Periodic Reporting for period 5 - ASCIR (Active Suspensions with Controlled Interaction Rules)
Berichtszeitraum: 2022-04-01 bis 2023-03-31
The main scope of this project is to identify rules in interacting particle systems which allow to understand the formation of living matter in swarms, flocks and swirls. Using a novel experimental approch, where such interaction rules can be implemented, we are aiming to investigate the conditions of emergent phenomena in experimental systems, i.e. in the presence of thermal noise but also imperfections regarding particle properties (shape, velocity etc.). Apart from a better understanding of communication in living systems, our work might provide useful information regarding minimal interaction rules in microrobotic systems which are considered as future autonomous systems to deliver e.g. drugs or other loads within liquid environments.
The aim of this ERC grant was the realization and application of a novel experimental approach to investigate collective behaviors and group formation in active colloidal particles which are interacting via user-defined rules. Opposed to previous approaches, this project is aiming to mimic biologically relevant interaction rules in a synthetic systems beyond reciprocal pair interactions as typically considered in the field of active matter. Such non-reciprocal interactions are believed to have a strong impact on group behavior and collective phenomena. Another important direction of this project was to demonstrate how viscoelastic fluids which provide the natural habitat of many microorganisms modify the motion of active particles. Because such fluids respond to particle motion in a time-delayed manner, this is expected to lead to considerable differences compared to simple Newtonian fluids, e.g. water.
We have successfully designed and developed the hard- and software required to realize an optical feedback-loop where almost arbitrary user-defined interaction-rules in active systems can be realized. A particular challenge was, to connect all those components in order to form an efficient feedback loop where the live video-information from single frames was evaluated in real time to steer the laser beam to a single particle. Amongst different interaction rules which have been realized during this project we investigated quorum sensing which allows organisms to change their motility depending on the local density of neighbours. Using colloidal microswimmers, we could not only demonstrate quorum sensing in an synthetic system but also that this phenomen is only observed within a rather small range of parameters. Only when the density-dependent threshold for the motility change is choosen properly, this leads to the formation of dense groups. In addition we have also studied vision-based interaction rules where the particles change their swimming velocity depending on the density of neighbours within their vision cone. Interestingly, we found that for certain vision angles, this leads – even without any attractive forces – to a cohesive swarm similar to e.g. those found in midgets. Those those results have been also corroborated by a theoretical model developed by us. A major part of this grant was devoted in understanding how a viscoelastic swimming medium is affecting the properties of a microswimmer. Such viscoelastic liquids (blood, mucus, …) provide the typical environment of many microorganisms but only little is known, how they affect their motion. With our experiments we could demonstrate, that the time-delayed response of viscoelastic fluids has a profound effect on the particle motion. A particularly suprising effect is, that the rotational diffusion of microswimmers in viscoleastic fluids can be increased by several orders of magnitude. This is caused by the time-delayed response of strains exerted from the fluid onto the particle which leads to a fluctuationg torque with zero mean. When the propulsion velocity of particles exceeds a critical value, such torques eventually lead to a persistent circular motion. Such behavior is rather astonishing since there is no asymmetry in the shape of the particle nor the structure of the fluid. We also applied the concept of Reinforcement Learning (RL) to our experimental approach. Opposed to static interaction rules which are followed by every particle, this is different in case of RL. Here, the motion of particles is evaluated regarding a defined goal and changed accordingly using neuronal networks. During this process, particles “learn” to optimize their motion in view of a specific goal. We have tested this approach using the example of the active selfish herd, where each particle is aiming to minimize its predation risk by seeking proximity with neighbours. Interestingly, this does not only lead to a cohesive group but also to a swirling motion of the entire group. Most stringingly, however, the selfish goal of each partice to minimize its predation risk leads to a fair risk distribution amoung all individuals within the group. We believe that this result is particularly valuable since it suggests that collective behaviours may even result from a selfish motivation of individuals.
