 should we just stay out for a few seconds so we can close once I put the screen down. So, pointer, this one, alright, thanks to the organizers for inviting me, this is work about bacteria and I don't do experiments, so the experiments that I will be talking about and some modeling of course have been done by this gentleman, it's actually will be about three sets of experiments about Japanese and Chinese bacteria from a group of Masaki Sano in Tokyo, this gentleman here is not doing experiments but working with me on the modeling side, the group of Iling Wu in the Chinese University of Hong Kong, if you can read this, and the group of Hapeng Chang in Shanghai. And so before actually proceeding to my talk I would like to let you know that I'm the lead editor of physical review letters and that's a journal that is doing physics of course, but includes also lots of work that many authors are sitting in the audience in fact, I would like to remind you that this journal publishes relevant works in the kind of thing we've been discussing here from this is actually this week robots, very simple robots, robots here doing collectively some interesting stuff, swarms, animals, this one here is actually about to appear and so on, I'm not sure we can read the titles at least, fish schools of course, flocks, termite nests, foraging problems, and humans and some others here, okay just to let you know that many of these papers have been actually highlighted in physics which is this electronic only highlighting journal and this one, this one, if I go back, this one, this one, this one and so on, these get a lot of press thanks to the work of the editors here, so anyway if you want to talk about the journal and you ever wanted, why should I publish there or not, you can talk to me, whoops, okay, I hope it's not the end of the talk, it's back, okay, all right, so okay, so why study collective motion of bacteria, as you probably know bacteria do all kinds of things collectively, you know, some bacteria crawl on surfaces, some others swim and fluids, some of it twitch, so-called twitching motility, they move together, they build together things and so this has been studied of course in microbiology for the sake of biology but here my viewpoint would be more okay can these systems be good systems for doing new, for discovering or studying a new interesting physics, so basically here this is not, you know, physics for biology to help biologists, at least that's not the main purpose or the first purpose but it's rather can biology do something interesting for physics and of course in the end we hope we can close the loop and whatever we say as a physicist has some interest for the biologist, okay, so and basically the talk will be revolving about this basic question whether can we stupidly consider bacteria swimming for example or bacteria crawling as little particles interacting with physical interactions only, so on long time scales and evolutionary time scales this is certainly, the answer is certainly no but on short time scales regarding the motion of these things together I will show you that in some cases at least it's, we have pretty good descriptions even quantitative descriptions by using strictly physical interactions between, between little particles indeed, okay, so this is within the general framework of what is now called active matter physics and before I really go to the bacteria would have to give you some background information about where we stand in active matter physics and using some results that would be useful to understand the bacteria results later, so if you really, if you really came for the bacteria you should bear with me for a few minutes and then you'll see the bacteria, okay, we've, so the bigger picture here is about statistical physics which basically is about understanding emergence or collective properties given microscopic or local rules and how can I predict and not just observe the collective behavior and we've heard about universality a few days ago in statistical physics we believe in universality we believe in it, in, it has various meanings the meaning that Andrea put forward a couple days ago Andrea Cavagna is, is well established mathematically but there is another level wider more qualitative level of universality and my favorite example for this is a very classical example but still one of the best in my view in my opinion at least okay as you may know most simple fluids are described by the Navier-Stokes equations so this is a set of partial differential equations which is very interesting mathematical object but it is well known well recognized and well understood why these equations are describing all kinds of fluids made of whatever molecules and even sometimes not even molecules little particles moving on some lattice respecting the right symmetries and conservation laws will be described at large scales by the Navier-Stokes equation and the only difference between this or that fluid would be in the coefficients of the terms of that equation so that level of universality or generosity or robustness of description is, is, is what I have in mind here when I speak of universality of universality in the sense of critical phenomena is also important but you have to have a critical system so here if you, if you assume that there must be of it could be some degree of universality at least at this general qualitative level that I described here studying simple models out of the many many many possible models giving the same collective behavior up to pre-factors studying the simplest ones is certainly a reasonable idea so that's the viewpoint we have here and so once this is said you will see I will show you very simple models and to situate to locate active matter physics here is a list of very different ways of being out of equilibrium so energy is injected in the system at some in some way you can be near equilibrium and use this proximity to have perturbative approaches you can go very very very slowly to equilibrium and that's the glass problem glassy transitions and so on you can be maintained out of equilibrium by some external field you know migration field or concentration field or the flux of something temperature gradient or you can be I would say genuinely out of equilibrium in the bulk at the level of the units in your systems which are coupled together these units spent energy to do some things and active matter falls into this class of course this is this general framework more precisely active matter is where the units actually spending energy collected from the environment or stored internally to move themselves or to move other