 to you, our bright current diffraction imaging can be applied to image particles on the RISD conditions. So this is part of the ESC project, Ro Carine, and it's a collaboration between CEA, the European Synchrotron, the Technion, XMASA University, I have 2NP, the LEPMI, and DESI. And in this recent day, so I had more to do with X-ray absorption, here you can find a current X-ray diffraction, so here you can see some X-rays that are from my relatives, and I can say that this demonstrates the power of X-rays in medicine. So now, going back to the subject, so first I will explain you the motivation of the work, then the results of nano-focused bright current imaging for hands catalysis, and I will end with a summary. So catalysis is everywhere in our lives, it's about 90% of chemical manufacturing processes, and now it's well-reliased at the surface, like in strain, the deformation can be caused by surface relaxation and as a tropics strain, strain on ground boundaries, strain due to some cautious structure, as well as defects. There are key components for catalysis, and quite recently appeared the concept of strain engineering catalysis. So you can tune the strain in order to improve the catalytic performances of the particle. So the structure, meaning the type of assets, defects, crystallinity and strain, can be key parameter for the activity, selectivity and reusability of nano-catalyst. Here I show you an example from a science paper during a reaction which is the oxygen reduction reaction, what they have done, so they have a cautious structure and they succeeded to create a tensile strain in the shell and this decreased the catalytic activity by 40% then they perform a compressive strain and this boosted the activity by 90%. So now it's well realized that the structure can be a key for the performance of nano-catalyst, but there are still some challenges which are to resolve the structure in real environments and at relevant scale and to develop some predictive models. And by chance, we have access to one technique which is called Brack or Heron diffraction imaging, which has been developed 10 years ago by Jan Robinson and their co-workers. So we need some coherent illumination, a beam size which is larger than the particle. Now it can be applied in situ or per handle in complex active liquid gas environments and what we can get, it's some dynamical structural imaging at the nano-scale. So we can get the strength, composition, shape, defects and diffusion. Most of the experiments I'm doing are performed at the ID1 beam line. Here you can see a sketch of this beam line and the diffractometer is positioned at a distance of about 120 meter from the source. I'm also applying this technique to over synchrotrons can be I mean daisy or soley where I'm doing also experiments. So how does it work? Our Brack or Heron diffraction imaging is sensitive to strain. So here I show you an example. So you can see a particle, so platinum particle. Here it has a size of about 300 nanometer. Then you can see the 3D Brack peak. So we are measuring in five minutes. The working curve is around two degrees. As you know, we are just measuring only measuring the intensity, so the phase is lost. So we need to use some phas or trivial algorithms. So we are using a pionics in our case in order to get the particle in real space. So what we can get is the electron density, meaning the morphology. Here you can see a top view of the particle, a bottom view. So at the beginning we didn't know that there was a hole at the bottom of the particle and we check with scanning electron microscopy and this has been confirmed. And so what is very interesting is that we can get the phase and the phase is proportional to the atomic displacement here in picometer. So we can access to the atomic displacement and to the strain. Here I give you the voxel size of the reconstruction. If we look more closely at the displacement, we can see here that the corners, edges are in blue color and the facets in red color, meaning that the corner and edges are in compression and here the facets are in tension. And this is quite important for catalysis because this can modify the absorption energies of reactants. So in a few words of the technique, it's a 3D technique. It has a spatial resolution of 10 nanometers. So we are working to improve the spatial resolution with a new source and it has a strength sensitivity of about few 10 to the minus four. So now the question, how can these results, these experimental results compare with atomistic simulations? So here I show you the work of Maxime Duprasse. So I told you that we can get the morphology of the particle, what is called the support from phase retrieval. We have a resolution of about 10 nanometers, but we can feel the support with atoms and then it's possible to do some atomistic relaxation. And with that, we can get the out of plane, the displacement. Here I show you the out of plane displacement and we can take the derivative to get access to the local strain. So we made a comparison between experiments and simulation. Here it's the same particle with the same shape. And if we look at the size of the particle, this particle is a little smaller than the measure of particle because at this time it was not possible to simulate a particle as large as the experimental particle. What I show you here is the strain, the out of plane strain along the 111 direction. And if we look at some of the facets here, I mean, in the experiments and in the simulation, there are in a tension here, the 113, if we focus on the 111 here, there are in compression. So we can really see that the strain state is well captured by atomistic simulations. So we can investigate the surface relaxation of platinum nanoparticles. Here I show you another view of the particle to look more closely at this comparison. So this is a very good point, so it's sensitive to strain and we can look at the surface relaxation of platinum nanoparticles. With our resolution. So as the technique is quite fast, we can do some particle statistics. So we can look at particle, I mean, here we can see particle with one defect. We can see a lot of twin platinum particles or some twin particles and with defects. So at the beginning we were quite surprised to see some twin platinum particles