We have successfully designed and developed the hard- and software required to realize an optical feedback-loop where almost arbitrary user-defined interaction-rules in active systems can be realized. A particular challenge was, to connect all those components in order to form an efficient feedback loop where the live video-information from single frames was evaluated in real time to steer the laser beam to a single particle. Amongst different interaction rules which have been realized during this project we investigated quorum sensing which allows organisms to change their motility depending on the local density of neighbours. Using colloidal microswimmers, we could not only demonstrate quorum sensing in an synthetic system but also that this phenomen is only observed within a rather small range of parameters. Only when the density-dependent threshold for the motility change is choosen properly, this leads to the formation of dense groups. In addition we have also studied vision-based interaction rules where the particles change their swimming velocity depending on the density of neighbours within their vision cone. Interestingly, we found that for certain vision angles, this leads – even without any attractive forces – to a cohesive swarm similar to e.g. those found in midgets. Those those results have been also corroborated by a theoretical model developed by us. A major part of this grant was devoted in understanding how a viscoelastic swimming medium is affecting the properties of a microswimmer. Such viscoelastic liquids (blood, mucus, …) provide the typical environment of many microorganisms but only little is known, how they affect their motion. With our experiments we could demonstrate, that the time-delayed response of viscoelastic fluids has a profound effect on the particle motion. A particularly suprising effect is, that the rotational diffusion of microswimmers in viscoleastic fluids can be increased by several orders of magnitude. This is caused by the time-delayed response of strains exerted from the fluid onto the particle which leads to a fluctuationg torque with zero mean. When the propulsion velocity of particles exceeds a critical value, such torques eventually lead to a persistent circular motion. Such behavior is rather astonishing since there is no asymmetry in the shape of the particle nor the structure of the fluid. We also applied the concept of Reinforcement Learning (RL) to our experimental approach. Opposed to static interaction rules which are followed by every particle, this is different in case of RL. Here, the motion of particles is evaluated regarding a defined goal and changed accordingly using neuronal networks. During this process, particles “learn” to optimize their motion in view of a specific goal. We have tested this approach using the example of the active selfish herd, where each particle is aiming to minimize its predation risk by seeking proximity with neighbours. Interestingly, this does not only lead to a cohesive group but also to a swirling motion of the entire group. Most stringingly, however, the selfish goal of each partice to minimize its predation risk leads to a fair risk distribution amoung all individuals within the group. We believe that this result is particularly valuable since it suggests that collective behaviours may even result from a selfish motivation of individuals.
We have created a novel and world-wide unique experimental setup which allows to individually adjust the self-propulsion of colloidal microswimmers depending on user-defined interaction rules. This enabled us to explore how social information being exchanged and processed by individuals enables them to gather in cohesive, ordered groups such as flocks, swirls or swarms. As propulsion mechanism we have developed a scheme which is based on the laser-induced local heating of colloidal silica beads which are capped on one side by a thin carbon layer. When suspended in a critical mixture which is kept below its lower demixing point, the light absorbing carbon cap leads to temperature gradient around the particles leading to a local demixing and thus triggering a self-propulsive particle motion. The propulsion velocity is controlled by the laser intensity. For the individual propulsion control of more than one hundred colloidal beads, we have designed and realized a rapid laser deflection system, where a single laser beam is rapidly scanned using acoustic optical deflectors across all particles in the system. This scanning system is connected to a imaging software which has been also developed within this project and which is capable to track particle positions and their orientations in quasi real time (i.e. the timescale of particle tracking is much faster than the time scale of particle motion). The laser-scanning and particle-imaging components are connected to form a feedback system which allows to make the particle illumination (i.e. their self-propulsion) dependent on the configuration (local density, particle orientation etc.) of their neighbours and thus allows us to make the particle propulsion to be dependent on interaction rules defined by us. Because such interaction rules do not depend on physical interactions but are imposed externally, this allows to design rules which go beyond classical pair interactions and may even break reciprocity. In addition to static interaction rules, we have also expaneded this approach towards reinforcement learning, where a specific behavior towards a task of the group is “learned” using neuronal networks. Such multi-agent reinforcement learning has been implemented for the first time in an active colloidal systems within this grant and we are expecting that this work is motivating further studies in this direction.