PhD

Neuromorphic photonics is a new field of research at the heart of recent progresses in analog computation and machine learning. Its goal is to investigate new ways to process optical information or to compute using brain-inspired concepts.

We propose to study coupled spiking photonic nodes in order to implement photonic artificial neural networks. Each spiking node is materialized by a micropillar laser with integrated saturable absorber, whose neuromimetic properties have already been explored in the team. In biological neurons, information is coded with spikes (electrical pulses) which are excited in an all-or-none fashion provided input stimuli to the neuron soma exceed a given threshold. This generic property is called excitability and has been demonstrated in micropillar lasers. Though, the optical spikes emitted by the latter are more than one millions times shorter in duration than biological action potentials. Hence, photonic neurons could in principle be interesting to build ultrafast artificial neural networks with low power consumption. The computing capability of optical neurons are enforced by the property of temporal summation also already demonstrated by us in micropillar lasers, and which provides universal computation capability.

The main objective of the thesis will thus be to take advantage of these neuromimetic properties to fabricate and implement neuromorphic computing architectures and demonstrate ultrafast analog computation. The thesis will take place at the C2N (Palaiseau) which hosts a first-class nanofabrication facility. The samples will be fabricated in the C2N clean-room. This study will take place in the framework of the recently funded ANR ANACONDA project, with one partner in FEMTO-ST (Besançon) and one in CERCO (Toulouse).

The work will consist in the design and fabrication of the samples, running the experimental setup, modelling and analysis of the results, and interacting with the project's partners.
The applicant is expected have a strong motivation, be rigorous and curious, with strong background in physics, optics, laser physics, semiconductor physics, and possibly semiconductor lasers and/or nonlinear dynamics and/or machine learning.

More information:
https://toniq.c2n.universite-paris-saclay.fr/en/activities/smila/neuromimetic-photonics/

Work Context

The PhD will be hosted at the C2N in Palaiseau. C2N is a joint CNRS/Université Paris-Saclay laboratory which hosts a first-class clean-room facility and whose research activities span basic to applied research in the areas covered by 4 Departments: Photonics, Nanoelectronics, Materials, Microsystems and Nano-Bio Fluidics. He/She will work in the Photonics Department in the TONIQ team whose activities concern quantum and nonlinear micro-nanophotonics. The PhD will be supervised by Dr Sylvain Barbay, CNRS senior researcher.

PhD

La photonique neuromorphique est un nouveau champ d'étude au cœur de progrès récents en calcul analogique et en apprentissage artificiel. Son but est l'étude de nouvelles voies pour le traitement optique de l'information et le calcul utilisant des concepts neuro-inspirés.

Nous proposons d'étudier des nœuds photoniques impulsionnels afin d'implémenter des réseaux de neurones artificiels. Chaque nœud impulsionnel est matérialisé par un micropilier laser avec absorbant saturable intégré dont les propriétés neuromimétiques ont été explorés dans l'équipe. Dans les neurones biologiques, l'information est codée avec des impulsions (électriques) qui sont excitées de façon tout-ou-rien pourvu que des stimuli d'entrées qui lui sont appliqués dépassent un certain seuil. Cette propriété générique est appelée excitabilité et a été démontrée dans des micropiliers lasers. Cependant, les impulsions optiques émises par ces derniers sont plus de un million de fois plus rapides que les potentiels d'action générés par les neurones biologiques. Ainsi, les neurones photoniques pourraient être potentiellement intéressants pour fabriquer des réseaux de neurones artificiels ultrarapides et de basse consommation. Les capacités de calcul des neurones optiques sont liées à la propriété de sommation temporelle que nous avons déjà démontrée, et qui leurs fournissent la capacité de calcul universel.

L'objectif principal de la thèse sera donc d'utiliser les propriétés neuromimétiques de ces lasers pour fabriquer et implémenter des architectures de calcul neuromorphiques et démontrer des calculs analogique ultrarapides. La thèse sera effectuée au C2N (Palaiseau) qui accueille une centrale de technologie de premier cercle. Les échantillons seront fabriqués dans la salle blanche du C2N. Cette étude s'insère dans le cadre du projet ANR ANACONDA récemment financé, dont les partenaires sont FEMTO-ST (Besançon) et CERCO (Toulouse).