because twinning is very rare in bulk platinum. If we compare to over material, platinum has a very high twin boundary energy. And but so in bulk it's very rare, but we can see that in nanoparticles, we have a lot of these twin particles. And this technique is very interesting to look at defects and here I show you an example of this location loop. So here I've shown you some result with where we are looking at just one reflection. And nowadays we are looking now at multiple reflection to get the strain in all the directions. So here I show you this multi-reflection. So some measurements that we have performed on a nickel particle and this nickel particle has here a nano twin here. So we have this empty, so this void at this position because this part is not diffracting at the same position in reciprocal space. So it was possible to measure several reflections. So now we are doing multiple reflection quite often to get the 3D strain state of the particle. So I show you here an example of this 3D strain tensor. So you can get the strain tensor, I mean the different components of the strain along the different directions. We also perform some simulation. So here we have simulated the out of plane strain so along the z direction. And this is in good agreement with our measurements for this nickel particle with a nano train. We can also get access to the 3D atomic displacements around the defects. So it's quite interesting. And here this work has been performed with a special resolution of 10 nanometer. So here I show you some examples, some X2 example and the current project is really to do, to look at, I mean to look at reaction in situ and operando. So to do it, to have access at the 3D structure, the strain and defects, at the same time monitor the activity of the particles and to develop some predictive model. For this, we need some tools, compatible with the nano beam. So we have developed in collaboration with T1 Dovan with ESRF, we have developed some tools like gas reactor. So you can see here an example. So of this gas reactor, it can go up to a temperature of 950 degrees. It's very interesting because it's 3D printed. I mean here you can see the water cooling. So the temperature, the stability of the temperature, it's quite stable. And this is really necessary for nano diffraction experiments. And here you can see perhaps it's difficult to see it, but the reactor is installed at the ID1 beam line. And here you have the gas panel with all the gas bottles and all the mass flow controllers to control the gas reaction. So for the in situal parameter experiments, we have developed a gas reactor. And in collaboration with T1 Dovan, we have developed an electrochemical cell and we have a commercial battery cell from the University of Picardie. And all these equipments, they are all compatible with nano focus X-ray beams. So first, I want to show you some results that we obtain during gas reaction. So we look at one reaction, which is the CO oxidation reaction, which is to convert the toxic CO into CO2, which is used in the hexo system. And first for this, I will show you some of the results. So during this CO oxidation, I will show you how the twin boundary can migrate. So here you can see the gas reactor. So we had to work at a temperature of 150 degrees. So the particles are active. I show you an example of the diffraction pattern. And this is the diffraction pattern of a twin crystal. So you can see here the 111 streak and this interface here, the signal of this interface, the other part can be measured for at the 111 reflection because it's 115 terminated. So we had a look at the evolution of the diffraction pattern as a function of gas. If we look at the diffraction pattern, we can see that at the beginning we have this nice tilted streak. And at the end of the reaction, we have a flat here horizontal streak. So really the structure, I mean the morphology of the particle it evolves during the reaction. And as we have some 3D diffraction pattern, we can get the retrieve a particle in real space. So we can get it as a function of gases. And what we can see here is that the volume of the particle here is changing during the reaction. And what we have observed is that when we flow CO, we can see some twinning phenomenon. So we have a reduction here. We have some twinning. So some migration of the twin boundary. And here for this gas condition and also at this gas condition. And then there is a detwining of the particle. So how to explain the driving force of this twinning? So for this, we had a collaboration with the CMAP and they have made some DFT calculations. So we know that we have two types of first surfaces. So for the twin crystal, so a 111 surface and for the over grain connected grain, it's a 115 surface. They have calculated the coverage, the expected coverage for the experimental temperature and pressure. And they obtain a lower coverage for the 111 platinum surfaces, so a lower coverage of CO. And if we look at the interfacial energy of CO, we can see that it's preferentially here it's lower in the case of a 511 surface. So meaning that we have a lower 511 interface energy for CO. And we think that this is the chemical driving force for this twin boundary migration. I saw quite interesting if we look more closely at the particle, we can see that at the beginning under CO, the particle is more, I mean, is more roundish. And after when we are inject some oxygen, we can see some very small facets forming on the particle. And perhaps here it's easier to see these small facets. And so we have some diffusion material transport during the reaction. And we can make a comparison with a work which has been performed previously by TM and by microkinetic simulations. So they had a look at the CO oxidation and it was for smaller particles. And what they observe is that when the gas, I mean, if it's in a very rich, a CO rich environment here, we are in a CO rich environment. And here it's their simulation. They see that we can, we should get, I mean a roundish particle here it's black. So meaning that we should get some more roundish particle, more round particle. And when we have less CO, here it's gray, meaning that we should get some more faceted particle. So in this case, we are more, we have some oxygen inside the