 I'm not from the life science department, so this is the first thing that it could sound strange, that the life science department is a group of researchers that do other things different than the one that I will show you now. Life science department is more related to bioinformatics, genomics and computational biochemistry, but we work in a different thing. The reason is a bit historically on our own personal histories there, but the fact is that what we do is computational engineering or computational physics. And we work on what we do is we develop code and software from the very beginning, from the very algorithms up to the parallel implementation and on the end we can think that biomechanics is one of the most beautiful fields for doing these things because of the challenges that these biological systems put on us for the on the simulation point of view. You have seen all through the talks how difficult it is and in particular maybe the closest talk that you have seen in this summer school, the closest to what we do is that of Blanca and you can see how difficult and challenging it is. It's the closest one because we work also in cardiovascular. I don't know if you have seen this TV show, it's very nice, it's a bit freaky, but well it's good. It's for people that I realized just five minutes ago that I have this supercomputing whatever so it's kind of today I get one of these t-shirts and I didn't realize that it's a supercomputing or whatever so you can think what a freaky guy but well anyway. So in this freaky TV show there is a they say these that computers are not the thing, computers are the thing that gets gets you to the thing in the sense that computers to me and to most of the people that are working in the field, computers are not that important. Computers are just the mean for doing interesting things. Well the mean is too interesting too but the fact is that we do not get distracted in these beautiful computers but we try to use them to discover things. In our department we have these research lines which are very technical related to computational mechanics. These are kind of the research lines that you can see in any computational mechanics department or group. So you don't see any bio in this any bio bio related word you know in all these because what we group thinks also is in the applications lines that we that what we do is used and because the fact is that as we are a supercomputing research center we have to collaborate with people in order to use these supercomputers and in order to use the software that we develop for these computers in different application fields and one of the main fields that we are working is biomechanics. So this is the the target of my of my talk today although as we do other things but the main application field is biomechanics. So in general the projects where we are involved in are these kind of projects where you have complexity and this complexity can come through different aspects of the of the system that you want to simulate and but the on the end what we look for is to develop efficient and accurate software that can run in supercomputers at least in large in large computers but most of the problems that we are working with are this kind of course I'm not saying that all the problems in biomechanics require large computers not at all absolutely not but we work for the problems where you really need them and I'm not saying that all the problem for instance of course we solve very small problems in order to try to evaluate what we are doing but the target is finally to solve complex complex problems. So what we do what the things that I will show you is that we do simulation tools for biomedical research we do some biomedical research of course but we try to help people to use these tools we mostly work at organ level or tissue level the problems are are complex and the fact is that we target the problems where you really need these large computational resources. So what is the kind of things that that we do well we do computational mechanics I am I am a physicist I'm not a bioengineer I hardly know biological stuff just for the that I learned from the people that work with us that really know so my field is this computational mechanics I'm a physicist by by training and and what we do is computational mechanics is kind of a well it belongs to theoretical physics it's a it's computational in the sense that it is mechanics but also but where you use computers in a in a massive way and we work in biomechanics so computational biomechanics is these are the is biomechanics but now using computers so in this very nice book and very seminal book on biomechanics these are some of the quotations that appear there which is mechanics applied to biology or the second one is is very beautiful because it tries to relate biological structure and function and this is something that will appear all along this talk so the goal of of doing this kind of thing is try is try to understand the function of these of these systems and also this is kind of a motivation in this very beautiful paper about continuum mechanics of soft tissue you see this very nice quotation that it comes a definition of biomechanics much more complex and then he says he states that that we have to take into account that what we do has a purpose and the purpose of try to look in for the improvement of the human condition to me it's also a very nice quotation and it has plenty of meaning because being a physicist we get so easily distracted by things because what we do well just when we look at the at the equations and the way to program them they are very beautiful per se so it is very nice to have also kind of a higher objective if you look at this at the normal material not a biological one you see that it has some order and it is more or less like this brick on the other hand biological tissue work more or less like this it's like a team of of wheels that are all of them with a with a certain common objective but they are very independent in the way they they feel so they in the way they are so they are they have some order but it is not of the same kind of of engineering material let's say so you see these things are kind of the basic stuff that you find in this biological tissue and the important thing is that finally what you see are emergent macroscopic