ISBN 9783843926362

978-3-8439-2636-2, Reihe Physik

Aune Koitmäe
Hybrid Neural Networks on Semiconductor Microtube Arrays

161 Seiten, Dissertation Universität Hamburg (2016), Softcover, A5

Zusammenfassung / Abstract

In this thesis a well-defined large scale hybrid neural network on semiconductor microtube arrays is realized. A quantum well is embedded in the multi-layered wall of the microtube as a special feature giving it optically active characteristics. Furthermore, a novel method of optical detection of action potentials via optically active microtubes is proposed and supported with experiments and simulations on artificial axons.

In the first part of the thesis, the fundamentals of neuron guiding and the mechanism of how neurons communicate with each other, called action potential, are explained.

In the second part the fabrication of the microtubes, based on the selfrolling of lattice mismatched layers, is described. In addition, three methods of bio-compatibilization of the samples are pointed out. To protect the cells from the toxic compound of the substrate, arsenic, the samples are coated with different biocompatible materials. To confine the adhesion areas of the neurons, the samples are coated with arrays of a specific amino-acid.

The third part presents the results of neuron guiding on three different microtube arrays: stripe, cornflower and chess. The best neuron guiding results were achieved utilizing the combination of chess arrays by printing an adhesive amino acid selectively in the entrance-areas of the microtubes.

In the fourth part a novel method of action potential detection via optically active microtubes is proposed and as a proof of concept measurements with artificial axons in the microtubes are carried out. Simulations mimicking action potential propagation in the microtube show a red shift in the emission energy of the quantum well in comparison to the case where no action potential propagates along the microtube’s wall.