reactor. So we are less, we are more, we are sorry, more CO poor. So we can get the same equivalent results with some faceted particles. So this was also quite interesting. So it was possible to connect the mass spectrometer. So we get the product of the reaction and it was possible to get an idea of this product of fraction of the CO2 as a function of a different gas condition. Here for sure, I mean, it's the results of the ensemble of particles. It's not on a single particle, but it's at least interesting to see if the particles are reactive or not. So here we have seen some defect dynamics, some twin boundary migration and some facet dynamics during the CO oxidation. Now I want to show you one of our work, another work on facet-dependent strain evolution. I mean the different particles have a lot of facets and how the strain can evolve depending on the orientation of the different facets. So this was the black time of SRF. So we went to the daisy synchrotron, so to the P10 beam line. So we install our reactor gas panel and we perform a silver CO oxidation. And here, so we work at 450 degrees. We again look at the 111 reflection. Here you can see the diffraction patterns and during the reaction, so here it's the stoichiometric CO oxidation reaction. We can see a clear change of the diffraction pattern compared to pure atmosphere, inert, sorry, atmosphere or oxygen rich atmosphere. So it's very interesting just to have a look at the diffraction pattern and to see that there are some evolutions during this reaction. What was very nice, so perhaps you saw it here, it's that we have a lot of streaks, so meaning a lot of facets for the particle. So we measured three particles, so a particle with the size of 300 nanometers, 650 nanometers and 700 nanometers. And it was possible to index all the facets of these particles. So either we can index in reciprocal space by doing a polar figure and we have all these streaks and by getting the polar figure, we can get all the directions, the normal of the facets or as we have the retrieved particle in real space, we can do the facet indexation using facet analyzer. So this is the plugin of ParaView. So it was possible to index all these three, I mean, index all the facets of these three particles. And here I show you all the different facets of the table. So we had some facets like 111, 100, 110, 113. So for instance, for the particle with a diameter of 300 nanometers, we have 32 facets. And for the particle with a diameter of 650 nanometers, we had 43 facets. But this is very interesting because we will see all the different types of facets that can evolve during the reaction. So let's have a look at the small particle. So it's like that we have a movie of the evolution of the morphology and of displacement of the particle as a function of the different gas conditions. Here I show you the first measurement in this gas condition and the last measurement of the gas condition. So first what we can see. So here it's the bottom of the particle. Here is the top. So if we look at the bottom of the particle, we can see very strong evolution of the displacement of the strain at the bottom of the particle. So we note that there is an interfacial dislocation network. So that means that the gas can diffuse inside this interfacial dislocation network. So this was quite interesting because sometimes in TM it's very difficult to know what's happening at the bottom of the particle. Then if we focus on the reaction on the stoichiometric CO oxidation, we can see a strain, that strain also opt during the CO oxidation and it's very, very impressive to see it. Then at the same time, we have a diffraction pattern. So with the diffraction pattern, we can get the full width at the maximum of the diffraction pattern. We can get the average strain. So there is no change of the average strain. From this retrieved particle, we can get the strain-filled energy. And here what we can see is that when we are in the condition of CO oxidation, so it's here, we have an increase of the full width at half maximum of the diffraction pattern and also an increase of the strain-filled energy. So now we focus more on what's happening during oxygen, I mean just when we are flowing oxygen. So do we have oxygen absorption or not? So we'll see in the next slide. So here it's just at the beginning. So we have our particle in inert gas. Here we are flowing oxygen. And just by, I mean, while flowing oxygen, we can see a change of the strain. So here we are looking at the strain along the 1-1-1 direction. So we can see a change of the strain on the surface of the particle. And it's reproducible. I mean, here we have just done two cycles, but for the second cycles, we have seen the same phenomenon. And here, so the color is the strain. And if you look here, you can see different colors there. And this is just the difference of the strain between the last state here and this initial state. And what we observe during absorption, we can see, so I will tell you, that the strain evolution is facet dependent during oxygen absorption. In fact, we have seen that all the 1-1-1 facets, they will go here in red, they will go into tension. And the 1-1-1 facets and 1-1-1-3 facets, they will go into compression. So we have seen this for the different particles. So we can say that during oxygen absorption, the strain is really facet dependent. And so how can we understand this? So we know that oxygen will preferentially adsorbed on corner and edges. And then they will preferentially adsorb on the 1-1-3 and 1-1-0 facets. And this explains why they are in compression and why oxygen with preferentially adsorbed on the 1-1-3 and 1-0 facets, it's because they have a lower coordination numbers than the 1-1-1 facets. And then to confirm this, we have performed, I mean, some DFT calculations. I mean, it's Quentin Chatelier, who is part of the team. He has made some DFT calculation. What he has observed, I mean, is that here, you can see the strain for the 1-1 facets. And given the strain value, we assume that for the 1-1 facets, we have a quite low coverage of oxygen. And here, I mean, for the 1-1-0 facets, if we look at the different results of the strain, we expect a very high, higher coverage of oxygen for these facets. So this also confirms that the oxygen will