Simulation of hot carriers by using ab-initio parameters for energy harvesting
Starting from October 2021
Nowadays, energy conversion devices are mainly designed by using macroscopic models (such as the drift-diffusion
and the heat Fourier’s formalisms) which assume local equilibrium and simplified material properties (energy dispersion, relaxation times...) that must be known a priori.
In the framework of future generations of energy converters that will most probably involve high energy carriers, out-of-equilibrium carrier distributions and complex band structures as well as strong quantum effects [1,2], these tools
reach their limit of accuracy .
Besides, these new routes of development will require to investigate a large panel of new materials  and nanostructures probably different from those which are well known and commonly used today in the microelectronics industry such as 2D materials (mono or
multilayers). In recent years, a spectacular progress in ab-initio DFT (density functional theory)-based description of the electronic and thermal transport has been achieved and this method can be used to investigate these recently studied materials.
The goal of this internship proposed in the COMputational electronICS group at the C2N is to developed transport
simulation of charge and heat by coupling existing simulation platform  with parameter calculated by using ab-initio methods. More precisely, in our home-made Monte Carlo simulator will be included the full band electron/phonon scattering rates, in particular electron/phonon coupling and impact ionization phenomenon (which is important for hot carriers in small gap materials like InAs) as well as interface transmission. Electrical thermal properties as well as the thermalization time of hot carriers in nanostructures
will be investigated and compared with the experiment results. The proposed modeling development will be carried out in close connection with the experimental activities performed by J. Chaste and his co-workers in the C2N.
This work is supported by the ANR project “Placho” and the Labex Nanosaclay via the MACACQU flagship. This
internship is expected to be continued in a Phd program.  D. Cakiroglu et al., Solar Energy Materials and Solar Cells 203, 110190 (2019).
 T. Abu Hamed et al., EPJ Photovolt. 9, 10 (2018).
 A. J. Nozik, Nature Energy 3, 170 (2018).
 I. Konovalov and V. Emelianov, Energy Sci Eng 5, 113 (2017).
 B. Davier et al., J. Phys.: Condens. Matter 30, 495902 (2018).
Artificial intelligence models for opto-mechanical silicon metamaterials
Starting from January 2021
Stimulated Brillouin scattering (SBS), mediating (THz) photons and (GHz) acoustic phonons, has an immense potential for opto-acoustic signal processing which have no analogue in conventional electro-optic or all-optic approaches. Namely, SBS has shown ultra-high-resolution filtering and remarkable low noise signal regeneration. Furthermore, the narrow linewidth and low phase noise of Brillouin lasers, make them an ideal solution for high-performance micro-wave signal generation.
SBS has been extensively developed in optical fibers, however the phonon leakage towards the cladding in conventional silicon-on-insulator (SOI) waveguides precluded the observation of SBS in Si photonics. However, the recent development of a new generation of Si optomechanic waveguides (see Fig. 1) has removed this barrier, revolutionizing the field and allowing the experimental demonstration of SBS nonlinearities surpassing Kerr and Raman effects (in 2013 ), complete phononic bandgap (in 2014 ), net amplification (in 2016 ), and Si Brillouin laser (in 2017 ). Still, fully exploiting the potential of these Brillouin optomechanical interactions in silicon requires of novel design strategies and, and advanced design tools allowing their optimization.
The goal of this internship is to develop artificial intelligence models to harness the unique degrees of freedom of subwavelength Si nanostructures to independently tailor photonic and phononic modes, providing simultaneous tight confinement and strong overlap, thus maximizing the efficiency of the SBS effect.
Deep learning models for miniaturized silicon photonics sensors
Starting from January 2021
Lab-on-chip sensors are miniaturized circuits that integrate all key functionalities within a single chip, including target preparation, detection and data analysis. Ideally, these devices should be compact, robust and low cost, allowing large-volume production. This would allow the widespread deployment of a wide range of high-impact applications such as invasive medical diagnostics, food quality control and air pollution monitoring. In addition, lab-on-chip sensors should comprise sophisticated and intelligence enough (data monitoring, processing and analysis) to be used by non-skilled personal.
Deep learning methods rely on a training process and different abstraction levels to yield models able to automatically discover the representations needed for detection or classification. Deep learning models can perform precise detection regardless the changes in operation parameters within a training range, thus obviating the need for tight calibration processes. The different abstraction levels are learned from data training, and not from engineering design. This provides deep learning methods with a unique flexibility and outstanding processing capabilities that are already being exploited in image or speech recognition or predicting DNA mutations. We experimentally demonstrate for the first time that the use of machine learning substantially improves the tolerances of on-chip SHFT spectrometers against temperature variations, thus opening a new route for their use in realistic applications outside the controlled laboratory environment.
The goal of this internship will be to explore the use of deep learning algorithms to improve the performance, robustness and flexibility of miniaturized silicon photonics sensors. Simplified models of Si sensors and deep learning algorithms will be combined to alleviate performance degradations due to non-idealities and to develop advanced detection and processing functionalities.