macroscopic properties so the the lowest levels have a strong influence in the in the upper level in how and modulating the function of the of the biological system if you look at the organization levels and the scales you can see that for instance I was talking about the life science departments in my in my center and they work mostly in omics so you can think that more or less looking at these scales what they do is about seven orders of magnitudes of what we do that we work with tissue seven eight orders of magnitude so the the difference between one and the other is really very very large but what is interesting of biological biological stuff is that you have to take into account that the lowest level has it had a have a decisive influence and in a way you have to consider them not not by simulating them but you have to consider them you have to understand how they behave even at the lowest level in order to do a better in order to do better models of the of the upper scales also if you look at the at the organization levels this is more or less with where we are placed in at organ level however as I said before not just for the lowest level but also for the upper level you have to take into account how which is the influence of these neighboring levels too on both sides to the to the organ to the organ level and not just for understanding the properties but also try to blanket up yesterday about the danger of using the word validation and what we do and as you said evaluate not don't just validate use evaluate or assess or whatever so in order to better evaluate what you are doing it's very important to consider also what is coming from the upper level in the sense of boundary conditions for instance and boundary conditions or initial conditions or things like that that is everything that is required to more or less define the system that you are working with so our research path is then we develop these simulation tools for biomedical research and we work at different we try to cover in a way all the influences of the different level so we do simulations so we do this kind of simulations like trying to understand the the cell models and its emerging mechanics then continuum mechanics and then these couple multiphasics which is very typical of these things so first of all we do these simple computational experiments to check the models and then use the models on complex geometries this is more or less the research patch that we that we have one goal is that is something that we well we kind of sort of concept of this computational man which is the or computational human not man so the idea is try to do the best possible dummy for biomedical research and then try to adapt it to patients it is we have seen many times that the system is so complex that that people really do not understand what's going on so it is what it's maybe i'm not saying that the other that another approach is wrong but i say that we we do it a bit differently we try to better understand kind of an average person with an average system and then try to adapt it to to patients i will example i will put some examples on the soft tissue that can be applied also to this is this would be for the mechanical action in soft tissue but you can think about this also with the electro physiology or or blood flow or many things in this in this regard so how these tissue cell properties can emerge to upper levels and at what extent these response to the system to the soft tissue to this uh of this electro physiology at this organ level of tissue in general can depend on the fine details of the underlying instructor if you think about forces if you think of a mechanical of a mechanical system the response of forces the response to forces to forces depend upon internal constitution so it is it is very interesting at least to try to understand why so complex in the case of biological biological tissue you can think that soft tissue is more or less like this you have cells and extracellular matrix so if you think about from an engineering point of view this is the first thing that that you can see and say oh it is like a composite so it is like a matrix where some fibers or a different material and two both materials are related very closely related and they are and this tissue is designed in a way to get some properties from the mixture of these two materials so this is the first idea that we see from an engineering point of view say I'm I'm thinking I'm more or less understanding what's going on here but well when you study each of when you just take a very brief look at the two components you start to see the differences um well typically the cells have the same genetic information same genotype but express different genes depending on many things that that most of the people doesn't know why don't know why so this way of behaving even if they come with all the information this way of behaving depends upon some things like the surroundings and and upon how they were created let's say on the it becomes on the genetic design of the of the set but then you have this extracellular matrix which is the one that gives this strength and resilience so it is kind of if you think about well again and from an engineering point of view it's like carbon fibers and epoxy things like that um so this extracellular matrix is the one that is given this strength and resilience the more closer it's kind of closer to what is an engineering material material let's say however as you as you have seen in the talk before about the the bone you see how complex it is even if you think of this how complex it is the the material but also how complex it is to discover the properties of the material so this extracellular matrix is there for these kind these kind of things now if you look at the at plenty of studies and papers and how and even at yourself if you look at your at at a mirror you see how complex is it is very heterogeneous you have seen before in in the talk before even it is highly heterogeneous um what is interesting is that it this is also in the paper by hand for instance that simplicity arises it looks like simplicity arises from complexity because the more complexity you add in the model the more regular and homogeneous how the properties of the are the stresses filled for instance in the in the tissue which means