preferentially adsorbed on the 1-1-3 and 1-0 facets. And this will lead to this compression. So during oxygen adsorption, I mean, it's very facet-dependent. Here, I show you the two cycles. We are here, the initial state, it's a metallic state. Then we have some oxygen adsorption, some change of the structure of the particle. Then we can reduce here by doing some CO oxidation. And then we have some oxygen desorption and we are back to the initial metallic state. So this is quite interesting just to see how adsorption occurs on the surface of the particle. Now, let's have a look at the CO oxidation. How it works? Is it facet-dependent, the CO oxidation? So for this, we have calculated the average displacement for different types of facets. Here, so for different types of 1-1-0 facets and for different types of 1-1-3 facets. And what we observe is that here, if we look at the 1-1-0 facets here, we can see that some facets will not evolve. I mean, they will not evolve. Some, they will go in tension, some in compression. So during CO oxidation, it's no more facet-dependent. So it's no more facet-dependent. We have some in-equal reactivity of in-antical facets. And if we look more closely, so if we look here at this type of facets, it's 0-1-1 or this facet here, which is minus 1-0-1, it's located here and it's close to the substrate. This is the same thing here. If we look at these blue points, it's the minus 1-1-1 facet and it's located close to the substrate. So what we have observed is that we have larger strain variations on the facets close to the substrate when we are doing this CO oxidation reaction. So this is support-dependent. So in a few words, we have seen some strong local structural changes associated to chemical interactions. The strain evolution is facet-dependent during oxygen adsorption and it's super-dependent during CO oxidation. So we have some metal support interaction which is important for performance in catalysis and this can give some new insights on the active sites during the reaction and the relation between the strain and chemistry. Now I will show you what we can get in the liquid conditions. So in electrochemical condition, I'll show you some example during gas experiments. So how is it during liquid experiments? And this is the work of Clemol. So here we have an electrochemical cell which is mounted at the ideal and beam line. The idea is to measure one single particle in the sulfuric acid electrolytes. Here you can see a CV curve. So it's typical of platinum and it's what has been obtained during the measurement. I mean, we have a potential stat which is connected to the electrochemical cell and the CV curve is really in the agreement with what can be obtained, I mean in a electrochemical lab. So we have looked at the evolution of the platinum particle and at the evolution of this strain as a function of applied potential, you can see here the evolution of the potential. So it's in a region which is called double layer region. And if you look here at the strain, so it's along the OO2 direction which is here vertical, we can see a real change of the strain of this particle during electrochemistry as a function of applied potential. Here I show you just a slice of the platinum constructed particle. And if we look more at the corner of the particle, we can see that it's more and more displaced here. We can see the arrows or more and more blue. Meaning, I mean in this region, we note that there are some B sulfate adsorption. So here we can see that the bisulfate here will preferentially adsorb on the corners. So I show you how the strain evolve and if we extract the strain at the edge and corner, this is going really in compression. Adversary effects is going more in tension. So here I will not say a lot about this, just to say that we have some potential dependence strain distribution between the facets which are highly coordinating atoms and the under coordinated atoms which are the edges and corners. And we can see it quite well during an electrochemical reaction in 3D or through these slices. Now I want to say some words about machine learning and defect recognition. So I show you some very nice examples of retrieved particles, but we all know that phasotrival has some limits which are that I mean phasotrival doesn't always convert. It's the case of highly strengthened particles or some particles with a lot of defects. So what was the idea? It was to stay in the four year space because there have been very nice demonstration in the paper of Maxime Dupre. He has shown, I mean here that the defects that very defined signature in the four year space. So the idea was to develop a convolutional neural network for defect recognition from this 3D, from 3D current diffraction patterns. So to stay in four year space and just by looking at one 3D current diffraction patterns to know if there is or not a defect, a dislocation inside this particle. So can machine learning help us to screen particles and detect defects? So for this it was necessary to develop to build the data sets. So the data sets we focus on metallic nanoparticle perfect or with a screw or edge dislocation with different shapes. So we have to cut different planes so they have different shapes, but close to the wolf shape. And we played with the position of a dislocation. This has been done, sorry, with the Merlin software to create so atomistic configuration with defects. Then to be close to reality you need to relax the, I mean the structure of the configuration. So then using lamps that have been some atomistic relaxation and to get a data set, I mean we need some 3D diffraction pattern, some simulated 3D diffraction pattern. So this has been calculated using the PyNex software. So then it was possible to have a data sets with perfect crystal crystals with a screw dislocation edge dislocation. It was about one million atoms with a size of a diameter of about 13 nanometers. As I told you, it was very necessary to do some atomistic relaxation to be closer to experiment reality. And after this, once you have the data sets you need to develop the convolutional neural network. So as an input, you have the 3D current diffraction pattern and as the output, so there is series of layers, conditional layers. And at the end, there are three numbers