arXiv subjects

Sylvain Gigan

Publications and source records attributed to Sylvain Gigan.

At least 73 records · Page 4Linked to original sources

arXiv2021

Chromato-axial memory effect in step index multimode fibers

Multimode fibers (MMF) are used in many applications from telecomunications to minimally invasive micro-endoscopic imaging. However, the numerous modes and their coupling make light-beam control and imaging a delicate task. To circumvent this difficulty, recent methods exploit priors about the transmission of the system, such as the so-called optical memory effect. Here, we quantitatively characterize a chromato-axial memory effect in step-index MMF, characterized through its slope $δz/δλ$ and its spectral and axial widths. We propose a theoretical model and numerical simulations in good agreement with experimental observations.

physics.optics↗

arXiv2021

LightOn Optical Processing Unit: Scaling-up AI and HPC with a Non von Neumann co-processor

We introduce LightOn's Optical Processing Unit (OPU), the first photonic AI accelerator chip available on the market for at-scale Non von Neumann computations, reaching 1500 TeraOPS. It relies on a combination of free-space optics with off-the-shelf components, together with a software API allowing a seamless integration within Python-based processing pipelines. We discuss a variety of use cases and hybrid network architectures, with the OPU used in combination of CPU/GPU, and draw a pathway towards "optical advantage".

cs.AR↗

arXiv2021

Large field-of-view non-invasive imaging through scattering layers using fluctuating random illumination

On-invasive optical imaging techniques are essential diagnostic tools in many fields. Although various recent methods have been proposed to utilize and control light in multiple scattering media, non-invasive optical imaging through and inside scattering layers across a large field of view remains elusive due to the physical limits set by the optical memory effect, especially without wavefront shaping techniques. Here, we demonstrate an approach that enables non-invasive fluorescence imaging behind scattering layers with field-of-views extending well beyond the optical memory effect. The method consists in demixing the speckle patterns emitted by a fluorescent object under variable unknown random illumination, using matrix factorization and a novel fingerprint-based reconstruction. Experimental validation shows the efficiency and robustness of the method with various fluorescent samples, covering a field of view up to three times the optical memory effect range. Our non-invasive imaging technique is simple, neither requires a spatial light modulator nor a guide star, and can be generalized to a wide range of incoherent contrast mechanisms and illumination schemes.

physics.optics↗

arXiv2021

Roadmap on multimode light shaping

Our ability to generate new distributions of light has been remarkably enhanced in recent years. At the most fundamental level, these light patterns are obtained by ingeniously combining different electromagnetic modes. Interestingly, the modal superposition occurs in the spatial, temporal as well as spatio-temporal domain. This generalized concept of structured light is being applied across the entire spectrum of optics: generating classical and quantum states of light, harnessing linear and nonlinear light-matter interactions, and advancing applications in microscopy, spectroscopy, holography, communication, and synchronization. This Roadmap highlights the common roots of these different techniques and thus establishes links between research areas that complement each other seamlessly. We provide an overview of all these areas, their backgrounds, current research, and future developments. We highlight the power of multimodal light manipulation and want to inspire new eclectic approaches in this vibrant research community.

physics.optics↗

arXiv2021

Accelerating ptychographic reconstructions using spectral initializations

Ptychography is a promising phase retrieval technique for label-free quantitative phase imaging. Recent advances in phase retrieval algorithms witnessed the development of spectral methods, in order to accelerate gradient descent algorithms. Using spectral initializations on experimental data, for the first time we report three times faster ptychographic reconstructions than with a standard gradient descent algorithm and improved resilience to noise. Coming at no additional computational cost compared to gradient-descent-based algorithms, spectral methods have the potential to be implemented in large-scale iterative ptychographic algorithms.