that um if you neglect some scales let's say fine it's okay neglected but take into account why you are neglected and under which hypothesis you can neglect neglect this lower level it's highly highly non-linear in in general is in elastic and it has in elasticity and viscoelasticity um it varies in time worse it varies from individual to individual not not just through species but also from individual to individual um and i said before a small scale multifysics and strongly coupled so to us two physicists or engineers that works in computational mechanics um what i'm saying is these all these were the reasons for a computational mechanics person to start study these biological stuff because it is so challenging and so beautiful that you said you just go for it now how does it describe at continuum mechanics level how to include all this complexity well first of all you can think about levels also first of all there is this constitutive relationship in the in the case of mechanical deformation is how stresses are related to strains this is kind of in the core of all your simulations in in solid mechanics there are people that are working on to design this constitutive relationship for all their lives even without and and doing very very simple simulations and very very complex experiments to define how this constitutive relationship behave um so these relationships come from the very very deep structure of matter once you have this constitutive relationship what you can do is to plug this in a differential equation so this is going to give the spacetime evolution once you have this then you plug this and and give the spacetime evolution this is the thing that um where stats computers start to appear if you look at the at the or large computers start to appear if you just look at the first part the first part can be analyzing on a single base single cell basis and well these can be or small pieces of tissue or based on on guessing some parameters from experiments but now this is the things start to get more complex even because there with with this you you would like to plug this and try to plug this constitutive equation in the in the differential equation and try to study the temporal and spatial behavior of your system and finally it comes up that okay but it is not you don't need only one equation you need many equations to simulate the full system all of them coupled and uh they the coupling can be really tight and also coupling can be just a different problem um it's very controversial for instance not controversial but there are plenty of different models for electrophysiology plenty of different models for um cardiac mechanics but there are plenty of different models for the coupling of electrophysiology and cardiac and and cardiac mechanics so it is like you have this local behavior you have this global behavior and then you have the system behavior so what we what we are trying to do is trying to give a comprehensive comprehensive study of all these things that are coupled together um now to me to to a physicist the heart is um it's the pump is the perfect pump it's a pump that was created after millions millions of evolution and is there and it works it is extremely efficient uh extremely resilient so it is kind of the perfect pump but in any case okay it's uh perfect but it is no more than a pump so it is a physical system but it is a physical system with a problem which is again uh all the complexity this was this is written in in this uh Daniel Streeter book so it is this is a handbook of physiology section it's like the grace anatomy of the of this kind of thing so it is very it's an old book and this sentence was written in 1979 but it is absolutely um real nowadays and it it it puts us with the with this evidence with this thing that um it is not it is very sophisticated and we can describe it as a physical system but it wasn't designed by an engineer so this makes that you have to try to understand this in a in a kind of different way um it is a pump but it is so complex as as that in the books it looks like this but in the real in the real life it looks like like this it is very real because it is myself that I was it was I volunteered for being one of these uh machines and well so there is my talk is to see how committed we are in in the research that we just go through these ugly machines just to get these images so if you think about the if you think about the physical system we can describe more or less like this you have this electrophysiology that is described with this um linear and isotropic diffusion plus some very complex uh nonlinear terms um then the mechanical deformation it we solve by choosing large deformations you have nonlinear material models as it is um finite deformation it is it's nonlinear also in the geometrical point of view um blood flow it's it's a navier stocks for incompressible flow it could be a Newtonian or non-Newtonian but the fact that this Newtonian non there is is is very simple how to program one or the or the other uh with respect to all the the other things um coupling the coupling between the electrical and mechanical deformation it's uh it's um as I said before it it's is in green because you can think that it is a completely different problem also very difficult and controversial with plenty of different models um where the calcium is the key in this in this coupling and this is done on the volume that is to say for each point uh you compute these coupling and you compute the stress that is generated by this electrical propagation for each point and then you have the coupling between the mechanical deformation and the blood flow which is done through the boundaries coupling between the mechanical deformation and the blood flow can have some numerical difficulties but it is not there is no physiological model on that this is just how you try you transfer forces on the other hand everywhere in all the rest there are plenty of physiological hypotheses that you have to admit so it's a three-coupled pro in this case is is uh you have these three coupled problems they are very intimately coupled so you have this electrophysiology and solid mechanics and blood flow all of them are transient and dynamic you can do some approximations at some point of doing some kind of things depending on on on the case that you're