Le travail consistera en l'étude et la fabrication des échantillons, le montage des bancs optiques, la modélisation et l'analyse des résultats, et l'interaction avec les partenaires du projet.
Le/La candidat-e devra montrer un forte motivation, être curieux et rigoureux, avec une solide formation en physique, optique, physique des lasers, physique des semiconducteurs. Des connaissance en lasers à semiconducteurs et/ou dynamique non-linéaire et/ou apprentissage artificiel seront un plus.

Plus d'information :
https://toniq.c2n.universite-paris-saclay.fr/fr/activities/smila/neuromimetic-photonics/

Contexte de travail

Le/La doctorant-e travaillera au C2N à Palaiseau. Le C2N est un laboratoire CNRS/Université Paris-Saclay qui possède en son sein une centrale de nanotechnologies de premier cercle et dont les activités de recherche vont de la physique fondamentale à la physique appliquée dans les domaines couverts par 4 Départements : Photonique, Nanoélectronique, Matériaux, Microsystèmes et Nano-Bio Fluidique. Le/La candidat-e travaillera dans le Département Photonique dans l'équipe TONIQ dont les activités concernent la micro-nanophotonique non-linéaire et quantique. Le/La doctorant-e sera encadrée par Sylvain Barbay, Directeur de Recherche au CNRS.

 

PhD

Centre de Nanosciences et nanotechnologies

 UMR 9001 du CNRS,

université Paris-Sud, université Paris-Saclay

91 405 Orsay – France

  

PhD research proposal

All-Dielectric Metamaterials for Terahertz metadevices

 Éric Akmansoy 

Département Photonique

__________________ 

 

General framework

Metamaterials have opened a new field in physics and engineering. Indeed, these artificial structured materials give rise to unnatural fascinating phenomena such as negative index, sub-wavelength focusing and cloaking. Metamaterials also exhibit near-zero refractive index [1]. These open a broad range of applications, from the microwave to the optical frequency domain. Metamaterials have now evolved towards the implementation of optical components[2]. 

We consider All-Dielectric Metamaterials which are the promising alternative to metallic metamaterials, because they do not suffer from ohmic losses and consequently benefit of low energy dissipation and because they are of simple geometry. They consist of high permittivity dielectric resonators involving Mie resonances. We have experimentally demonstrated negative effective permeability and/or permittivity by the means of all-dielectric metamaterials [3]. Previously, we have also demonstrated a negative index all-dielectric metamaterial [4]. Recently, we have numerically demonstrated a metadevice, namely, a metalens that focuses an incident plane wave, is less than one and a half wavelength thick. Its focal length is only a few wavelength and the spot in the focal plane is diffraction - limited.[5]. We also study role of the coupling of the modes of Mie resonances in an all-dielectric metamaterial so as to achieve negative index at terahertz frequencies (see fig. 1)[6].

Work Plan

During this thesis, all dielectric metamaterials will be designed, fabricated and characterized for the terahertz range. Negative index will be studied in this frequency domain, which is a challenge. Various devices which will be investigated such as flat lens, gradient devices, etc. This work takes place within a group of scientists of different disciplines (chemists, material scientists, physicists). 

 

 

Fig. 1: Spatial mode coupling : frequency of the first two modes of Mie resonances in function of the distance  between two resonators which is half the lattice period . The shaded area corresponds to negative value of the effective index . It evidences the mode degeneracy [6].

Bibliography

[1]    I. Liberal and N. Engheta, “Near-zero refractive index photonics,”  Nature Photonics, vol. 11, pp. 149 EP –, 03 2017.

[2]    N. I. Zheludev and Y. S. Kivshar, “From metamaterials to metadevices,”  Nat Mater, vol. 11, pp. 917–924, 11 2012.

[3]    T. Lepetit, E. Akmansoy, and J.-P. Ganne, “Experimental evidence of resonant effective permittivity in a dielectric metamaterial,” Journal of Applied Physics, vol. 109, no. 2, p. 023115, 2011.

[4]    T. Lepetit, É. Akmansoy, and J.-P. Ganne, “Experimental measurement of negative index in an all-dielectric metamaterial,” Applied Physics Letters, vol. 95, no. 12, p. 121101 (, 2009.

[5]    F. Gaufillet, S. Marcellin, and E. Akmansoy, “Dielectric metamaterial-based gradient index lens in the terahertz frequency range,” IEEE J Sel Top Quant, vol. 10. 1109/JSTQE. 2016. 2633825, 2017.

