 Ok,so good morning,so I'm going to talk about a related topic which is thermal properties of a specific liquid which is graphene nanofluids, so my name is Paolo de Holm, but actually I'm going to talk about something that I was only incidentally involved, the work was actually done by Francesca Costanzo who is a postdoc in our institute in my group and by Ben Enzig who is a professor in the University of Amsterdam, so these are the ones that did the work. And I hope I won't say anything too stupid in my talk. So this is part of the MAX project and this is one of the parts of the project which were supposed to be linking the outputs and the ideas of MAX, the materials modelling with the needs of the users and in particular with the industry. In MAX we defined several pilot cases where we tried to address industrial problems of industrial interest and this is one that came up a few years ago related to work of a Spanish company, Avengois and it's a company on energies in general. At that time they were very much interested in studying liquids for thermal storage, so the idea is that they are company for a strong activity in solar energy, so you have these plants where you have a lot of mirrors that concentrate the light of the sun in a small spot and then you heat a fluid and store it that has the advantage that you have energy even when there is no sun, which is one of the main problems of solar energy. So you heat, you store the energy in a liquid that you store at very high temperature, so you have these huge tanks storing essentially a material which is a molten salt, a molten salt which is stored at about 500 degrees, then you use it to process electricity, it cools down and then you circulate it again to the source of heat. So you can have energy when there is no sun, this is one of the important things. So what materials do they use to store the heat, they use salts that melt and they are quite common salts that you can have chlorides like sodium chloride and so on, you can have carbonates, nitrates or mixtures of all of them, so they find which ones are the most interesting. Again they have melting points relatively high, about 200 Celsius and they operate about 500 Celsius. So of course they want to use liquids that store heat more efficiently and one way to do that is to improve the thermal properties of the liquid by including nanoparticles in the liquid. This is a very popular topic, many groups have been doing that, these are two papers that appeared when we started this project in 2014 showing that including quite small amounts of nanoparticles in the molten salts, it would increase the thermal properties quite dramatically, so there are reports of about even 100% of increasing the heat capacity of the salts just by including small percent amounts of these nanoparticles. So one main problem of this is the accurate quantification, the experiments are not so easy and there is a quite large dispersion of values, you won't see anything in this table but this is a table showing increases in the specific heat of certain materials by different authors, so typically you have amounts of nanoparticles of about one percent in weight and you may get increases in the specific heat of about a few tens of percent, so it's a quite dramatic increase. Even in some cases like this one I'm highlighting here, this is graffing nanoparticles in one of these materials, it gets about 100% of increase in the thermal properties in the specific heat. So of course these people wanted to understand why this happens so to be able to predict and to build liquids which are much better including nanoparticles. So there are questions for us where can we understand what is going on and predict the thermal properties of the liquid through simulation, can we get some basic understanding of that. These are papers that appear also in the same year so they were computing the thermal properties of the molten salts like the PV curves or the effect of inclusion of nanoparticles in the molten salts. This was a silicon nanoparticle and they saw that the structure of the liquid in the surface of the nanoparticle which goes up to here is quite dramatically affected. So there is some layering, some compressed layer that they claim is the origin of this increase in the thermal properties. So the pilot that we define is precisely to try to see with simulation how to improve and why the thermal properties of these liquids are improved by the action of the nanoparticles. This is what we proposed in Max. However life is tough sometimes and in the end of 2015 this Spanish company went into a lot of trouble. These were newspapers, titulars, so for instance in Forbes there was this one saying that Spain's renewable energy powerhouse Avengoa tears towards bankruptcy. This happened in 2015 and in 2017 it actually happened so Avengoa went away. So this one from El País, the main newspaper in the country was saying that the biggest laboratory in renewable energies in Spain had been dismantled. So the reasons for that have nothing to do with the simulation and the materials. It has to do with finances and actually the people that were running the company are now being put in trial. So this had a very dramatic consequence for us that the people working in that in the company went away of course and we had to rethink the whole project. So out of lemons we made lemonade. So we thought about what to do with this, we had invested quite a few months of time. So we look for a different system looking at the same properties and in particular we are looking now at heat transfer fluids. So these are fluids that work at room temperature instead of these molten salts that are used at very high temperatures. And they are used for many things and Stefano was mentioning a few in his talk. So for instance cooling and thermal management in devices in engines and so on. And again as I show you before graphene is one of the most popular nanoparticles that they use for these nanofluids because it seems to have a particularly large effect in the thermal properties. There is a very large literature on that and the nice thing is that there are some experimental groups at my institute who are working on that. Both of the preparation and stabilization of these nanofluids and in measuring the thermal properties. So in particular the group of Pedro Gomez at my institute is an experimental chemist. He is using organic solvents to disperse nanografite or graphene nanoflakes. So they have very stable dispersions with very low concentrations even lower than 1%. So they have concentrations from 0 to 0.05% in weight so very very dilute. But they find quite dramatic effects in the thermal properties. The size of