- Surface self-organization with buried dislocation networks
D. Buttard, J. Eymery, P. Gentile, F. Leroy, K. Rousseau, J.L. Rouvière, F. Fournel CEA LETI

To pursue the race to miniaturization of electronic circuits, self-assembled configurations rather than artificially drawn structures are expected to play a major role in the design of nanometric devices. Many methods of self-assembling have been proposed ranging from epitaxy to colloïds but the majority of them suffer invariably from a statistical yield in the control of size and spatial position of the nanostructures. The research we have developed opens a way to control the lateral ordering of nanostructures at the scale of a silicon wafer. The lateral self-organization can be obtained by the preferential patterning of a solid surface with local strain or morphology created by a periodic array of buried stressors. This array is formed artificially by the dislocations which appear at the interface of two identical silicon (001) crystals bonded with some crystallographic disorientations. As it will be shown in the next contribution, such bonding creates two dislocation networks according to the flexion and rotation of the two crystals : the first one is constituted of parallel mixed dislocation lines, whereas the second is a square network of screw dislocations. The general features of these arrays have been studied by Grazing Incidence X-Ray Diffraction (GIXRD) at ESRF [1]. Fig. 1 shows an example of the very good periodicity of the lateral atomic-displacement that we have modeled with kinematical calculations [1].

 

The propagation of the strain field through the bonded layer up to the surface, also measured by GIXRD, has been checked by continuum elasticity calculations. Experimentally, the surface-strain is increased by thinning the bonded layer by sacrificial thermal oxidation to less than 10 nm. The clean and crystalline surface is then obtained by an HF deoxidation. Flexion and rotation misalignment give rise respectively to long-range undulations and short-range embossing than is enhanced by strain-driven overetching. This surface is a template for the Ge-growth as shown in Fig. 2. After annealing, quantum dots can be ordered with a short-range correlation with a fourfold symmetry related to the buried screw-dislocation networks [2]. When two “twin” surfaces of Si(001) produced by the splitting of a single wafer are bonded together, we have only a relative in-plane rotation of the two crystal [3]. For a small rotation angle, the periodicity of the network is inversely proportional to the rotation angle. We achieve to control the rotation angle with a precision of 0.01° with an original patented method [3]. So that, it is possible to tune the periodicity of the dislocation network from few nanometers (<10° of rotation) up to 2.2 µm (~0.01° of rotation). Moreover, by bonding twin surfaces at low rotation angle, there is almost no other defects at the interface, except some residual steps.

 

To directly transfer the fourfold symmetry of the interface to the surface morphology, we use an etching solution sensitive to the strain with a mixture of oxidant and deoxidant chemicals. The reaction rate R can be written as the product of a mobility term depending on the stress tensor by a non-linear kinetic law depending on the etching driving force. This driving force takes into account the free energy between the solid and liquid phases of silicon, the local elastic energy and an energy related to the curvature of the surface. The determination of the local variation of R is quite difficult because the stress modifications induced by the surface morphology evolution during the etching must be taken into account. But it can be shown with simple considerations that a small reaction rate can be sensitive to very low variation of energy (below thermal energy), so that the spatial frequency corresponding to the dislocation periodicity will develop a correlated roughness at the surface. An example of nanopatterning of the Si surface is shown in Fig. 3 with a scanning tunneling microscope (STM) measurement. The patterning consists in a square array of silicon nanostructures, regularly spaced, with a lateral period of 25 nm consistent with the bonding angle of 0.88°, and a peak-to-peak amplitude of about 5 nm. The bonded layer thickness is decreased from its initial value (110 nm) to less than 30 nm. The regularity of the buried dislocation network is preserved as indicated by GIXRD. In this sample, the thinning procedure is obtained with a mixture of fluorhydric, nitric and acetic acid (so-called Dash-etch), and similar morphologies are reproduced with another solution (chromium oxide and fluorhydric acid). The bonding process and the chemical etching may be performed with full 300 mm wafers, so that this method is fully compatible with very large scale integration process.

 

The interest of this surface patterning is illustrated in Fig. 4 by growing 9 Å of germanium by molecular beam epitaxy at 490 °C (growth rate 0.7 Å/min) in the STM set-up. A very original growth occurs with the localization of flat and rectangular Ge islands on top of the Si bumps, when the same deposition gives standard “hemispherical” islands (16 nm in diameter) on a flat substrate. This behaviour, may be explained by the Ge dots elastic relaxation due to the shape and size of the Si patterns. The example of Ge quantum dots growth on nanopatterned Si surfaces opens the way to study the dependence of the surface curvature and strain. Such templates are very convenient because the lateral periodicity and the depth of the patterns can be changed by varying the bonding or etching conditions. The pattern regularity is expected to narrow the size distribution, and therefore the physical properties of the nano-objects. Surface stress and morphology engineering by wafer bonding and chemical etching constitutes a general approach to assembly nano-materials in well-defined functional networks, with the high density required by applications. This approach is not restricted to semiconductor materials and is also potentially useful for organizing organic as well as non-organic nanostructures.

 

[1] J. Eymery et al., Phys. Rev. B 65, 165337 (2002). [2] F. Leroy, J. Eymery, P. Gentile, and F. Fournel, Appl. Phys. Lett. 80, 3078 (2002). [3] F. Fournel et al., Appl. Phys. Lett. 80, 793 (2002).

 

Last update : 10/19 2016 (74)

 

Retour en haut