Research in quantum dot lasers in order to implement all optical neurons for high speed processing systems

Postgraduate Thesis uoadl:2876294 252 Read counter

Unit:
Κατεύθυνση Σχεδίασης Ολοκληρωμένων Κυκλωμάτων
Πληροφορική
Deposit date:
2019-06-19
Year:
2019
Author:
Skontranis Menelaos
Supervisors info:
Χάρης Μεσαριτάκης, Αναπληρωτής Καθηγητής, Τμήμα Μηχανικών, Πληροφοριακών και Επικοινωνιακών Συστημάτων, Πανεπιστήμιο Αιγαίου
Original Title:
Μελέτη ολοκληρωμένων λέιζερ κβαντικών τελειών με σκοπό την υλοποίηση αμιγώς οπτικών νευρώνων για υψίρρυθμα συστήματα επεξεργασίας
Languages:
Greek
Translated title:
Research in quantum dot lasers in order to implement all optical neurons for high speed processing systems
Summary:
This work is focused on the theoretical study of excitability in quantum dot lasers in order to simulate a biological neuron. The constantly growing needs for high speed processors and greater bandwidth, in combination with the saturating performance of electronic devices has driven the scientific society to study and develop new bio-inspired data processing systems. The basic unit of a biological neural network is the neuron. Taking into account, the inherited advantages of the quatum dot lasers and the recent studies that confirmed their functional similarities with their biological counterparts, quantum dot lasers are the best candidate for the realization of an artificial neuron.
The struct simulated in this work consists of two single mode single section quantum dot lasers connected in Master-Slave configuration. The two lasers have stable bias conditions, which forces them to emit only from the Ground State (GS). The Master ’s optical field is injected in the Slave’s field, disturbing its constant function and thus changing the Slave ’s output field. The two lasers have a slightly differation df in the emiting frequency, while the amplitude of the injected field is controlled by the injection coefficient rinj. For the simulation of the struct a MATLAB code was developed, based on the multipopulational rate equation model.
Firstly, a free running quantum dot laser was simulated, in which the increase of the pump caused the stimulated emission from the GS. Further increment of the pump current caused the start of stimulated emission from the Excited State (ES). Greater pump current caused greater field amplitude in both states. Most importantly if the pump current increased more than the ES threshold then the GS field quenched and started to decrease its amplitude. This effect was attributed in the asymmetry of transport times of electrons and holes.
For the case of the two quantum dot lasers the injected field can radically alter the Slave ‘s output. Specifically, the Master ‘s GS injected field, dependind on df and rinj values, can increase the amplitude of the Slave ‘s GS field, alter its output from constant to periodic or cause the start of the ES stimulated emission with simulataneous GS quenching. The start of the ES stimulated emission is of high importace since it can be used for the realization of the inhibitory neuron. Lastly, the ES output shows periodic, constant or pulsating behavior depending on the values of df and rinj.
In the final section, the struct was polarized with constant pump current, df and rinj. With simulataneous injection of negative rinj pulses the injected field is altered causing the generation of sharp ES pulses. At the same time negative GS pulses are observed. Further, by altering the frequency, amplitude and width of the negative rinj pulses, fire-and-integrate behavior has been confirmed which is one of the paramount characteristics of the biological neurons.
Main subject category:
Technology - Computer science
Keywords:
quantum dot lsers, bio-inspired structs, otpical neurons
Index:
Yes
Number of index pages:
6
Contains images:
Yes
Number of references:
49
Number of pages:
116
MSc_Dissertation_Menelaos_Skontranis.pdf (5 MB) Open in new window