the nanoflakes about 100 to 400 nanometers in diameter and the number of layers go from 1 to 10. So they refer to them as nanografene. I prefer to call them nanografite flakes so it's quite long. So these are some examples of the kind of results they have. So they've been studying for example the thermal heat capacity as a function of the concentration of graphite or graphene nanoparticles in the liquid. And again they have a quite dramatic change so they can reach values of about 15% for only 0.12% in weight incorporation of graphite. The thermal conductivity is also very much affected and you have values which are even larger in the enhancement of the thermal conductivity. This is work again in my institute by the group of Sotomayor who is a physicist working on thermal properties of materials. Ok so they went a bit ahead of what most people do in this field which is just compute or determine experimentally the thermal properties. They have tried to go a bit more inside and try to see what happens at the microscopic level. So one of the things they are measuring is the vibrational properties of these melt, of these solutions. So for instance they are measuring the Raman spectra and they do find that there is also quite large change in the sum of the Raman peaks with the incorporation of graphene. The Raman peaks that come from the DMF molecules so they are DMF vibrations. So for instance this peak which is at 10-90 cm. It does shift considerably when you increase the number of graphene nanoparticles in the liquid. So you have a quite sizeable effect again with a very small concentration of graphene nanoparticles. They are measuring all the things but I am going to show you just this one. Ok so the idea was to apply what we had been trying to set up with Max to do these particular liquids. And we started forgetting about the ab initio. The ab initio is quite too complicated. These systems are very large and Stefano was mentioning the hiccups that you have with ab initio. We started trying to understand things in much more simple ways with using classical molecular dynamics simulations using classical potentials. We used a few of them and in particular we are using this DMF liquid with some model nanoparticles going from single layer graphene to very small to quite a bigger to a few layers graphite nanoparticles. So we were doing molecular dynamics out of this and trying to get solutions. So one of the things in Max is we want to automatize all this process to be able to do the same thing for different materials in a quite automatic level. So we kind of make a catalogue of the steps that you have to in producing these runs. And this is just an example of one of the molecular dynamics. You haven't seen these things many times. It doesn't impress you here. But it tells you that in molecular dynamics runs which are quite long in terms of ab initio simulations, nothing really happens. I mean you have diffusion which is much longer time than the simulation time that you have here. So this is something that has to be taken into account when you want to apply ab initio to this kind of materials. Ok, so what do we get out of the simulations? The very first thing we looked at is what is the structure of the liquid when the nanoparticles are diluted. So we looked at things like the radial distribution function between the solute molecules. So for instance I'm showing here this assimilation of a single graphene nanoflake in a large number of molecules. So what we should show here is the nitrogen-nitrogen percolation function as a function of the distance from graphene. So what we are plotting is not the 3D radial distribution function but a 2D distribution function. We have a nitrogen at a certain distance from graphene. What is the number of molecules in the same plane of that molecule? Ok, and we do that as a function of the distance from the molecule to graphene. So we are seeing what is the change in the liquid structure when you get very close to graphene and when you move away. So when you move away what you have essentially is this single peak which has to do with first neighbors and then there is a liquid structure so everything is relatively flat. You can see that essentially here in this green line. So you have a first neighbors peak but then everything is relatively flat. When you move very close to graphene you have this very sharp peak of first neighbors which actually happens at a distance longer than the peaks in the liquid and then you have this very clear structure which indicates a strong order within the plane. So what is happening is that the molecules which are close to graphene in contact with graphene are very much ordered. They are not disorder in the liquid so you have a very strong first order peak but then you have second or third neighbors very clearly identifiable. When you move away from that you still see disorder for a few layers and then you go to the liquid. So there is a layering, a very clear layering effect. You have a single layer at about 3.5 amstrons then you have another layer here, another layer here and so on. So you have a very clear layering even if it is still a liquid but it's a very much structure liquid, much more structure than in the bulk. So you can see that clearly here in this contour approach this is the same thing as this one but just looking at the contour plots. So you can see very clearly graphene would be here at zero and you can see here one layer which is very clear and very structure and then you see a second and even a third layer with much less internal structure. So there is a very strong layering here and graphene affects very much the structure of the liquid a few layers away from graphene. So the thermal properties of the liquid are supposed to change not only because graphene is there but because the structure of the liquid also changes near graphene. This is just a snapshot with a very small flag of the molecules which are touching graphene and you see that all of them even if they are moving and they are actually diffusing they do have this planar structure, they are lying flat on the graphene surface and they form this kind of network around graphene which moves out of the plane very quickly. Ok so we wanted to see how is the effect of this or what is the reason for this layering and we were looking at now electronic structure calculation with DFT we see that there is very little interaction between the molecules and graphene there is in the frontier orbiters there is essentially graphene you have to go quite deep in the balance band to see some orbiters due to the molecule and there is a quite clear pi stacking so the pi orbiters of the molecule interact with the pi