things and the general problem of the collective properties of these active particles of self-propelled particles is often in terms of collective motion that's one of the main themes of not strictly not only okay so lots of examples of course relevant to this audience animal groups cells bacteria but also subcellular components like molecular motors and biofilaments and also artificial swimmers of some sorts and various colloidal particles okay now so on all scales and then the questions we have is whether we can have at least some cases make predictions on some understood reach some understanding which are basically across all scales another distinction which will come in the discussion I'm putting all this here because it will be interesting later in the with the bacteria experiments in active matter we often distinguish wet and dry active matter so dry active matter is when you can neglect or you want to neglect the fluid in which the particles surrounding the particles so if I if I walk here okay I displace the air around me of a wildebeest moving you know indeed the air displacement between the wildebeest is not going to be very important for fish and birds you can think about it that's not so obvious and for some swimming bacteria without confinement clearly you cannot neglect the fluid if a bacteria very confined your surfaces you will see in a moment that you can ask yourself whether this is a wet or dry system okay and of course neglecting the fluid makes things much much simpler and our current understanding of dry active matter is actually far no more advanced than of wet systems of swimmers okay now only for now again for just a dry active matter there's been lots of work in the general framework of as simple as possible models okay and typically the simple setting is you have a constant self-propulsion force a little constant speed for particles when they are at least alone it's is our physics word for saying that this is dynamic is over damped and first-order in time no inertia over we we have some words about inertia here before the interactions are strictly local for simplicity and the noise often is a stochastic component you know these persistent random walkers acting on their direction of motion rotational noise okay now of course it's important oh it's interesting only when you have interactions and these interactions can be of various types and you can have just alignment between particles and if you have point particles you know like in the big check model this is really only thing you have you have alignment competing with noise given the self-propulsion of the particles you can have just repulsion between particles and we've heard some of this something about this yesterday about MIPS MIPS is a motility induced phase separation where particles only interact by repulsion which means in that when they are entering a dense area their speed will decrease or they will just stop and so they can actually phase separate into large microscopic dance clusters without any explicit attractive force between them so that's okay and now if you think of anything realistic like elongated objects moving together and colliding into each other then you have a mix of repulsion and induce alignment and this is typically a simple models of bacteria in terms of these elongated rods self-propelled colliding and so on okay you see here there's phase separation in MIPS there's also phase separation in fact in the big check world in force point particles only interacting by alignment where alignment is clearly the dominating interaction in other terms and here I have one example of a big check style model so this is a big check model point particles constant speed some noise and the alignment here is what we call thematic alignment so that's a formula for coding it but in a nutshell basically when two particles come at some acute angle they will align and continue together and if they come at some obtuse angle larger than part of the two there where they will entire line and go away from each other but in opposite directions okay so that's an ematic alignment for a four words with words and in all these big check systems when you have for example large noise basically the noise dominates the alignment and your system is disordered some very short correlation lengths and times and if your noise is actually low enough and not necessarily zero you can have a globally ordered phase and here maybe you see that half of the little arrows are going up and half of the little arrows are going down that's a global nematic order that has a reason spontaneously from this local interactions okay in between all these big check systems in fact undergo something which is like a phase separation scenario there is a region of parameters in noise here or in density where the ordered liquid here will be ordered phase here coexists with the gas you see here you don't see it but you have to believe me this is a very dense band along which the particles are traveling 50% to the right 50% to the left which has spontaneously emerged and outside it you're left over sparse gas so bits of you have a gas here and you have the ordered phase order liquid here and you can see that tuning this parameter you can go for a very low dance fraction of dense nematic order regions to dominating them nematic order so that is a phase separation is at stake here what is also known in big check style models and all kinds of active matter systems is that in the ordered phase here you have interesting fluctuations and correlations which set up generically typically particles will super diffuse in the transverse dimension to the order okay so you don't need to input any levy walk distribution etc from these local interactions collectively the particles set up an ordered regime in which they move super diffusively another signature or correlated signature to this is a fact that from number fluctuations in these ordered phase are enormously strong you know you might want to believe that if I look at this system here and I measure the number of particles in sub boxes of increasing size okay basically I'm going to have a low of large numbers and the fluctuations of a number of particles in a given box will be like square root of a mean the root mean square of these fluctuations will go like square root of a mean number of particles that's the low of large numbers in fact that's not true and we understand why it's not true and typically what happens that you do have the here the variance of the square and I remember going like some power law this is log log scales of a mean mean number of particles and variance of this