 Exactly. If you have some other comment or something that is not clear, just write us in the chat. Okay. Thank you both. Okay. It's okay. There was also another question regarding, in general, if CP2K, for example, is a quantum code that has already QMM, some QMM features, let us say, inside. Why? What is the reason to do an interface? And then was mentioned also that other codes like CPMD that has QMM inside and there is Mimic, which is another interface that was devised to couple CPMD to Gromax as well. And in this, during the recent year, other interfaces of this kind that couple in particular Gromax, very popular, classical molecular dynamics code, are coming out, coming up. So what is, in general, the reason why we should make this interface and we should use this interface? I would say that, in general, there are three main reasons why you want to develop a new interface that can couple a quantum code with a classical molecular dynamics one. The first one is performance. Quite often, not always, but quite often the QMM routines inside the quantum code are, let us say, not, the performance are not so good in particular could be, for example, not so good in strong scaling or not so good in quick scaling. So, which means scaling with respect to the size of the system. And so you want to improve somehow the performance of your code because what you would like is to be able to do QMM with system which are larger and larger because quite often during the year, let us say, in the years, your interface is on system that are larger and larger. Another motivation is usually because the features of the QMM inside the quantum code could be limited. And of course, coupling with the popular program like Gromax, which has lots and lots of tools inside to do analysis, etc., can, of course, be very, very useful because it allows a user to have a plethora of new analysis tools. The third reason, even if it's not the case for the interface of CP2K to Gromax, but for example, it was a reason for Mimic, was licenses copyright reason. For example, in CPMD, CPMD is free for all academic users, etc., but the QMM interface inside the code is not free, you have to pay a small, but you have to pay a license of Gromax in order to use it. The idea of making an interface which was totally free as well with CPMD was in this way in order that you could use a, by the way, a modern, more updated version of QMM with CPMD and that could be totally free and so it could be available to anyone without any restriction. And regarding Mimic in particular that was mentioned in the chat, as I wrote in the chat, the official release of Mimic with some support, let's say, will be very short time, probably at the beginning of May. And so I would expect that it will be announced or the CPMD website, probably also on the channels of BioXL, so just stay tuned if you want also to give a try to this other interface. Dimitri, you have something to add? First of all, why we started developing new interface because as already wrote in the chat previously by some other people that old QMM interface which was in Gromax since Gromax 3 or even earlier, it is now having decommissioned, so it is not, the newest version of Gromax does not have it. The second one is that the Gromax now does not support, for now does not support on periodic system, so UQMM code should also be periodic. One of the possibilities was here CP2K, because it has a periodic QMM. The third one that this interface which is developed now for CP2K could be further used for another packages in a later use, so it made in a way that it could be extendable to other QMM packages. There is no concrete plans on that, but yeah, I mean it's free to develop. And of course, another point was why we should use Gromax CP2K interface instead of just pure CP2K QMM. Well, first of all, I can say you that if you're used at any point in the CP2K, you probably know how painful it is to set up the QMM system in CP2K. First of all, your Gromax topology will not fit. You need to convert your topology into charm topology or unburnt PRM top topology. And it means already that you have a different systems for your Gromax and dissimulation and for your CP2K QMM simulation. And why it's important, I can tell you that, for example, since Gromax, since CP2K 304, something's already more than 10 years, CP2K, for example, had a bug in under topology reading and PRM top files. And it's have been only fixed in January in CP2K 8.1 and only because we start comparing our interface with CP2K implementation. And we found that the CP2K reads PRM top files wrongly. And yeah, probably everything which had been done for 10 years with CP2K QMM and using the unburnt topology was wrong. Sorry, but yeah, it's fixed now. But that kind of stuff, and I know why it's happening because not many people are using QMM and CP2K and not all of them are aware of such kind of things. And CP2K developers also did not know about that problem. So yeah, why? Because Gromax provides a way how, at least in Gromax, you can be sure that your force field is correctly used and you have a unified topology for both QMM, classical and dissimulations and QMM simulations. Yes. So, yes, so already some people writing in the chat that they agree that using CP2K is painful right? Yes. And more importantly that we want to provide user, we want to push the QMM into the larger community which is obviously usually uses Gromax. Yeah, so the Gromax community is much larger than the QMM community and we just want to push that you have kind of an option how to pretty easily move from completely