orbiters of graphene but it's a very weak bonding. You can see that also here, here we are using this indicator which was developed by Wei Taoyang some years ago of the reduced density gradient that allows you to identify what is the characteristics of the bond between two systems these green contours mean that you have a quite weak interaction which is actually this kind of pi stacking. Ok so one thing we did with this is try to understand if we could shed some light in the Raman spectra and we computed what is the frequency of this particular Raman peak when you have the DMF with no graphene and we have this number here there is another Raman peak that we also measured that we also predicted quite well and we were trying to see how is the effect of graphene in the molecules which are touching graphene on the frequencies actually we see that both peaks are increasing in frequency for the molecules which are in direct contact with the graphene which is in the right direction we don't think actually that this is the reason for the shift in the Raman peak because in the Raman peak you don't see a superposition of peaks you see a peak which is quite moving quite rigidly with a half width which is relatively insensitive to the concentration so we think this is not the reason we are still trying to understand what is going on you must have in order for this to be experimentally observable you must have a shift in the molecules which are also far away from graphene because the number of molecules touching graphene is very small it's just a fraction of a percent so this cannot be the reason for this whole shift peak shifting ok we also looked at what is actually industrially important which is the thermal properties and we started with the specific heat and here the same way that the thermal conductivity is difficult to get the specific heat is also quite hard what we did is ok so the problem if the liquids is that you have this structure of the density of states so if you have the density of states that you can get from the simulation from the velocity of the autocorrelation function you can get the thermal properties of course you have this structure when you have a zero frequency you have a finite density of states because you have diffusion but then you have something which is like a mixture of a gas behavior which is this one for a gas and a solid behavior for which the density of states go to zero at zero frequency now for a solid you can work out the thermodynamics using the quasi harmonic approximation it's a quite good one for the gas of course you cannot do that because you have diffusion and harmonicity but for the solid this quasi harmonic approximation allows you to include quantum effects which of course are critical for the specific heat so what we do is to follow a procedure which was proposed a few years ago by the group of Goddard in which if you can have this splitting of the vibrational density of states into a gas like behavior and a solid like behavior then you can compute thermodynamic properties by adding up these two pieces separately and for instance for the solid you can use the harmonic approximation using just the statistical quantum weights and you can get the values for the thermodynamic properties so for instance for the specific heat you would use this weighting function multiplied by this density of states that come from a solid which is this one so we apply this to this liquid, to the DMF and this is the density of states that you get from the velocity alter correlation function so if you zoom this thing here and have a logarithmic scale in the frequencies you see that the solid one is a situ so you have all these peaks here at high frequencies and you have a rotational and translational contribution that go to a finite value at finite frequencies and this is what you treat differently from the harmonic frequencies so using this way it's quite involved but you can get an estimate of the specific heat which is about 150 which compares very well with experimental value which is 148 of course this tells also that the potential we are using is quite good we have started and this is work which is unfinished just a very preliminary calculations seeing how the presence of the particles affect the specific heat and we have something which we still don't understand very much for some of the sizes we consider a particular amount of concentration of nanoparticles but we were changing the size of the particles keeping the same concentration and for some sizes of particles we do see that the nanofluid is considerably more has more specific heat, higher specific heat than the pure DMF but for some other particles which are bigger actually it goes down it goes against experiments so we are still trying to analyze and understand these results ok and finally we also did work on the thermal conductivity here you have all the problems that were mentioned by Stefano before we are using so far empirical classical force fields so we can define very easily these terms in the heat flux so we are using that for computing the heat capacity the thermal conductors but you will find the other problem that Stefano was talking about which is the noise it's very noisy if you look at timescales of a few hundreds of picosexons you have all this noise but if you look at even longer timescales it doesn't really converge it shifts and then it kind of goes wild which may have to do with instabilities in the democratic dynamics or with reentrons of particles in the other side of the box we are not sure what's going on yet but actually we have to solve these issues to provide some numbers which are reliable so we have again two issues can we achieve converse results with much shorter simulation and this was addressed in the talk of Maroni and then if we can do that can we still compute it using Avinishon models which are going to be much better than these empirical models ok so just an abstract of what I told you in Outlook we created this workflow for the automatization of the simulations of thermal properties of nanofluids we are trying to validate these procedures by comparing directly with experimental information provided by some groups in my institute we went to this graph in nanoparticles which is quite a string case you have a very large change in the thermal properties including a very small number of nanoparticles so you should be able to see something quite easily here we are trying to identify the microscopic mechanisms like this layering I talk about and we are next trying to adopt these new developments by Stefano and co-workers for the thermal conductivity of these samples and with this I conclude and I thank Max for his support