This work, which is the fruit of a collaboration between the groups at the CNR-IPCF and the University of Gothenburg, shows the potential of machine learning for faster and more accurate optical forces calculations. The extended abstract can be found here.
Machine learning to enhance the calculation of optical forces in the geometrical optics approximation David Bronte Ciriza, Alessandro Magazzù, Agnese Callegari, Maria A. Iatì, Giovanni Volpe, Onofrio M. Maragò Submitted to: OSA-OMA-2021, AF2D.2 Contribution Date: 16 April Time: 17:00 CEST
Short Abstract: We show how machine learning can improve the speed and accuracy of the optical force calculations in the geometrical optics approximation
Extended Abstract: Light can exert forces by exchanging momentum with particles. Since the pioneering work by Ashkin in the 1970’s, optical forces have played a fundamental role in fields like biology, nanotechnology, or atomic physics. Optical tweezers, which are instruments that, by tightly focusing a laser beam, are capable of confining particles in three dimensions, have become a common tool for manipulation of micro- and nano- particles, as well as a force and torque transducer with sensing capabilities at the femtonewton level. Optical tweezers have also been successfully employed to explore novel phenomena, including protein folding and molecular motors, or the optical forces and Brownian motion of 1D and 2D materials.
Numerical simulations play a fundamental role in the planning of experiments and in the interpretation of the results. In some basic cases for optical tweezers, the optical trap can be approximated by a harmonic potential. However, there are many situations where this approximation is insufficient, for example in the case of a particle escaping an optical trap, or for particles that are moving on an optical landscape but are not trapped. In these cases, a more complex treatment of the light-matter interaction is required for a more accurate calculation of the forces. This calculation is computationally expensive and prohibitively slow for numerical simulations when the forces need to be calculated many times in a sequential way. Recently, machine learning has been demonstrated to be a promising approach to improve the speed of these calculations and therefore, to expand the applicability of numerical simulations for experimental design and analysis.
In this work, we explore the geometrical optics regime, valid when the particles are significantly bigger than the wavelength of the incident light. This is typically the case in experiments with micrometer-size particles. The optical field is described by a collection of N light rays and the momentum exchange between the rays and the particle is calculated employing the tools of geometrical optics. The limitation of considering a discrete N number of light rays introduces artifacts in the force calculation. We show that machine learning can be used to improve not only the speed but also the accuracy of the force calculation. This is first demonstrated by training a neural network for the case of a spherical particle with 3 degrees of freedom accounting for the position of the particle. We show how the neural network improves the prediction of the force with respect to the initial training data that has been generated through the geometrical optics approach.
Starting from these results for 3 degrees of freedom, the work has been expanded to 9 degrees of freedom by including all the relevant parameters for the calculation of the optical forces considering also different refractive indexes, shapes, sizes, positions, and orientations of the particle besides different numerical apertures of the objective that focuses the light.
This work proves machine learning as a compact, accurate, and fast approach for optical forces calculation and presents a tool that can be used to study systems that, due to computation limitations, were out of the scope of the traditional ray optics approach.
Abstract: We numerically investigate the morphology and disclination line dynamics of active nematic droplets in three dimensions. Although our model incorporates only the simplest possible form of achiral active stress, active nematic droplets display an unprecedented range of complex morphologies. For extensile activity, fingerlike protrusions grow at points where disclination lines intersect the droplet surface. For contractile activity, however, the activity field drives cup-shaped droplet invagination, run-and-tumble motion, or the formation of surface wrinkles. This diversity of behavior is explained in terms of an interplay between active anchoring, active flows, and the dynamics of the motile disclination lines. We discuss our findings in the light of biological processes such as morphogenesis, collective cancer invasion, and the shape control of biomembranes, suggesting that some biological systems may share the same underlying mechanisms as active nematic droplets.