physics.optics↗

arXiv2021

Scalable spin-glass optical simulator

Many developments in science and engineering depend on tackling complex optimizations on large scales. The challenge motivates intense search for specific computing hardware that takes advantage from quantum features, nonlinear dynamics, or photonics. A paradigmatic optimization problem is finding low-energy states in classical spin systems with fully-random interactions. To date no alternative computing platform can address such spin-glass problems on a large scale. Here we propose and realize an optical scalable spin-glass simulator based on spatial light modulation and multiple light scattering. By tailoring optical transmission through a disordered medium, we optically accelerate the computation of the ground state of large spin networks with all-to-all random couplings. Scaling of the operation time with the problem size demonstrates optical advantage over conventional computing. Our results point out optical vector-matrix multiplication as a tool for spin-glass problems and provide a general route towards large-scale computing that exploits speed, parallelism and coherence of light.

physics.optics↗

arXiv2021

Spatial entanglement engineering by pump shaping

The ability to engineer the properties of quantum optical states is essential for quantum information processing applications. Here, we demonstrate tunable control of spatial correlations between photon pairs produced by spontaneous parametric down-conversion. By shaping the spatial pump beam profile in a type-I collinear configuration, we tailor the spatial structure of coincidences between photon pairs entangled in high dimensions, without effect on intensity. The results highlight fundamental aspects of spatial coherence and hold potential for the development of quantum technologies based on high-dimensional spatial entanglement.

quant-ph↗

arXiv2021

Deeply Sub-Wavelength Localization with Reverberation-Coded-Aperture

Accessing sub-wavelength information about a scene from the far-field without invasive near-field manipulations is a fundamental challenge in wave engineering. Yet it is well understood that the dwell time of waves in complex media sets the scale for the waves' sensitivity to perturbations. Modern coded-aperture imagers leverage the degrees of freedom (DoF) offered by complex media as natural multiplexor but do not recognize and reap the fundamental difference between placing the object of interest outside or within the complex medium. Here, we show that the precision of localizing a sub-wavelength object can be improved by several orders of magnitude simply by enclosing it in its far field with a reverberant chaotic cavity. We identify deep learning as suitable noise-robust tool to extract sub-wavelength information encoded in multiplexed measurements, achieving resolutions well beyond those available in the training data. We demonstrate our finding in the microwave domain: harnessing the configurational DoF of a simple programmable metasurface, we localize a sub-wavelength object inside a chaotic cavity with a resolution of $λ/76$ using intensity-only single-frequency single-pixel measurements. Our results may have important applications in photoacoustic imaging as well as human-machine interaction based on reverberating elastic waves, sound or microwaves.

physics.app-ph↗

arXiv2020

Large-Scale Optical Reservoir Computing for Spatiotemporal Chaotic Systems Prediction

Reservoir computing is a relatively recent computational paradigm that originates from a recurrent neural network and is known for its wide range of implementations using different physical technologies. Large reservoirs are very hard to obtain in conventional computers, as both the computation complexity and memory usage grow quadratically. We propose an optical scheme performing reservoir computing over very large networks potentially being able to host several millions of fully connected photonic nodes thanks to its intrinsic properties of parallelism and scalability. Our experimental studies confirm that, in contrast to conventional computers, the computation time of our optical scheme is only linearly dependent on the number of photonic nodes of the network, which is due to electronic overheads, while the optical part of computation remains fully parallel and independent of the reservoir size. To demonstrate the scalability of our optical scheme, we perform for the first time predictions on large spatiotemporal chaotic datasets obtained from the Kuramoto-Sivashinsky equation using optical reservoirs with up to 50 000 optical nodes. Our results are extremely challenging for conventional von Neumann machines, and they significantly advance the state of the art of unconventional reservoir computing approaches, in general.

physics.optics↗

arXiv2020

Hardware Beyond Backpropagation: a Photonic Co-Processor for Direct Feedback Alignment