solving you can solve the mechanical as a quasi-static or whatever but you have to take into account that all of them are transient and dynamic um we use in general one single mesh for electrophysiology and solid mechanics which we do all this is not uh this is what we do it can be done in a different way but this is just to give you an idea of what we do so when you see the next slides you will have a better idea of the things that we do um and we use one parallel code that is ours to simulate the full problem it is what we use is a staggered strategy so we solve the problem sequentially for each time step it can be done monolithically this should be another strategy but in the case in our case it's fine with the staggered is it's okay so this is staggered that each multiphysics equation is solved sequentially by blocks at each time step and then at the end of the of the time at the you have for each time step you have a iteration strategy that you can follow in order to couple everything the the good point is that on the end after all these iterations you can reach the you can reach a situation as if you have solved it monolithically the downside is that maybe you never reach this because it is not stable it is very difficult to converge but well you can do things in order to get this improved so this is again as i said before this is kind of a simplification of the of the system that you want to solve but remember the pictures i show you about this you things are much more complex than that even like this it is well you can see it's relatively complex but if you look at this well this is much more complex and we are trying we are targeting to do something that it is kind of comprehensive and trying to add more and more things to the to the code to the simulation this is also another thing that that it goes in our adn in the one that we are that my group and the people that work with me um as we as we do multiphysics we would like to put everything all together and try to solve everything all together so what we are targeting are things like this i mean these complex very um multiphysics problem so this is a very this is a very large difficulty that nature is not a cat user so you are not going to have a cat from the kind of system that you want to solve in my in my group we do and we use the code for doing several engineering stuff that are very complex too but always you have a cat um or at least you can try to have a cat like this one or even much more complex problems see for instance this one this is the city of barcelona so this is a cat of the city of barcelona including well you can see here the city and then the mountains you see the god river there you see the the harbor as you see it is indeed very very complex this cat but it is a cat so with this cat you can you can have a mesh complex one but you can have a mesh it's a lot of work that you have done and and finally you can you can come up with a very large mesh to solve a really complex problem but at least you you can have this mesh it is can be done there are plenty of things i remember one of you in in the talk asked me something and i said i will show i will show you some meshes i don't know which of uh of you were but well this is the mesh of the city uh what the just i say um side comment uh it's about this mesh is about uh 50 million elements it's not that large because it is the resolution is not that that's more but in any case if you have a cat you can do that if you have a cat you can do this also these are the things that we do this is a combustion problem in a combustor for an aircraft engine so as you see it's it is indeed very complex cat it is the geometry is very complex these are parts of the different parts of the of the geometry this part here all this is all this where fluid flow is getting into all this part this all this part so as you see it's really very complex but the problem is that you have this so how do you get how can you generate a mesh for these kind of things that is moving all the time and and and changing and and whatever well you work a lot on that we have a mesh generator that it is specifically done for doing this kind of mesh look at this beautiful example that we are doing with with people here in in like oscar and and costa here in in upf and fedrica you have seen one of the posters of this look how complex it is this is the inner part of the the inner part of the ventricles so you can have the mesh and you can see this jungle of things that are within the within the ventricle so if you would like to solve fluid mechanics within a ventricle it definitely be very important to have these kind of things it is like simulating the wind over a city you don't think that the city is just plain land you think about the buildings that are there so you need these kind of structure in order to properly simulate the fluid flow so you see how complex it is um so if you don't as you don't have the cat this means that the mesh generator tools that you use that you are going to use in your to prepare the mesh for doing your simulations is not going to be the normal one that is used by engineers now what is this kind of organ multiphysics simulation what i will going to do is just to give you some brief examples of the things that we do this is a hard these are two different different cases on the on to your right you see the typical geometry that this is used for just electrophysiology in electrophysiology the typical geometries are they cut the heart they get the ventricles and and somehow cut the heart because it and get the two ventricles because ventricles are really very kind of they were the first and they are the most important or it is not a very big field but there are plenty of people doing research on the ventricles the electrophysiology of the ventricles one good reason is that it is relatively easier to get the geometry of the ventricles than that of the atria so the first thing that has been done was to study electrophysiology back to the 60s as blanka said yesterday so well you can blanka show you yesterday plenty of images of of that for the electrophysiology the problem is that if you want to solve also now the mechanics then these geometries are very bad these kind of geometries because the heart it is not cut the heart there are plenty of things up there