The last round table of this workshop regarded the topic advanced control of active matter. As organizers of round table, Audrey Nsamela, Chun-Jen Chen, Sandrine Heijnen, Harshith Bachimanchi and Alireza Khoshzaban, we welcomed and introduced our esteemed guests, namely Jérémie Palacci from University of San Diego, Clemens Bechinger from Konstanz University, Frank Cichos from Leipzig University, and Lucio Isa from ETH Zurich.
The round table started out with a clarification on advanced
control of active matter. Active matter can be controlled by numerous external stimuli but implementing control on individual particles or artificial entities is what qualifies as advanced control. Currently, the control of active matter is still far from the behavior and control micro-organisms have on that scale; hence
a big challenge lies there for us. Jérémie Palacci introduced an interesting research topic where they found a way to regulate the swimming process of E. Coli by light illumination. Here genetic modification was used to control the proton pump involved in the energy transportation process.
Advanced control of active matter can be applied to model
systems where the control is lacking, for example biological systems. In a biological system the control over an organism is limited to the external stimuli that are applied and won’t always result in the same reaction.
Therefore, using active particles showing predictable and reproducible behaviors when exposed to a stimulus works perfectly to model and to probe different parameters and thus provide a deeper understanding of the system. The fact advanced control of active matter doesn’t have an application outside of modelling
systems is something we shouldn’t be ashamed of.
We concluded the meeting by asking every one of our guests what
the promising research directions in the advanced control of active matter are.
All of them had a different perspective. Starting with Clemens Bechinger, who was most invested in the further exploration of the applications for model systems. Lucio Isa is mainly looking forward to explore the different materials that we can use to create active material that can subsequently be controlled. Frank Cichos
mentioned the importance of looking into new ways to create active particles. So far nature was able to achieve production of active entities with limited waste whereas human production is rather inefficient. Jérémie Palacci pointed out that the current man-made active matter systems are reacting to a strong signal in a well-controlled environment, where nature faces many more factors and
still works. It would be interesting to design a system that is resistant to noise.
The fourth roundtable was an opportunity for all students to discuss the topic “Collective Behavior” on Zoom with a panel of guests: Clemens Bechinger from the University of Konstanz, Ivo Buttinoni from Heinrich Heine University in Dusseldorf and Caroline Beck Adiels from Gothenburg University. The event was organized by Daniela Pérez, Danne van Roon, Davide Breoni, Jérémie Bertrand, Laura Natali and Liam Ruske on March 24th.
Although the guests had different background they seemed to agree on the fact that complex behavior can emerge from an ensemble of entities that obey a small number of simple rules. Indeed, minimalistic models such as the Vicsek model account for phase transition from a disordered motion to large scale motion and more; phenomena that appear to be universal.
A question on the role of intelligence and communication in collective behavior started the discussion. Although some animals or colony of bacteria may seem intelligent (e.g. escaping from a predator in a clever way or making long-lasting symbiotic microfilms), we must bear in mind that collective behavior is… collective, and rarely arises from decisions made individually. It may be said that in the animal kingdom, the need for survival requires a need to adapt and therefore to be intelligent, but this need for intelligence can be outsourced and solved at the level of the group rather than hardwired in the physical brain of each animal (or human).
It is also conceivable that one of the entities acts as a leader and ignites a collective behavior. Giovanni Volpe made an interesting remark, stating that a leader is the one who defines the objective function to be optimized by the group. The idea of leadership in collective behavior of microscopic systems remain largely unexplored by physicists.
After one hour of fruitful discussion and back and forth between the students and the guests, the session was finished and we resumed our activities with a better understanding of collective behavior. We thank the panelists for their inputs and attendance!
A lot is understood about the ways in which single cells move, but there are still many questions about the motion and organisation of cell aggregates where cells coupled through intercellular junctions show a range of collective behaviours.
This work, which has been recently published Phys. Rev. X 11, 021001 (2021), shows the potential of active nematic continuum models to describe collective cell motion in a three dimensional environment.