which are probabilities of these three classes. So either no defect, either screw dislocation or edge dislocation. So to test the neural network, the neural network has been tested on simulated data. So about 100,000 diffraction intensity and this has to be split into a training data set, validation and test set. So for the training, so here are two numbers. So it has been obtained through Adam optimization with a learning rate of 0.001 and a batch size of 64. So the accuracy score on the test set was about 95% and in machine learning it's quite nice to look at the confusion matrix. It says how good works the network and how good it can predict for instance, perfect dislocation. And here we can see that it works quite well. I mean, it can very well predict perfect nano crystal and yes, so more trouble to detect a crystal with edge dislocation. Then which was very interesting and all this work, I mean, it's the work of the internship of Bruce Lee of A1Belec who is working at the ID1 beam line and Maxime. All this work, I mean, has been also compared with experimental data. Here I just show you some experimental data that we have taken at Soleil. So we have a diffraction pattern from a particle with our defect and diffraction pattern from a particle with screw dislocation. So here to be sure that we have a screw dislocation we have to do phase retrieval and to confirm that we have a screw dislocation. So if we look at the results, so here it's our diffraction pattern with no defects and the prediction is quite good. I mean, it would say that it's a perfect crystal. So here the probability is high for a perfect crystal and for the screw dislocation it's the same thing. It says quite well here that it's a screw dislocation just by looking at this three diffraction pattern. So here I saw a look at different data sets, experimental data set taken at Soleil and Daisy at P10 beam line and they all work and so we can set a defect recognition works on experimental data. Then I wanted to say a few words about the news from the EBS upgrade. So what's the news? I mean, news from the ID1 beam line and they're connected to the project. So I show you what we can do in situ a pronto and I mean the dream of everyone is to go to atomic scale and real time. So can we measure smaller nanoparticles with better resolution and do faster measurements? So we hope, I mean, we are not yet there but to be closer to the atomic resolution and to measure faster. And for this, so we had the EBS upgrade with an increase of 40 for the brilliance and coherence. So the idea is to be more and more surface interface sensitive and to have access to a slow motion movie to look at right limiting steps during operation and look at particle refaceting, defect formation, absorption, absorption diffusion. So I mean in the bright X-ray, so the ultimate goal is to have a ultimate bright X-ray microscope that works close to the atomic scale and real time. So and to work at high energy with currents with brilliance. So we have met some measurements on some platinum particles with a size of 20 nanometer in 3D and we have seen that it's, I mean, these particles are embedded into sapphire substrates. So they were stable under the beam and we can see that with the new EBS beam, it's possible to measure the 3D diffraction pattern of some 20 nanometer platinum particles. There have been some work with Maxime showing that now we can look at super structural reflections with weak signal. And here it's for a platinum nickel particle. So we can see here the signal from the 200 reflection and from its super structural reflection. And also we can measure at high energy. So there have been some demonstration with Steven that it's possible to do some measurement on 20 and 33 KV. I mean, it allows, I mean, we are using the higher X-ray penetration and we can look at embedded grains in polycrystal and we hope for less beam damage. Then another point, I mean, I show you a lot of results of phase retrieval and I want to tell you that when we are doing some in-situ or parallel experiments, we really need to do some online analysis. I mean, we need to know if we are going in the good direction, if we can see effects, if we have to change our parameters or not. So for this, we have David, I mean, all the PhD students and postdocs of the ESC project, they have worked how to improve, I mean, the reproducibility of the BRAC-CDI data treatment. And David has made a very nice Jupiter, I mean, graphical interface based on a Jupiter notebook. So it's called gray. So it allows to use at the same time and it's very easy to use. It allows to use BCDI package from Jerome Karnis for the pre-processing part. Then the phase retrieval, the PyNX package from Basson-Favon Nicolas and the post-processing part, BCDI from Jerome Karnis again. And to do it on the same Jupiter notebook, so it's quite easy to use it and it helps us a lot during our experiment to get some online analysis. So I would say it's very user-friendly. So if you want to test it, it's really welcome. And also one question was the reproducibility of the results. So one of the good idea was to set all the parameters. So all the parameters of the pre-processing part, phase retrieval and post-processing part in just one H5 file. So someone can repeat and use the same parameters to compare and check the data treatment. And everything is saved in one H5 file and it's then very easy to reopen with software, with a graphical interface. So if you want to know more about this graphical interface, based on Jupiter notebook, I mean, I hope there will be soon the paper from David or you can have a look at all the videos he made about how to use the software. And then to end this presentation, I want to thank all the team worked on the project. So all the PhD students, so I don't show you the work of Sarah and Nikita, but they're working on battery and their indentation, all the work of the postdocs and previous postdocs, I mean, nice collaboration with the ID1 beam line here. I mean, it's an old photograph, but I want to thank her a lot as Steven for his very nice help. I mean, during all the experiments, the collaboration with Letme, Tecdion and my collaborators from Marseille, from the SEA and the funding, so the ESC funding. So there are two open postdoc