number goes like a power law that this power law is not one or not one half a root mean square something larger so the for the root mean square here we typically observe something but of the order of 0.8 instead of 0.5 which would be the law of large numbers this is very generic feature semi-quantitative feature of all these systems whenever you have alignment competing with noise another interesting signature of this case of the so-called Vickshack style rods I call these rods because they mimic the fact that two elongating objects gating into collision might have this pneumatic alignment okay here we have measured in the same ordered pneumatically ordered fluctuating phase the intensity of a quality of a pneumatic order as a function of system size or subsystem size and you see here again in log log scales that this pneumatic order parameter decreases but it decreases slower than a power law this is if it would be a power law it would be a straight line it goes like a power law to some asymptotic finite value these numerical results if I extrapolate them boldly lead me to say that in this system you may have true long-range pneumatic order which is for physicists in the room still a surprise a matter of debate okay all right all kinds of interesting physics there now can we see this anywhere okay so beyond little particles and Vickshack simulations we do have lots of theoretical work behind this and of course this is not the audience and I don't have a time to go into any of this so that just to say that we do better than just simulating Vickshack models okay me and many other people but there is a relative scarcity of experiments showing the same sort of behavior as these very simple models this is not too surprising because anything you can think of is going to be more complicated than these simple things so that's not a you know it's on it's important to understand the simplest possible situation it's not a big deal if any realistic situation is more complicated because it's still important to understand the core but would you do find some simulation some experiments some realistic situations in which what we observe is actually very very close to these silly Vickshack like things okay so I show you bacteria now and this can be also done on other systems where you have both very large number of moving objects and reasonably good control on on them you know here for biofilaments and motor proteins if you purify properly your proteins then you know exactly what's in your soup for active colloidal particles you typically they are complicated but it's just physics bacteria is more mysterious and more of a challenge but they are offering some control right so the rest of the talk is about three sets of experiments these are free all right one it would be equalized cell strongly confined between two glass plates that's what you see here and you see already maybe you see it here that there is some pneumatic order at least on this scale number two would be a quasi to the wet active pneumatic systems now I will tell you what it is of swimming bacteria you see this this is the dynamics that have been shown by an Andre at the end of his talk and it's typical of wet active pneumatics and this is the most mysterious weak synchronization of very very dense equalized suspension again okay so first first hello yes so I'm not sure if you see this on the screen here this is the same this is these equalized cells are elongated by they've been drugged so they are longer than usual they're typically 20 microns in this in this movie and they are confined between two glass plates they can hardly pass across each other but they do but just just just and I'm not sure if you see it on the screen but here there is very large scale pneumatic order in the diagonal direction from where I from where I stand I can't I cannot see it but anyway in spite of obviously the party the particles the the bacteria are swimming they are not strictly aligned in this pneumatic direction but nevertheless statistically they are moving 50 percent up there and 50 percent down here okay here it's a millimeter square field of view with very very many bacteria all of them are marked by fluorescence so you can see them rather easily when they meet each other they often collide okay and if they collide with some small but finite probability they actually end up being more aligned pneumatically aligned than they were initially that we have done this on on on individual you know binary collisions statistics okay so this global pneumatic order in fact you can study the way I've showed you for the the vixac rods very very silly model dry model so these are swimmers here that are so highly confined that you can wonder and you can see here for example here in blue it's a sparse system and not enough bacteria to have emerging long range order and the number fluctuations these are normal fluctuations here this is the root mean square divided yeah the root mean square divided by square root of mean okay so normal fluctuations low of large numbers for non-interacting or weakly interacting bacteria would be flat and it's what you have when you have not enough bacteria to have this large-scale global pneumatic order emerging but when you do have a large-scale pneumatic order you see that this thing departs from the flat here and goes up with an exponent smaller than the 0.8 of a vixac model but okay this this is experimental data it could be all right and here is a signature of again if it's disordered the pneumatic order goes down as a function of subsystem size here like it should be in a disordered system with finite correlation lengths but in the ordered situation you cannot see it here but it's basically constant it actually decreases slower than a parallel this is again log log scales it going like a parallel to a finite asymptotic value just like in the vixac model so this behaves at this semi-quantitative level very much like this very dry silly model and it's kind of a mystery why the hydrodynamic interactions are completely screened by the two surfaces in below between which the bacteria are confined okay so that's a system that as far as I can tell behaves very much to myself like the vixac system or any system of a class of these rods system vixac like rods but it is a nominally about swimmers makes you think about wet and dry you know the effect of confinement on systems all right I quickly go to the other experiment so here this is again elongated cells and so here we don't we produce pneumatic order locally you see that locally the cells are forming this pneumatic order without putting them in some pneumatic liquid