so if you want if you really want to solve the mechanics and really want to study how the heart behaves then you have to be very careful on the things even if you don't need them to to simulate I can talk about later about this geometry but you need to properly put the boundary conditions the kind of anchor in the heart in order to do something that it is closer to this one not to the other one if you do this kind of simulations in this incomplete geometry then you will find some very weird things even if the electrophysiology is fine but the movement of the heart is not well it's not that great so under this respect something that we do is we got this this real well real this closer to real geometry that we have there are some details that you can say well the septum is very thin things like that okay fine it is not that real but it is closer to real that the that the cat one so what we did is we get the we get this large geometry and we cut it and and we do on this geometry simulations for electron for electromechanic contraction to compare with the one that it wasn't cut and try to see what if you want to use a cat geometry a shorter geometry only of the two ventricles what are the things that you have to take into account in it so one of the things this is a very coarse mesh but in any case it gives a good idea of what's going on some of the things that are important are the boundary conditions if you do electrophysiology you have boundary conditions that are related related to electrical potential but if you if you do electromechanics now boundary conditions are of different kind you have to see you have to understand how you grab this mechanical this mechanical system in order to do it properly because it is moving so in this case we are studying different boundary conditions on the pericardium and on the on the well on the pericardium trying to see which are the good ones for studying also you can study different ways of you know as I said before these these now incomplete geometries but they're there just to to give an to give an example are colored by the electrical activation so they are deforming but they are called by all of them are colored by the electrical electrical propagation so these are two different waves two different models for electrical electrical propagation one of them is absolutely simple with no physiological meaning it is just a way of you just create a wave with the more or less with the shape that you like that you want the other one the second one is much more physiologically much more with with a much deeper physiological meaning so it is more useful if you want to really study physiological issues however you can see that at some respects both of them are contracting in the same way but there are some characteristics that are different but most of all especially for the for the entry of this wave it is more or less the the same which is the moral of this kind of studies is that depending on what you want to study you will use one model or the other on one hand and on the other hand it is very good idea to try to see which are the real differences that different models produce and not say no this is a bad model or if you're looking in paper people talking about one model I say no no this is a very bad model it's very simple it's too simple the problems that we are solving they are not simple at all so it's not good to just a priori to have some a priori of these kind of things and this is the one one of the last things that we are doing we are well there is one brilliant student here that did this simulation well this simulation was done by the by the full group but in this case the setup and and running so in this case it is something related to a company called metronic which is a this is a very this is a proof of concept it is not exactly like that not at all it's just a proof of concept so in this case what you what you have is the the ventricle that is one ventricle that is incomplete because it it ends here which is contracting and expelling the blood flow so here blood flow is computed and within you have this that is a it's a micro it's a pacemaker so this is no more than a than a proof of concept so in this in this case what what we what they wanted to study is try to see how the stresses are accumulated around this due to the fact that the ventricle is moving the ventricle is contracting the blood flow around this micro is acting against the micro and moving it the mesh is relatively coarse but it is just as i said before as a proof of concept this is other kind of things that we are using it is by is done by hasmin which is here also it's about these studies on that can be helpful to pharmaceutical industry in order to see uh how these anti arithmic drugs can act on the on on people on patients with arrhythmia so in this case it is kind of trying to reproduce experimental experimental to reproduce experiments that are really done on wedges on small pieces of tissue in the heart and trying to do the same using kind of the same protocol that they use to compute to evaluate this and to do these experiments with a big well and and it is this is very nice to try to provide to industry these kind of studies but also it is very nice for us to try to improve the models that we are using and try to evaluate the models that we are using blank i yesterday show you something that it is closely related to that but done on a single cell model in this case is this single cell model that it is used on tissue so it then it was a cell now it is tissue so you can use the same models that blank show and you can carry on the same studies but now on on these pieces these small pieces of tissue so as a sort of conclusion blank yesterday gave it was well this is the good point of being after i mean later talk because you can have you can get ideas from the previous talks so she gave a twist to his talk about to her talk about what about problems that we have in this in this domain so as most of of you are students then it's it's nice to for you to to see it and to foresee what's going what what's coming in your in your life if you do modeling if you computational mechanics sometimes people do not believe that you are doing science i mean if you have to develop computational met methods to do something usually the experimental or the medical side