positions, so don't hesitate to apply. So one is on time-resolved BCDI, another one is on nano-imaging with deep neural networks and there is one open PhD position to continue the work on BCDI and electrochemistry. So yeah, don't hesitate to apply for these positions and thank you for your attention. Thank you very much, Marengre. This was a very, very nice overview of your recent work and now I invite everyone to raise their hands and unmute the microphone to ask questions and I hope to have a lively discussion. Don't be shy. I would also like to maybe ask David Simon if he's interested, you can copy the GitHub address in the chat so everybody who's interested can copy it. So Marengre, I do have a few questions actually. One is a bit of a curiosity. I recently participated to a conference on surface X-ray neutron and there is still lots of measurement on catalysts and using surfaces. Can you tell, can you comment about this complementarity or can you see that actually this operando studies on nanoparticles will actually take over and completely wash out all the ideal surface measurement? Yeah, so yeah, this is a good point. So, and this is, I saw a part of the PhD phases of David. So he's doing his PhD phases at Soleil at the success beam line. And the idea is to compare what we can do with surface diffraction and Bragg CDI. So he's done some beam times on Bragg CDI looking at the same platinum particles during a reaction. So the idea is really to compare what we can get with surface diffraction and Bragg current diffraction. And one of the goal will be really to match the two techniques because I mean with the Bragg current diffraction we can see these very nice tricks. So then the idea will be to go very far along the sitiers of the particle. And one of the goal to access, I mean to the atomic resolution will be to try to combine, I mean, these two techniques. Alex, please. Thanks, Dina. I had a question immediately and I couldn't find the hand at the end I made it. Thank you for this. I thought it was very nice to see this clear potential dependence in this train of this particle. I don't, I mean, maybe I've missed something but I don't think we've seen that before, right? In that clear way. Yeah, no, there was the central work. I mean, Jan Robinson has shown a very nice work with fuel and he has shown that the poor of BCDI to look at adsorption. There was a work of a previous work from Jan Robinson on this. Sure, sure. But like in the double layer, it's very impressive. I'm sorry, you mean, yes, in electrochemistry, yes. Yes, yes, yes, exactly. Yes, I mean. So you see it in the double layer region. Yes. Of course, if you say it's the double layer region, it means that you don't have, you don't expect to have sulfate adsorption there but there might be some at low level. But I'm like, what I wonder about is how do you, I mean, how do you interpret that? Because you might think about, you know, quite a few different processes which could give this strain effect, right? Yeah, so up to no. I mean, so there are a lot of papers, surface diffraction papers on electrochemistry and in the double layer region, they observe for platinum 111 and 100, clear demonstration of B-sulfate adsorption. So the very, yes, there is this type of adsorption because you were thinking of over phenomena to explain. Yeah, I mean, I don't know what they would be but I mean, you have, at the very least, you're changing the surface charging, right? So. Yes, so, yeah, yes, I agree. I saw because here we don't show it. I don't, I didn't show you here but just we have compared one particle out of the electrolyte and in the electrolyte and just without applying, I mean, the potential, we can see some changes of a strain, yes. Just inside the electrolyte, yeah. Yeah, okay, cool. So I just, I mean, I just wonder how you, like what your opinion is on this, like will this be a useful tool given that the interpretation is not straightforward? I mean, we can see where stuff happens, right? But what is it? Yes, so, up to now, yeah, we have to make some hypotheses, I mean, from previous literature. And, but we have also ideas of having some collaborations with groups making some like DFT calculations and to predict, I mean, this trying state, yes. So just my follow-up question then is like, what can you see anything in the hydrogen absorption region? Yes, so, so. Because there you know, I mean, there you really know what's going on and that's very well established from single crystal electrons. Yes, so here it was not the topic of this work but recently we have made an experiment and we look in the region of hydrogen absorption. So we are still looking at the data doing the data treatment, but yes, we have seen and this is, I mean, region where we can do a bright CDI. Yes. Yeah, because I'm for, you know, for, I mean, sulfate absorption happens on terraces, right? On one one terraces, but hydrogen absorption, those two peaks that you have there, that's one one zero and one zero steps. So I would imagine that if like, if absorption really gave a clear signature in the BCDI strain, then you would light up kind of edges instead on the hydrogen peaks there. But I guess that's still to do, I mean. Yes, yeah. Okay, thank you very much. I see a comment from Clemo who says that you do have some data in this region. I don't know, Clemo, if you wanna add anything. I just wanted to add that we got some data in the hydrogen region. And as Merringrid said, we just have to analyze them. So that's it. But yes, this is probably a more interesting region in terms of electro catalytic reactions. Cool, that sounds fantastic. Thank you. Thanks. Other questions, comments? Vonsok, please. Amid yourself. Hello. Thank you very much for very nice talk. I have one, maybe two questions regarding electrochemistry experiment. How did you adjust the thickness of the electrolyte layer over the sample? Because we have been, we have tried years ago and we have some problem to set the right thickness of the electrolyte. Some, if we set, if we set the thickness of the electrolyte layer to thin, then we couldn't see any behavior, electrochemistry behavior on the sample. Yes, yes, here the thickness is about 300 micrometers, is hundreds of micrometers. And we can, I mean, and we are working at 13 KV just