The scaling hypothesis motivates the expansion of models past trillions of parameters as a path towards better performance. Recent significant developments, such as GPT-3, have been driven by this conjecture. However, as models scale-up, training them efficiently with backpropagation becomes difficult. Because model, pipeline, and data parallelism distribute parameters and gradients over compute nodes, communication is challenging to orchestrate: this is a bottleneck to further scaling. In this work, we argue that alternative training methods can mitigate these issues, and can inform the design of extreme-scale training hardware. Indeed, using a synaptically asymmetric method with a parallelizable backward pass, such as Direct Feedback Alignement, communication needs are drastically reduced. We present a photonic accelerator for Direct Feedback Alignment, able to compute random projections with trillions of parameters. We demonstrate our system on benchmark tasks, using both fully-connected and graph convolutional networks. Our hardware is the first architecture-agnostic photonic co-processor for training neural networks. This is a significant step towards building scalable hardware, able to go beyond backpropagation, and opening new avenues for deep learning.

cs.LG↗

arXiv2020

Directional control of weakly localized Raman from a random network of fractal nanowires

Disordered optical media are an emerging class of materials capable of strongly scattering light. Their study is relevant to investigate transport phenomena and for applications in imaging, sensing and energy storage. While such materials can be used to generate coherent light, their directional emission is typically hampered by their very multiple scattering nature. Here, we tune the out-of-plane directionality of coherent Raman light scattered by a fractal network of silicon nanowires. By visualizing Rayleigh scattering, photoluminescence and weakly localized Raman light from the random network of nanowires via real-space microscopy and Fourier imaging, we gain insight on the light transport mechanisms responsible for the material's inelastic coherent signal and for its directionality. The possibility of visualizing and manipulating directional coherent light in such networks of nanowires opens venues for fundamental studies of light propagation in disordered media as well as for the development of next generation optical devices based on disordered structures, inclusive of sensors, light sources and optical switches.

physics.optics↗

arXiv2020

Programmable linear quantum networks with a multimode fibre

Reconfigurable quantum circuits are fundamental building blocks for the implementation of scalable quantum technologies. Their implementation has been pursued in linear optics through the engineering of sophisticated interferometers. While such optical networks have been successful in demonstrating the control of small-scale quantum circuits, scaling up to larger dimensions poses significant challenges. Here, we demonstrate a potentially scalable route towards reconfigurable optical networks based on the use of a multimode fibre and advanced wavefront-shaping techniques. We program networks involving spatial and polarisation modes of the fibre and experimentally validate the accuracy and robustness of our approach using two-photon quantum states. In particular, we illustrate the reconfigurability of our platform by emulating a tunable coherent absorption experiment. By demonstrating reliable reprogrammable linear transformations, with the prospect to scale, our results highlight the potential of complex media driven by wavefront shaping for quantum information processing.

quant-ph↗

arXiv2020

Mean path length invariance in wave-scattering beyond the diffusive regime

Diffusive random walks feature the surprising property that the average length of all possible random trajectories that enter and exit a finite domain is determined solely by the domain boundary. Changes in the diffusion constant or the mean-free path, that characterize the diffusion process, leave the mean path length unchanged. Here we demonstrate experimentally that this result can be transferred to the scattering of waves, even when wave interference leads to marked deviations from a diffusion process. Using a versatile microwave setup, we establish the mean path length invariance for the crossover to Anderson localization and for the case of a band gap in a photonic crystal. We obtain these results on the mean path length solely based on a transmission matrix measurement through a novel procedure that turns out to be more robust to absorption and incomplete measurement in the localized regime as compared to an assessment based on the full scattering matrix.

cond-mat.soft↗

arXiv2020

Gain As Configurable Disorder-Adaptive Pumping for Control of Multimode Fiber Amplifiers and Lasers