is kind of far away of what you do and the languages are very different so they don't see any science there they can be you as skill people doing technology or doing or programming skill programmers but they don't see the difficulties in everything in all these kind of things but not the difficulty they don't see science there but there is a very large piece of science in the things that we do so sometimes it could be a bit ungrateful because you do plenty of things and then you show these kind of things to the or you discuss these kind of things with the people that have to use it or have to compare the experiments with and and they see well it's not that good and and you've been working on that for ages and trying to get the best of it so sometimes it could be it's it's difficult as i said before it's easier for them to see the science in experiments or theoretical studies that's in computational ones also you have to take into account that what we are doing in this case is multi physics in multi physics one plus one is much much larger than two so the difficulties are not just as it is not an addition of the of the difficulties of each of the of the problems even you can solve properly one problem and the other but then when you come up to the couple system no it is not possible you are not going to have any any solution of them and not just because of the bugs but also for the because of the mathematical the computational the computational setup of the of what you're doing so if you do this tool development you first suppose that the physiological models are okay i mean you don't question about the physiological model so if you do if you show something in a conference about i don't know the 10 to share electrophysiology model someone can come and say well but this is so proceed but fine okay excellent just give a good give me a good model so i can program and analyze in my code so here the goal is we first suppose that the physiological model are okay and then we do all the other things so we develop the computational model which is the math the physics the programming visualization analysis so this is the price you pay for this comprehensive stuff is that it's under it's like this is done under this hypothesis so under this hypothesis i do all the rest so if you believe in the first hypothesis then i do then just criticize what we are doing on the at the next level visualization is very interesting problem also this is an example it's kind of all the visualization that you or most of the visualizations that you see here they look like cartoons but then they are the fact is that we have a very nice visualization team that can help us to produce very nice visualization but most of them are well all of them are based on real results so in this case this is a respiratory system simulation so you have all this respiratory system that comes from a geometry from some patients that with some people that work with us in imperial college so they gave us this this geometry and under this geometry we simulate incompressible flow with with particles so in this case particles are not done in a in a this is not not this not as a post-process step but then on the fly so while you simulate you integrate the trajectories of the particles why because what you want to study you need a very accurate model for these particles tracking because you would like to study how these particles are distributed all along the system how which is the deposition of the particles where where are they going according to their masses where they go in according to their initial distribution so that's why like like here so you study the aerosol deposition of for instance drugs for asthma for for obstructive diseases things like that so even the visualization is a complex complex thing so Blanca talked about yesterday about the you have to find the biological question so our goal is to try to help the collaborators that work with us to ask and answer these biological questions so we put ourselves the asking an answer of the computational ones so if you can do this kind of dual work then it's fine and this collaboration work is is is fine so this is the good the good point of working with people like Blanca Oscar Bart Costa has mean the good point is that they to me and this is my personal opinion they took the very ungrateful job of talking with the doctors so I don't have to talk with the doctors I prefer to talk to them so it's much better so they they take these very the dirty part so I can be with my with we can be with our computer's programming and doing these clean things no no smell no stains no cancer cancer agent stuff and things like that so now what was the use given this is another question that Blanca say yesterday what was the use given to these computational models in answering these biological questions how do you use this for doing so all thanks to data all the experiments are extreme of course you need to do these experiments they are very difficult plenty of science but then to this computational science so this is a real multi-disciplinary field so what is interesting is that as I said in one of the previous slide it's a prey it's a price to pay it's a very beautiful but the problem is that and in general people like to talk about doing multi-disciplinary but the problem with multi disciplinary is that then it is very difficult to assess because it is very difficult to find multi-disciplinary experts I know one thing one part of the problem not the other part of the problem so it is very difficult for me to assess the problem in a complete way but well and try to assess the real value of the things it's difficult for me to assess the real value of the experiments I can imagine that they are very very difficult but it is difficult for me to assess so in this case it is difficult the other way around so let me just use water as one of the this is one of the things that we do so this is a dam break example so it is water just going against the swirl so it's water is very nice because with water we can we can do kind of some analogies so like water what we have to do is try to to solute on ourselves the all the knowledge that it is coming from the different fields so it's really in that respect is really a multi disciplinary a certain Leonardo did it some time ago um which was kind of a