to be sure to penetrate the electrolytes and to stabilize, I mean, we are, we are, I mean, the electrochemical cell is connected to peristatic pump and we try to keep this thickness, I mean, it's few hundreds of micrometers. And at 13 KV, it works quite well with, I mean, EBS or a BIM, yeah. I see. And the peristatic pump will push the bubbles away from the sample as well? Yeah, yeah, so we have seen that if we are not using it, we are forming a lot of radicals on the surface and we can, I mean, it's, I mean, on the CV curve, when you, we are not using the peristatic pump, we can see an increase of the current when the BIM is on. So it's very, I mean, extra sensitive. So we have really to be careful and to work in the good conditions. Okay, thank you. For electrochemistry, because yeah, it's very sensitive and we can see it with the CV curve, yeah. Yeah, yeah, I haven't been watching that. So we were chasing two crystals and one, we were excited to see the evolution, strain evolution inside the crystal. Actually, we observed the huge big changes in the diffraction pattern and then we moved to the another sample and we didn't see anything. So actually induced a lot of changes on the sample. Yeah, I totally agree. And there, we either work a little, yeah, with Climbo just to look by changing parameters of the electrochemical cell or the flow, how it can impact the measurements. And the there, yeah, it's so, I mean, there are, yeah, we have to be really careful about BIM damage, yes. Okay, thank you. Ross, I think you're next. Yeah, I was just curious how you control the, or where the electrodes are and how you control the sort of evenness of the electric field across the sample. Are these thin film cells good at that or do you have a sort of in how much you may need? So yeah, we are using here this electrochemical cell and Climbo is doing a, so I mean, it's a collaboration with Lepmi. So it's a laboratory focused on electrochemistry and is looking at the same sample using the cell at the BIM line and her conventional cell that is used by electrochemist. And we obtain really the same CV curve. So that's why we are quite confident about the process. Nice, so you haven't like looked at crystals near the edge of the sample and crystals in the middle of the sample and seen different behavior or anything like that. It's true that, yeah, for this study, we just, yeah, we measure several particles, I mean, and we focus on one particle, but yeah, it would be nice as you said to do some statistics to look more at the different particles around the substrate. Yes. Thanks. Tillman. Yes, it was a really nice talk. I really enjoyed it. I have one perhaps a bit technical question on the peristaltic pumping that you mentioned. I take it that you also keep on pumping during the experiment. So I was wondering, because I noticed quite a lot of pressure spikes from this kind of pumping. So I'm wondering if this is slightly moving your crystalline, if this has an impact on your data. So yeah, the peristaltic pump is on. We haven't seen some movement of the particles. So here the trick is that the particles that are de-weighted on the glassy carbon substrate, the electrode, and as they are de-weighted, they are strongly linked to the electrode. And we have made some measurements with chemically grown particles and they were not very stable. But here we have some, we are working with de-weighted particles that have a link with the substrate. And I mean, we haven't seen the particle moving during the experiment except in the oxygen desorption in this region. I mean, when we go to a potential higher than 0.6, so there is, we are in the oxidation region and there is some corrosion of the electrode. And at this point, the particles starts moving, but it's due to the corrosion of the electrode, of the supports. Okay, well then, good on you for fixing this problem. But yeah, yes, to fix this problem corrosion of the electrode, yeah, it's a little more difficult. I mean, the vis double layer region and the hydrogen region is easier to measure, yeah. Okay, thank you. Other questions, comments or curiosities? I have one about the machine learning. It's actually something that is, it's gaining more and more interest from this community. You show that you can recognize certain types of defects. How do you use this knowledge for the phase retrieval? Can you, have you found a way to use it? Could you, I don't know, use it as an educated guess for the starting conditions for retrieval or? Yeah, here it's, it was just classification. I don't know if we can use classification. I mean, this results in the fact that we have a defect or no in the phase retrieval. But I mean, here I just show you results on classification, but we are working on using neural network for phase retrieval. Yes, I saw. So we have a work on just 2D diffraction patterns and it seems to start working and we have to see in 3D how it works, yes. But yes, I think there have been some very nice demonstrations from the group of, I mean, of Janne Robinson and from APS that machine learning can help, I mean, for phase retrieval, yes. And I mean, we still need to make some tests that perhaps to see if it can replace phase retrieval. That's some hope, yeah. We have, but yeah, a lot of work to be done, yes. But at least for classification seems to work. But for phase retrieval, I mean, to get the phase and modulus with machine learning, I mean, we have to get a lot of data set or I mean, otherwise this demonstration that it's possible just by one data set that it can fit well. But yeah, we will work in this direction, yes. And I also would like to highlight the fact that you are basically making data analysis available. Do I understand well with this interface? So that's actually very, very good. Yeah, don't hesitate to try, yes. Yeah, I see that some people are leaving. I think that I can officially close the session. Thank you so much again for your participation. But then as usual, whoever wants to stay a little longer for saying hi or having a little bit more informal chat, we stopped the recording also now so we can say anything we like. And thank you again. And I, well, appointment to the next time. Thank you, Dina, yes. So Maringu, that was very curious about the, you know, all this data that you're