Active matter describes systems—living and synthetic—where a continuous influx of energy at the level of individual components leads to striking collective behavior among the individual components, such as self-organizing bacteria colonies, bird flocks, or polymers in the cytoskeleton of cells. Understanding their behavior has attracted interest for studies of biological systems—from the spread of cancer to the development of organisms—as well the development of mesoscopic engines. Here, we numerically investigate 3D droplets composed of active matter and the ways in which their shapes change in response to the continuous input of energy.
One striking observation is the continuous formation of fingerlike protrusions, reminiscent of the collective motion of invading cancer cells. By changing the mechanical properties of the drop or the activity level, we find several different dynamical responses: For example, the droplet surface can wrinkle in a way that resembles a walnut or the active forces can drive a dimple in the droplet to grow, leading to a cup shape. Such invagination is reminiscent of patterns seen during morphogenesis.
Understanding the behavior of model systems, here a continuum model of active material, is an important step toward the goal of understanding the role of physical theories in the life sciences.
On Tuesday 23 March the fourth round table of the initial training on experimental methods for active matter took place. The topic of the round table was “Optics, Spectroscopy, Micro and Nanofabrication, and Nanotribology”, and the discussion was led by Ayten Gülce Bayram , David Bronte Ciriza, Dana Hassan, Carolina van Baalen and Jesús Manuel Antúnez Domínguez. The panelists included Maria Grazia Donato, Pietro Gucciardi, Antonino Foti, Shivaprakash Ramakrishna, and Felix Holzner.
The importance of the topic of the round table to the field of active matter was motivated by the panelists from different perspectives. The discussion ranged from the main differences and challenges that come along with working on the micro- and nanoscale, to how changing the dimensions of your system allows one to change the properties of a system’s response, as well as the challenges involved in bringing a product idea to the market. The main conclusion was that the nanoscale is exciting, but the smaller you get, the greater the challenge.
Optical tweezers in a dusty universe P. Polimeno, A. Magazzù, M. A. Iatì, R. Saija, L. Folco, D. Bronte Ciriza, M. G. Donato, A. Foti, P. G. Gucciardi, A. Saidi, C. Cecchi-Pestellini, A. Jimenez Escobar, E. Ammannito, G. Sindoni, I. Bertini, V. Della Corte, L. Inno, A. Ciaravella, A. Rotundi & O. M. Maragò
Eur. Phys. J. Plus 136, 339 (2021)
Abstract: Optical tweezers are powerful tools based on focused laser beams. They are able to trap, manipulate, and investigate a wide range of microscopic and nanoscopic particles in different media, such as liquids, air, and vacuum. Key applications of this contactless technique have been developed in many fields. Despite this progress, optical trapping applications to planetary exploration are still to be developed. Here we describe how optical tweezers can be used to trap and characterize extraterrestrial particulate matter. In particular, we exploit light scattering theory in the T-matrix formalism to calculate radiation pressure and optical trapping properties of a variety of complex particles of astrophysical interest. Our results open perspectives in the investigation of extraterrestrial particles on our planet, in controlled laboratory experiments, aiming for space tweezers applications: optical tweezers used to trap and characterize dust particles in space or on planetary bodies surface.
The third round table session of the experimental training was about machine learning and its role in science, in particular physics and active matter. The panelists invited to the discussion were Carlo Manzo from Vic University, Benjamin Midtvedt and Saga Helgadottir from Gothenburg University, Onofrio Maragò and Alessandro Magazzù from CNR ICPF-Messina. The discussion was organized and lead by Jesus M. A. Dominguez, Davide Breoni, Liam Ruske, Chun-Jen Chen and Alireza Khoshzaban, who are students attending the training.
The discussion touched topics like the applications of machine learning in fields like optics, biophysics, medical research, the potentialities and the reliability of the method. Questions on when a machine learning approach is advisable and how cautious one must be when applying machine learning were also addressed. Current important logical and practical aspects of the method were also discussed, together with the need of testing machine learning applications against more classical ones. The panelists also stated the importance of reliably checking the results obtained to avoid biases that can lead to false conclusions.
After one hour of fruitful discussion we gained a broader perspective and a deeper understanding of machine learning.