successful guy so it is like coming back this multi disciplinary stuff is like coming back so like water also um what we have to then once you have absorbed all this knowledge and you have accepted the things that you don't know you try then so with this multi disciplinary study you try then to understand the the deep um the deep meaning of nature and the things that that you are doing so with this uh well it is yesterday Blanca said many times it is you have to be an an even in informal talks about the the uh that you have to be patient and and humble but in any case strength you have to perseverate so it is like water I would like to finish this with this um sometime you have maybe you have seen this it's very nice just go to internet put be water my friend and you will see Bruce Lee talking about water so I think that this is yes this is all right thank you thank you very much for the stock maybe partially some some some practical questions as first you say that you use your own code solving this what's the main reason why you had to develop your code because there's plenty of multi physics code around in the world why did you make again another one yes because the the we've been talking about a bit of this yesterday is that in our case the the the path that we followed from the very beginning was not that for instance I am I do fluid mechanics I would like to use or write a code it was different it was our background was in multi physics in computational mechanics and we've been working in different codes we've been developing different codes uh for doing fluid mechanics and solid mechanics and we said oh we can do a code that it is multi physics and that it is uh efficient for running in supercomputers in the same way as it is efficient for running in one with one problem it's a it should be efficient for running in several problems so this is the differential things the differential thing of most of the codes that are that are around there is no commercial code that can solve efficiently in hundreds in hundred thousand of course a multi physics problem so that was a very nice a very nice field to to work on because this this means not just program but this means that you have to study a lot about different aspects of the full computational mechanics problem like starting from the algorithms the very algorithms because they gave you a different perspective so you study different algorithms you study plenty of different computational issues like programming parallel when you're solving in parallel two different codes two different problems how to couple them in an efficient way so if you look from the with some perspective you see well it is another another group of guys doing this but if you look more carefully you say ah well but there are some uh things that are different which means that we we are not very smart guys but um uh it's uh the good point is that we work in a super computing center so they pay for doing they pay us for doing that and if you if you look that there are plenty of work for everyone there are plenty of things to be done still and partially related to that is do you develop your code differently knowing that you work in a supercomputer and that it will run on a supercomputer or totally independently and you say like I just find the best way of doing the multi physics this this is this is a very nice question because um right now it's not that much but some years ago people talk about parallelizing a code so you get a sequential code and then you say oh we have parallel machines here uh let's parallelize this code and you study how to parallelize it but in if you want to use uh 100,000 cores not 100 just hundreds of course if you do this strategy it's not very it you are not going to go very far so what you have to do really is to start from the from the very beginning thinking that you have a parallel code so you don't write a sequential code and then parallelize it you write the parallel code so this puts you in a different perspective and everything you see you you do and your program on the algorithms you try to develop are related to this so it is um in this way it kind of conditionates a lot the kind of algorithms that you're going to develop and then also when you do this it's like you have different architectures for supercomputers and they're changing over time yes how dependent is your code or how dependent or how much do you need to think about the architecture rather again than on the physics problem yes this is this is another beautiful question is that uh as as we work in a supercomputing center we can have a better idea on a better idea ahead the kind of architectures that are coming so what we try to do is we try to write the code this code must be used by people that have no uh clue on how to parallelize things but we will we like that people use the code and program things there in a way that that parallelization let's say is kind of transparent so if you do things that are very very targeted targeted to one architecture to run very very efficiently one architecture dangers are everywhere because maybe the architectures do not depend on me architecture depend on IBM Intel engineers um and be the engineers so they are going to come up with a new architecture and give give it to you and say well this this is the new architecture and you have to program everything all over again and the second thing is about the flexibility of the code so what you what you what we do is we try to do things in a way that the code is flexible that it can be um programmed by people with no uh almost no knowledge on parallelization but it can but on the other hand we have plenty of um researchers working with us that are mostly devoted to uh do the deep low-level computational things of the code in order to make it flexible and then coming to the code what's your opinion on for example open source codes and and like group coding and this kind of things to really get into it oh it's it's okay I think we have with with my friend that are the the the other leader of the group we always say that okay it's a joke but with part of of truth is that you have to program everything because when you program you understand okay this you can say but you need plenty of time and resources to do okay fine but it is if you think about this as the as the last