doing in electrochemistry and in these conditions. I mean, how do you solve the particle stability? I've seen, you've shown these results on 20 nanometer. It's really cute that you managed to do so, but how do you keep the particles still? Yeah, here, I mean, for this 20 nanometer particle, they were embedded inside sapphire. So this is a big trick, I mean, they are embedded. But I mean, there were a lot of issues about beam stability. At the beginning, we wanted to work with the focus beam with the size of 15 nanometer, just to be sure to focus, I mean, to have a very high flux on the particle. And I mean, it was very difficult to make this 15 nanometer beam stable on our particle. Stephen is working very hard on interferometer for the ID1 beam line. So, I mean, maybe in a few months, we will have this interferometer and this will be maybe more easier. So we had to enlarge the beam to be sure to stay on the particle. And here the trick is that, yeah, the particles they were embedded, but I think, I mean, on the weighted particles, it should have to work. And they're, yeah. So I understand, well, they were not staying fixed for the 3D or they were not just not staying fixed for the 3D? I know, yeah, here, these particles, they were fixed for the 3D. But just we wanted, I mean, because they were embedded inside their sapphire, so it was fine. But you need, if you work with a beam of, I mean, just the size of your particle, it's difficult during the working curve to, I mean, to make that the beam is always looking at your particle. So you have just to enlarge the beam just to be sure that you are really well on the center of rotation. So it's more... And the nanometer is a phrenosome plate, right? Yes, yeah, yes. So it was more to be sure that the particle, I mean, the beam, everything was in the center of rotation. Yes. Okay. So any other common questions? I'm adding that you are, you have the right to stop us anytime. It's enough. I don't know. I think it becomes really, really difficult to keep the center of rotation on this 100 nanometer scale. I mean, this... On nanometer scale, I would say it's fine because we are doing a lot of experiments at ID1 and it's working quite well. I mean, for 50 nanometer, it's more difficult, but I would say 100 nanometer, it seems okay. Yes. How is it working? The energy scan, this is always a really huge question, isn't it? Yes, sir. And we all seem to have some issue with retrieving data with the... So I've not shown her... I mean, there is a Sahaa, she's doing a PhD phases on inundation. So she needs to do, I mean, energy scans. And I think so we have performed the phasor travel and it seems to work well. So we are using the BCDI package. And I think I would say, so during the experiment, it was working quite well. I mean, that should both largely these issues. How far are you scanning energy? Is it like a kilovolt or a kilovolt and a half? It's all right, but it's more... When you do the energy scans, what's the range? Is it like a kilovolt or more? I would say at least, yes, 100 there, half, 400, yeah, 100 EV or... 400 EV. Yeah, yes, 500 EV, yeah. Yeah. But yes, you really have to track the... I saw the peak, so to track in Dell. And yeah, and Steven has made a very nice crepe and I do one beamline and it's working quite well. So yeah, it has worked a lot, Steven, to make it work and to keep the... I mean, to keep the particle on the detector and to track as so the Dell movement, yeah. Actually, it's a pity that Alex left, but I think that they were trying because Nanomax has a detector arm, like a robot. Then they were also trying to kind of move away, to change the distance so that you didn't have to reshape and rebain all the pixels for different energy in the Q-range, but I don't know where they are. But it's true that we are... I mean, we are not reshaping the pixels and we compare the measurements done by normal measurements, conventional working curve and energy scan and it was quite okay, yes. Okay. Yeah, I kind of wonder what resolution you have to get to before it matters. We looked into the same thing, Andrew Alvastod actually did a measurement where as a function of energy, we moved the detector closer and further and it worked out to be just a few centimeters, over a half a kilovolt scan. Okay. And it just didn't really seem to matter, but I suspect when you're doing a really large resolution, meaning you're doing a long energy scan, you have to maybe pay a little more attention to it or resample in a simple way. Okay, you mean that you are moving... I mean, the detector is more or less closer because we are just moving... Okay, okay, we are not doing... Yeah, we would chase them in all three dimensions. We'd be chasing in Delta and Gamma and Distance. I think in this study, we only chased in one, we only moved the distance because we weren't scanning very far so the whole signal fit onto the detector. But yeah, we've done a couple of energy scans that were maybe a kilovolt and a half once, I think is that right, where we actually chased the diffraction pattern with two diffractometer motions. Those were quite painful measurements. So you have a longitudinal translation, motorized translation for the detector arm, okay. Yeah, we have about a two meter long stage and the detector can motor forward and backward. The measurement wasn't that crazy. Yeah, well, yeah, we wrote a little back macro to move the detector and then measure and then move the detector and measure and it would always have an overlapping frame. So it would measure, it would move the detector, it would re-measure and then it would start moving the energy again, so you could realign everything because the detector doesn't move perfectly and it all worked, but it was just a nightmare to assemble all the data and do anything with it. And then we don't have the photons to really make anything out of these high energy differences. Okay. Yeah, in the BCDI package, Jerome Karnes worked a lot on how to deal with energy scans and we are using it and yeah, seems to be fine on our side, yes. Yeah, we just use actual utilities. You just tell it your energy stuff and it gives you- Yes, but is that so using? I mean, is that so using as far as utilities? And yes, that's so write a code to make it differently, I saw, yeah.