crystal is just because again if these cells are slightly longer than their white type state they have been drugged again like we call I before to grow typically to larger size and when they are long enough they align strongly enough to produce local pneumatic order and what you see here what you saw here is the motion spontaneous motion of these cells so these cells is just observing this is about 300 microns over one or two millimeters of a growing colony of these cells in standard conditions at the edge of a colony you have a one or two millimeter ring in which that dynamics on almost it's not strictly monolayer but very very thin layer this dynamics sets in for on time scales which are much faster here than the general growth speed of a colony and on land scales which are shorter than this millimeter scale so we have this it's actually a very very large system of which we observe here a portion of spacetime okay here all the cells are labeled fluorescently and what's what you see here and it's so dense that you cannot it's hard to distinguish individual cells because it's so packed okay so what happened is that it is so packed that they can actually cannot swim they rotate the flagella activating the fluid transferring force momentum into the fluid but they cannot really swim because there are too many people around and they're blocked most of the time so they don't swim they agitate the fluid the collectively agitate the fluid and the fluid is set in motion and this motion of a fluid advex and rotates the cells and one way to see this maybe more more explicitly is sorry this is the same thing next slide please all right can you start here only a fraction of the cells have been marked by fluorescence and you can see that most of them are not moving along their axis they're actually advected by the flow you see this especially the long ones this one here see this they're moving like this okay sometimes there is a little space in front of one cell it can zoom in fast but basically this is a system of shakers we would call them they don't need to swim actually they cannot swim by just jamming frustration but they agitate the fluid and this fluid feeds back to this to move them and and so on giving rise to this active pneumatic chaotic regime so all right so what do we do with this we can measure superimpose here I'm not sure you see anything we can measure locally on some coarse graining scale the orientation of this pneumatic order using standard techniques I don't want to go into the details of this on the similar images we can also measure the velocity below the velocity field by piv meaning pixel image velocimetry so the way these pixels move is one way of on coarse graining scale again of measuring the actual the actual speed or velocity field of bacteria in the lab frame okay so for and that's the experimental data we have for we have movies like this from which we extract the two fields orientation orientation field and velocity field okay and now we can do which can try to match these data to or can study things here what we do in particular to to do the modeling or to help in the modeling later is to follow defects you saw the defects in the sorry I forgot to mention them in the orientational field movie here you see the little red and blue things these are the locations of the topological defects of the orientational field so here is so-called plus defect a little red things moving and blue triangular symbols are for the negative minus one half defects these are represented here as a plus defect as this shape of a pneumatic orientation is like this here with a singularity here and this is a minus one half defect with this typical triangular structure and the core here at the middle so we can track these defects because this is just always a chaotic regime there is no way to stop or regularize this efficiently experimentally we can use the defects follow them in their reference frame and average their properties to obtain fine well-defined precise accurate measurements of the orientation orientational field around them velocity field also around them and so on and these maps of velocity and orientation fields we use later to actually match a model quantitatively okay and first thing we do here is we can uh given an experimental orientational field field q or average yeah orientational pneumatic orientational field uh this the fluid velocity field v should be a solution of the stokes equation that you see here forced by a force field which is just this so-called active stress term uh which is represented here with arbitrary coefficients mu some viscosity some friction coefficient because it's all this is on a agar substrate it's pretty high friction and some some pre-factor here okay so we take an experimental q field we vary these three parameters which are two of them are two independent ratios these and this one and this one okay we solve the stokes equation not too hard we get the v field and then we can compare this reconstructed v field to the experimentally measured velocity field okay and we show that there there are optimal parameters these are the two combination of parameters here and you see there's a minimum of this quality function which is represented here in colors there's a minimum point here giving you optimal values for uh matching uh for this for the solutions of the stokes equation for the solution solution such that the solutions of the stokes equation match uh quantitatively the experimental velocity fields and what you see here is an experimental velocity field at given time and the reconstructed one from the experimental q field in passing we get good values or interesting values for this parameters i mentioned here's a movie that shows you that this is true also a long time here the parameters are fixed at the optimal values you have the original velocity field in the experiment and the reconstructed velocity field reconstructed from the orientational field okay that's one step but we want to do better and we did better but i have no time to go into too many details now we introduce a vikshaq style wet model so for those of you who know vikshaq's models in continuous time this uh this is the rotation of a little particle carrying a nematic angle with thematic alignment with neighbors and some rotational noise that's a vikshaq system provided you move at a constant speed but here the particles are also rotated by the flow field v okay according to classic terms that i cannot i don't have time to explain with some arbitrary coefficients in front of them so that's a general dynamics of alignment at local scales which we treat effectively