and dissimulation, classical dissimulation to QMM simulation with just a few parameters. And most of other will be generated automatically to something which should make sense. It is not always best parameters which you can use, but it will be something which makes sense. And yeah, as a starting point, but I mean tomorrow I'll probably show you how you should check your parameters more precisely, but as an initial point the standard parameters should be okay. Percent of the cases I can say. Yeah, so this short answer to the question of the people. I also saw another question regarding the old Gromax QMM interface. It now have been decommissioned so you cannot, in the modern version of Gromax, you don't have interface with Gosha, with Orca, with everything like that, with games. Yeah, so for if you want to steal to use that it is possible but you need to use all the versions of Gromax 2018 I think is the last one here. I wish you can perform that simulations. Can we use MKL instead of BLAST and ScalaPak separately also? Yes, you can. If you know how to compile Gromax with MKL and you know how to compile CP2K with MKL, then you can do that. No problem, but yeah. Dimitri? Do you really want to do so, yeah? I also have a question about the water model a little bit. Ah, yes, yes, yes, I saw that. Okay, so the question about keep 3P and or keep 4P water model. For now, I would say, I would say that of course the keep 4P water model is better than the keep 3P, but it comes at the price. In keep 4P, keep 4P water model is much more heavy. It's 33% heavier in terms of computation. So you need to be aware of that. I mean, in most of the cases, what you want to do in QMM, so usually what I'm doing personally, you do a simulations, normal Gromax simulations and dissimulations with keep 3P water. And then important waters, which are nearby the active site of the protein in QMM, you just switch them to QM. So basically you make a QM waters. But in all other cases, it might be very important in some system to use also keep 4P water. So actually the general question that Thomas is asking, which is that, would your answer be the same? Would exactly, what is that question? Your answer applies to the water. Okay, yes, it will be in general, it could be important because in QMM you have this, for example, dependence on the point charges. Of course it can, but you, in order to leave that out, you need to check how your results are converging with respect to the size of your QM system. There is a lot of papers, I guess, in that sense, that you always, when you're doing QMM, you always need to check if your QM system makes sense. I mean, it means that if you extend your QM system, so you put in more and more. So you are kind of reducing the effect of the force field, but increasing the effort to QM and you should check whether you have a consistent result with your current size of your QM system. So for example, I can say that if you, like if you have a QM system and you have several hydrogen bonds with MM waters around here, that usually if you switch that water from MM to QM, you immediately see how your system will start changing. Why? Because typically your QM system interacts much strongly with MM point charges than with another QM system. You see my point, so basically what will happen if you switch waters from MM to QM, they will probably try to swap with another MM water. Because MM waters have a strong interaction with QM system than QM water, another QM water. So it's a really tricky question and there is no clear answer to that even now, but there is a lot of papers on that topic in reality. And in general, I would like to add that in general, if you think about how usually QMM schemes are devised, you have from one side the classical force field and on the other side the quantum system. The classical force field was devised really all the parameters of the classical force field to do, let's say to take into account all van der Waals dispersion interaction and so on, everything without considering that you have a QM system somewhere. Ideally, to do things in a very accurate way, you should re-parameterize all your classical force field considering that now you have to deal with the QMM system. But of course, no one does it because it would be extremely long in terms of time and all the gain that you have it, you would do it in this way. But from the theoretical point of view, this should be the right thing to do it. And since we have to reach a compromise, so you can still using a classical force field parameterize with only with classical molecular dynamics, let's say for classical molecular dynamics. Reaching this compromise is possible but you have always to verify that what you are, let us say, what you are doing in terms of convergence, etc. makes sense. Can you replace QMMMD polar force field in classical MD to make the calculation cheap QMMMD with a polar polarizable force field in classical MD? Okay, so the question should be, if I understand well, correct me if I'm wrong, Shiv, could I do a classical molecular dynamic with polarizable force field in place of a QMMM molecular dynamics? The answer is, in principle, you could do it. Of course, it depends which are the physical quantity that you are interested in. If your physical quantity are well represented by a classical molecular dynamics, in principle, yes. I have to say that first that, however, I have seen so far very limited, let us say, application of polarizable force field in the current community. Which means that, of course, the polarizable force fields are still in