Media featuring both optical gain and disorder, such as random lasers, represent formidable challenges as subjects of research due to the high complexity of the light propagation within them; however, dramatic advances in this nascent field have been furnished by the paradigm of applying wavefront shaping techniques to the beam pumping the system. We present here a theoretical and experimental study employing this approach in a gain medium where the disorder arises not from random scattering but rather from the multimodality of a waveguide: an amplifying multimode fiber. The shaping of the core-guided pump prior to its injection affects the complex, speckle-like patterns of excitation within the fiber volume. Thus we are offered the intriguing prospect of manipulating the spatially heterogeneous gain in a system where the disorder, albeit highly complex, may be fully understood in terms of the discrete eigenmodes of a well-known waveguide. We study our medium in two different configurations: as an optical amplifier, and as a fiber laser. In the first configuration we show that dependence upon the pump configuration, of the amplifier transmission function, surprisingly survives several physical mechanisms which at first sight would appear to severely limit it. This insight is then carried on to the lasing cavity configuration, where we find striking parallels between the behavior of the system and that of random lasers; more importantly, the pump shaping shows a strong ability to control the complex emission, in particular - to selectively favor individual spectral lines and stabilize single-mode operation.

physics.optics↗

arXiv2020

Spectrally-resolved point-spread-function engineering using a complex medium

Propagation of an ultrashort pulse of light through strongly scattering media generates an intricate spatio-spectral speckle that can be described by means of the multi-spectral transmission matrix (MSTM). In conjunction with a spatial light modulator, the MSTM enables the manipulation of the pulse leaving the medium; in particular focusing it at any desired spatial position and/or time. Here, we demonstrate how to engineer the point-spread-function of the focused beam both spatially and spectrally, from the measured MSTM. It consists in numerically filtering the spatial content at each wavelength of the matrix prior to focusing. We experimentally report on the versatility of the technique through several examples, in particular as an alternative to simultaneous spatial and temporal focusing, with potential applications in multiphoton microscopy.

physics.optics↗

arXiv2020

Non-invasive single-shot recovery of point-spread function of a memory effect based scattering imaging system

Accessing the point-spread function (PSF) of a complex optical system is important for a variety of imaging applications. However, placing an invasive point source is often impractical, and estimating it blindly with multiple frames is slow and requires a complex non-linear optimization. Here, we introduce a simple single-shot method to non-invasively recover the accurate PSF of an isoplanatic imaging system, in the context of multiple light scattering. Our approach is based on the reconstruction of any unknown sparse hidden object using the autocorrelation imaging technique, followed by a deconvolution with a blur kernel derived from the statistics of a speckle pattern. A deconvolution on the camera image then retrieves the accurate PSF of the system, enabling further imaging applications. We demonstrate numerically and experimentally the effectiveness of this approach compared to previous deconvolution techniques.

physics.optics↗

arXiv2020

Light-in-the-loop: using a photonics co-processor for scalable training of neural networks

As neural networks grow larger and more complex and data-hungry, training costs are skyrocketing. Especially when lifelong learning is necessary, such as in recommender systems or self-driving cars, this might soon become unsustainable. In this study, we present the first optical co-processor able to accelerate the training phase of digitally-implemented neural networks. We rely on direct feedback alignment as an alternative to backpropagation, and perform the error projection step optically. Leveraging the optical random projections delivered by our co-processor, we demonstrate its use to train a neural network for handwritten digits recognition.

cs.LG↗

arXiv2020

Non-invasive focusing and imaging in scattering media with a fluorescence-based transmission matrix

In biological microscopy, light scattering represents the main limitation to image at depth. Recently, a set of wavefront shaping techniques has been developed in order to manipulate coherent light in strongly disordered materials. The Transmission Matrix approach has shown its capability to inverse the effect of scattering and efficiently focus light. In practice, the matrix is usually measured using an invasive detector or low-resolution acoustic guide stars. Here, we introduce a non-invasive and all-optical strategy based on linear fluorescence to reconstruct the transmission matrices, to and from a fluorescent object placed inside a scattering medium. It consists in demixing the incoherent patterns emitted by the object using low-rank factorizations and phase retrieval algorithms. We experimentally demonstrate the efficiency of this method through robust and selective focusing. Additionally, from the same measurements, it is possible to exploit memory effect correlations to image and reconstruct extended objects. This approach opens up a new route towards imaging in scattering media with linear or non-linear contrast mechanisms.

physics.optics↗