noise but also at the rotation by the fluid flow and of course the particles are affected by the fluid flow as well in addition to the suffer portion here we have a little repulsion term between particles but that's actually not really important and the stokes equation here again is here we have a force field given by this expression which is nothing but the active stress term that i talked about before different notation sorry for this okay so we have estimated the party the parameters here at the stokes level from the previous matching and now we claim that we can find and we did actually all the parameters of this microscopic part of the model from the experimental data using in particular of the defect properties and i have no time to go into this but it works and these models give uh quantitative perfect matching to anything you can measure from the experimental system including the fine details of the defect and their dynamics uh they are very very efficient numerically so that we can also simulate easily uh numbers of swimmers which are realistic given the experiments we have can go to millions of swimmers very easily so i have no time to go into to convince you that we did this but here uh we have a truly quantitative estimation of all model parameters and evidence that these model parameters are optimal in some sense to match the data then we vary the control parameters of the experiment which is translated into variations in the control parameter of the parameters of the model here all these parameters and that for at least to us physicists tells us brings us a lot of information about the experimental system what is really important what are these terms doing is this for example this this one this term is not changed by any of the uh experimental parameters we should not surprise those who know fluid dynamics but this one is which is kind of a surprise and then etc so we have lots of knowledge about the experimental system from this quantitative matching done at every single experiment we have all right this is uh experiment number two where we do active pneumatics and match it to a microscopic level at fully quantitative stage now i'll go to the maybe yes third part i'm not sure how much i've left one minute that's not going to do it okay so in the previous two examples you could see every single bacterium almost and see how it participates to the collective dynamics and now and i'm just show you a movie of of something where you don't see what or you don't understand how the bacteria can produce a collective behavior because it's it's not readable from their trajectory so okay if you grow equalized cells in the lab this is done in every biology lab standard growth conditions optimal growth conditions and you let it grow to very very dense swarming liquid you observe something like this in in in uh in simple face contrast movies uh this is you know 20 microns and you have uh maybe some correlations over 10 microns 20 microns and the cells are two microns long here these are white type equalized nothing special so over lamb scales of 10 cell bodies and short time scales you have so called bacterial turbulence or something like this nothing too exciting this is very very dense system in which the volume occupied with bodies of the bacteria is about 20 or 40 percent now uh if you put uh little oil droplets sitting on the surface of this here the computer is following them for you to help you you see that these droplets describe very regular elliptical trajectories and these trajectories are in face and they are in phase here over you know some tens of microns but in fact they're in phase over millimeters thousands of microns and even centimeters so here you have hundreds of millions of cells producing this collective oscillation and um I skipped this and if you try to follow now in an experiment in which only a fraction of cells are marked by fluorescence to follow some of them this is very hard to do because they're so dense so this was done manually by some poor Chinese student you can see that these traces I mean look follow any of them it's very hard to guess that these sort of trajectories produce this uh collective red elliptical motion here which is here given by some impurity floating on the surface so here I have a system where individual bacteria do something like this okay but collectively their velocities when you average them over large enough lamp scales they do they produce this collective motion so in the noise strong erratic component of these trajectories in fact there is an underlying deterministic collective mode that you don't see at all unless you know what to look for okay and I will stop here but basically what we can do is we can have a minimal simple model that shows here you can see for example uh you can see nothing that's the point of this movie but if you follow individual guys marked in colors different from the others okay you can see that you can see nothing basically the yet collectively these produce a circle here not an ellipse but anyway so you can find very simple models in which at least qualitative phenomenology is can emerge and makes it less mysterious okay of course we have a better version of the model that produces an ellipse but I will jump to this here you have you have to take into account the fluid sorry normally the two should go together and you see that if you take into account the fluid the elliptic trajectories are recovered pretty much like in the experiments in fact all right I stopped here there are a slide of conclusions here and basically my conclusions are you know okay you have new interesting or interesting physics provided by bacterial systems most of this can be accounted for by simple physical local interactions well not so local in case of a fluid but anyway of course you can wonder biological significance of this for example there's this last mysterious synchronization occurs whatever the species whatever you know almost everything provided it's dense enough suspension very dense and you know it's done it's happening as it's just before biofemes you let the colony grow and grow and grow and then they dry into biofemes before it goes into biofemes you will have these mysterious collective oscillations and we don't know the biological significance of this but my viewpoint would be that there may well be one you know and because it's a very robust phenomenon that we have observed in many species that we can perturb and it resist and so on that is not linked to the history of a colony it's really a physical thing that's happening with super super dense bacterial suspensions I stop here thanks