development, so then probably they have to be improved from this point of view and they will do it in the next year. But as far as I can see from the current literature, it seems to me that they are not still at the level of accuracy, good enough, let us say, to compare in terms of accuracy with QMMM approaches yet. Maybe in future, or maybe now that the application with the polarizable force will become more and more popular, I don't know. This could be the case. Currently, it seems to me that is not yet the case. Okay, and I cannot hear that you can use polarizable force fields in principle, but you still cannot study the chemical reactions. Because bond breaking is still not here, and it will be not here because the bonds have a quantum nature and without correct electron modeling, without the QM, you cannot accurately and universally make reactive force field. There is a reactive force field, of course, but they are still a lack of, they still only works in a way that they are parametrized. QM is not dependent on the parametrization usually. There is also a question about QM waters, which I know, Emiliano, at another point you discussed how to deal with QM waters, maybe relieving or the site of interest, just above. Okay, this is the question on Unmesh regarding the fact that the quantum box, usually in the schemes, for example, that you have seen in the QM scheme that you have seen in the presentation, in the second lecture, has a fixed quantum box, which means that the size of this box is decided at the beginning, while the content inside this box can change. And what happens if I have in my box, at the beginning, 10 water molecules, and at the end I have 20 of them, or maybe less, 5 only, etc. How can I describe that in an accurate way? The answer is that so far, any QM scheme devised so far, I would say, has this drawback. And so you have to deal a priori, at the beginning, about how big should be your quantum box, quantum part, in order to be sure that during your simulation, all the relevant parts stay within the quantum box. There are, of course, in principle, theoretically, ideas about how to correctly deal a case like that. These are, what is the name, I would say that are called adaptive QMM scheme, adaptive is the key word, to work really with this approach. For example, these are such a kind of things who have been done at a classical force field level, when you have, for example, classical molecular dynamics, plus, for example, coarse-grain approaches, so to an interface between a full quantum and coarse-grain approach and particles that are flowing from one area to another. And this has been done in a very, let's say, accurate way. There is a theoretical framework that allows to deal with that in that case. And the same things we would like to do it also when there is a QMM system. And so there are these two kind of parts in your system that have to be interface and where also particles can move from one area to the other. As far as I know, there is not yet a theoretical framework for a QMM approach that is working. So people are investigating on that. There are a few groups in Europe that are working on this problem, very complicated problem from theoretical point of view. I would expect that in a few years we will get it, but no solution yet in terms of treated theoretical level in a correct way. So what are the ways to overcome a problem like that one? Of course I told at the beginning is to try to think at your system a priori how big should be, which content should be in the quantum six, etc. to avoid as much as possible the problem. Moreover, you can sometimes in literature people try to force somehow the system to remain in the way that you want. By adding some barriers, for example, close to the wall of the quantum part, of course these are approaches that have been used in the literature, but of course they introduce additional artifacts that have to be there with in the sense that all the artifacts that you produce have to be controlled, verified, that they do not modify significantly the physical quantities that you are interested in. Okay, and I think that's all about the questions in the sense of, yeah, there is only leaving the questions about the first practical part. Can we answer them as now? I think, well, there may be this one follow-on from what the Milano said about this. Unmesh asked actually, so if you did have a case where you wanted to actively track the waters, so there's a follow-on from that, maybe it's possible, I don't know if it's possible to... The problem is how you can couple, yeah. The problem here, you can track the waters of course, but the problem is that quantum and mechanical, molecular and mechanical waters, they have a completely different property. So once you will make your water, once you switch your water on the fly from mechanical, classical mechanics from MD to QM, it will cause a huge jump in energy. And what will happen after that? Well, probably the simulation will start breaking out. You have a temperature rising immediately. You have whatever problems you have just because you're swapped. Yeah, so the good question on the Milano said there is no good answer how to do that. If you want to study proton transfer actually, Unmesh, I have a paper for you in the chat now. They studied the transport of protons through the gramisidine channel using actually CP2K. So you can also check that if you're interested. But it's channel, so there is a water channel, a channel with waters and you pull the proton through that channel hoping from one water to another. So this idea. Yeah.