idea is very good because with that you can start your own coding um there are plenty of codes everywhere for doing plenty of things open source so you are sometimes researchers are tempted to say why should they program these if there are so many codes there try it it's good to try this right now you have matlab and you have this kind of um matematica you have this kind of tools where you can do your own little little things there and even you can grow uh larger and then you can come you then you can use other people codes and especially the good point of open source is that you can see what they do because the commercial code sometimes you don't know what they do so you get the results and well these results are okay because everyone uses these uh ansies or abacus everyone use abacus then abacus is is correct no you don't know what is inside so well thank you mariano for this brilliant talk uh indeed this question a little bit related to to this uh you didn't comment anything about uh the verification and the possible need of of code certification especially when you're simulating like strong couplings where you need to have your own solvers and uh so but not only is the processing code but also the post processing individualization where you have quite a quick comment on that please yes yes yes um as we if if you solve very complex a hard beating things like that the validation on a on a case with a hard beating is is relatively um loose because the it's kind of fine direct but in order to reach to do to perform these simulations first of all you have to pass plenty of these validation stuff uh now real validation because it is like uh you solve um simpler problems academic simple problems and try to uh repeat either experimental results when they are for simple problems you try to repeat um um there is a way of doing this kind of thing that is called manufacture solutions that you do this uh you you say this is the problem i want to solve this is the solution i would like to uh to attain so you program this in your code and and you see that you can and you see that you attain this solution then there are other uh things that you can do is that to compare with the other codes that are around and there are for instance um in electrophysiology it was a paper done by by people from king's college that they do they propose a relatively simple test and say well everyone to solve this test and try to and there there were plenty of differences for everyone but then try to explain these differences so the what i what i would like to to to come up to is that if you want to to assess uh fluid mechanics for instance turbulence or fluid mechanics you have some very um well established way of validating but with biological stuff like biomedical simulations the way of validating is uh the process of validating is more complex i don't see that it is different i wouldn't say that it is not possible but it is different and we try to do our best for doing this kind of thing so let's say at low level it's relatively validated and validated every time so at higher level it's more complex but we try other questions thank you very much for the talk it's a general question on computational fluid dynamics but specifically related to the particle tracking on turbulent flow i would like to know your opinion comparing traditional approaches like solving the navier stokes conservation equations or lattice Boltzmann yes um well the lattice Boltzmann is uh um it's it's very good this is a personal opinion lattice Boltzmann is fantastic for microfluidics for things that are really small um i we didn't we don't have in the code lattice Boltzmann so we use now traditional more traditional navier stokes the fact is that we prefer navier stokes for that's because our training comes from that but also because uh in navier stokes the navier stokes equations are solid are rock solid there's something that is there and you are absolutely sure that what you are solving is within the ranges within the range application range of navier stokes people would say Boltzmann equation is rock solid too fine it's rock solid too for um at a very very lower level a much much lower level so in the ranges that they that people use in general lattice Boltzmann well lattice comes from the discretization of the Boltzmann equation and it is a particular discretization so at the range that people use right now lattice Boltzmann there are plenty of examples that they work well and they and you can have good results on that but sometimes to me it's a bit difficult to understand i cannot follow all the steps because some dark points and obscure points on the other hand with the navier stokes based code even if it is finite differences finite volumes finite elements spectral methods whatever uh it's easier to follow you can rewrite and reprogram everything and getting the same results with no strange constants put there here and there but this is a very personal opinion but in my opinion for microfluidics lattice Boltzmann is very good this is to say the higher the Reynolds the better is the use of navier stokes even if you have can have very nice results we'll let this work on i i won't say thanks for the great talk just a question or comment because we present like a causality that okay the physiology model is okay and then i do all my simulations but i would rather say is this physiological model okay for the question i want to look at in the end you are in the unique position to look at all these integrated results yes yes yes and go back to the model and maybe simplify it exactly yes yes yes i i did this simplified idea but it is exactly as you said is um is that like i have some hypotheses so under this hypothesis i program so i do the model i then start to the to do this validation and to try these different problems and then with that i can i can come up with let us suppose that you you are the guy that developed the physiological models then you come you come to me you can come to me and say well now run these cases to see which are the things that i have to correct or modify my physiological model so in this sense it is a very it's very rewarding because we can work a lot with people with people that develop the physiological models so it yes it is exactly as you said thank any other questions okay so thank you again