[6]    S. Marcellin and É. Akmansoy, “Negative index and mode coupling in all-dielectric metamaterials at terahertz frequencies,” in 2015 9th International Congress on Advanced Electromagnetic Materials in Microwaves and Optics (Metamaterials), pp. 4–6, Sept 2015.

PhD

  

Centre de Nanosciences et nanotechnologies

UMR 9001 du CNRS,

université Paris-Sud, université Paris-Saclay

91 405 Orsay – France

  

PhD research proposal

Graded Photonic Crystals Devices for Graded Index Optics

 Éric Akmansoy 

Département Photonique

__________________ 

 

General framework

On the one hand, Graded Photonic Crystals (GPC) allow to efficiently control the flow of light thanks to the shape of their photonic bands, which we verified by demonstrating a “photonic mirage” at wavelength scale[1]. In the other hand, GRaded INdex (GRIN) optics is undergoing a renewal because it allows to downsize optical systems, opening a new way to optical design[2]. But GRIN optics was limited by a lack of easy to implement fabrication techniques. Nanotechnology enables to efficiently fabricate photonic crystals, which we will implement to fabricate GPC for GRIN optics. 

We are designing, numerically simulating and characterising GPC devices (see fig.1)[3, 4]. More specifically, we have conceived a GRIN flat lens which we have characterised in the microwave[5]. Recently, we have designed and simulated a Luneburg lens[6] (see fig.1). We conceive thin and thick flat GRIN lenses[7]. For the sake, we engineer the equi-frequency curves which convey the dispersion properties of photonic crystals. Our aim is to go further up to the optical domain. The experimental evidence of such devices is an end in itself because it is a technological challenge to fabricate such GPC. Nevertheless, we will investigate the great variety of applications of GPC flat lens, which concerns Photonic Integrated Circuits, Lab on Chip, optical pumping of organic materials, OLED, biophotonics, fluorescence of cells, etc.

PhD work plan

GPC devices will be designed, fabricated and characterised at the Centre de Nanosciences et Nanotechnologies (C2N). This work will concern numerical simulation and optical design, nanotechnological processes and optical characterisation. These devices will operate at telecommunication wavelengths (1,55 μm). The Silicium On Insulator (SOI) channel of the C2N will be involved, because “state of the art" photonic crystal -based devices have already been fabricated in its clean-room [8]. These devices require the technological mastery of the position, of the size and of the shape of the patterns, which is adapted for the realisation of GPC devices. The size of the patterns of these latter is of several tens of nanometers, which will be challenging. The devices will be also characterised at the C2N. GPC devices operating in the near infrared (NIR) spectrum will also investigated (from 1 to 20μm) because, in this domain, lies a lot of molecular signatures, which is fruitful for NIR spectroscopy, chemical sensing, polluant detection, etc. 

      

 

Fig. 1: Right : Focusing by a GPC GRIN flat lens[3] ; left : focusing by a Luneburg lens[6] 

Bibliography

[1]    Éric Akmansoy, Emmanuel Centeno, Kevin Vynck, David Cassagne, and Jean-Michel Lourtioz, Appl. Phys. Lett. 92, 133501 (2008) 

[2]    Predrag Milojkovic, Stefanie Tompkins, Ravindra Athale, Gradient Index Optics, Optical Engineering, November 2013/Vol. 52(11), p. 112101-1 

[3]    F. Gaufillet, É. Akmansoy, Graded photonic crystals for graded index lens, Optics Communications, Volume 285, 2638 (2012). 

[4]    F. Gaufillet, É. Akmansoy, Design of flat graded index lenses using dielectric Graded Photonic Crystals, Optical Materials, Vol. 47, 555-560 (2015) 

[5]    F. Gaufillet and E. Akmansoy. Design and experimental evidence of a flat graded-index photonic crystal lens. Journal of Applied Physics, 114(8) :083105, 2013. 

[6]    F. Gaufillet and E. Akmansoy. Graded photonic crystals for Luneburg lens, IEEE Photonics Journal, vol. 8, 11 (2016) 

[7]    CVI Melles Griot. Gradient-Index Lenses – http://pdf.directindustry.com/pdf/cvi-melles-griot/gradient-index-lenses/12567-66963.html. 

[8]    Z. Han, X. Checoury, D. Néel, S. David, M. El Kurdi, P. Boucaud, Optimized design for ultra-high Q silicon photonic crystal cavities, Optics Communications 283 (21) (2010) 4387 – 4391.