski Krastanov (SK) technique. The fabrication relies on the heteroepitaxy of a highly strained layer (for example InAs over GaAs) [41]. When the grown thickness rises above a threshold, islands of InGaAs nucleate at random positions. These islands are the SK QDs as shown on the AFM picture of Fig. 1.5(a). To prevent the interaction of QDs with surface states, this layer is then covered by a cap of GaAs. These QDs have heights typically between 1 and 10 nm and a horizontal size between 10 and 100nm [27] and are randomly distributed over the GaAs surface. The thin layer from which the QD are formed is called the wetting layer. Such QDs were fabricated with emission wavelengths ranging from 850nm to telecom wavelengths at 1.3μm. They are easy to fabricate and exhibit sharp emission linewidths close to their lifetime limit (a fewμeV). However, the size of these QDs is not controlled precisely and induces an inhomogeneous broadening of QD lines usually around 50meV [42].
These QDs can be integrated within PhCs or other optical nanostructures. How- ever, it is difficult to control the QD position in the photonic nanostructure because of their random nucleation site. In typical experiments, many structures are fabricated and the sample is scanned by photoluminescence measurements to find structures in which the QD is efficiently coupled to the optical modes. An alternative approach is to align the nanostructure directly over a selected QD [43]. However, these approaches are not applicable for integrated photonics in which arrays of many QDs would need to be placed at precise locations in photonic circuits.
1.2.3.2 Site-controlled QDs
Several approaches were proposed to fabricate arrays of position controlled QDs. These usually rely on structuring the growth surface to control their position of nucle- ation.
Typical methods include the fabrication of dielectric mask aperture to etch nano- holes in the substrate [44] from which average optical linewidths down to 156μeV were obtained [45]. In a different technique [46], several layers were grown to obtain stacked QDs, in which the nucleation sites of the QDs were determined by the strain induced by the presence of the lower QDs. In the upper layers, the QD quality was improved with linewidths down to 110μeV. In a different technique [47], the nanorecesses were etched directly inside the growth chamber to ensure minimal contamination. Record linewidths for such site controlled QDs were obtained down to 43μeV [48] (Fig. 1.5 (b))
(a) (b) (c) (d)
Figure 1.5 – (a) Atomic force microscopy image of an SK QDs; (b) scanning electron microscope image of an array of nanoholes used to grow site controlled QDs; (c) Atomic force microscopy image of an array of individual QDs; (d) Distribution of linewidths for site-controlled QDs; (a) Reprinted with permission from [50]. Copyright (2001) with the permission of AIP Publishing; (b) Reprinted from Publication 1.5, Copyright (2011), with permission from Elsevier. (c-d) Reprinted with permission from [49]. Copyright (2013) American Chemical Society.
and more recently [49] down to 6μeV (Fig. 1.5 (c-d). However, when using techniques based on the preferential nucleation of SK QDs, zero or more than one QD may be formed [45] which compromises their site control.
1.2.3.3 Pyramidal QDs
An alternative technique for growing site controlled QDs is based on the growth of In- GaAs/GaAs or AlGaAs/GaAs QDs in arrays of pyramidal recesses. The QDs are formed at the apex of each pyramid by the interplay of growth rate anisotropy and capillarity [51]. They are called pyramidal QDs. Because the formation of these QDs is not driven by strain, only one QD is grown in each pyramid. Besides, no wetting layer is formed during the growth of QDs, although under certain conditions, QWRs are formed at the edges of the pyramids [52]. These QDs offers promising properties, including repro- ducible optical spectra [53] and relatively narrow linewidths down to the resolution limit of 20μeV. The inhomogeneous broadening is typically around 5-10meV, and a record value of 1.4meV [54] was observed. Thanks to the three fold symmetry of the crystal and the pyramids permitted by the symmetric (111)B orientation of the growth plane, highly symmetric QDs can be formed in which the FSS is lower than in SK QDs [55, 56, 57].
1.2.3.4 The perfect QD
All these experimental efforts aim at improving the quality of QDs to make them suitable for quantum optics experiments. The desirable properties of QDs depends on their application. We propose here a non exhaustive list of these properties [59].
Some properties aim at obtaining a high quality optical spectra. QDs exhibiting a large confinement energy are more easily protected from unwanted interaction
(a)
(b)
(c) (d) (e)
Figure 1.6 – (a) Atomic force microscopy image of one pyramidal QD intentionally not planarized; (b) 3D schematics of the growth in pyramidal QDs (c) Atomic force microscopy image of an array of pyramidal QDs; (d) Spectra of individual pyramidal QDs; (e) Distribution of the resonance energies of excitons in pyramidal QDs; (a-c) Reprinted with permission from [58], Copyright (2008), AIP Publishing LLC; (e-f ) Reprinted with permission from [54] (2010) Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.
with residual defect states. Besides, deeply confined QDs are optically active at much higher temperatures. Optimally, QDs would therefore be confined deep enough for room temperature single photon emission as was demonstrated with GaN QDs [60].
Ideal QDs would also exhibit perfectly reproducible spectra including a low inho- mogeneous broadening of exciton lines [54]. Although a vanishing inhomogeneous broadening would be appreciated, nanofabrication limitations will probably pre- vent reducing the inhomogeneous broadening much below 1meV. This issue can be circumvented by controlling the wavelength of each QD individually at the cost of additional complexity, for example by applying a strain [61], via the Stark effect [62] or the Zeeman effect [63].
High quality QDs would also exhibit narrow lifetime limited linewidths as was observed in SK QDs, and a high dipole strength [27] to attain a strong light-matter interactions and a bright emission.
Other desirable properties are more focused on fabricating QDs that are easily in- tegrated into scalable technologies such as the position control [64] and the possibility to fabricate QD emitting over a broad energy range [65].
Last but not least, QDs are expected to play a role in quantum technologies. The two main quantum properties expected from ideal QDs are the emission of single photons with a high degree of indistinguishability [66].
Although each of these requirements was the subject of separate publications indicating promising result, achieving all of them simultaneously and reproducibly on a large array of QDs is extremely challenging. All these aspects will be considered in this thesis except for the photons indistinguishability.