 With this talk, I want to share some development that we've done on Black Tiger Coffee recently. So, some of the work actually was done during my post talk with Virginia Shamad. And some of the work are basically ongoing work at my current I13-1 diamond light source. So, firstly, I want to briefly review Black Tiger Coffee. Then I want to talk about the challenges we are facing and the development we've done trying to cope with those challenges. In the end, I want to share some Black CDR and Tiger Coffee work. I'm currently working on an I13-1. So, I assume the majority of the audience here knows quite well the CDR or Black CDR this technique. So, I won't spend too much time here. The general idea here is you have extended bin. Then your crystal is smaller compared to your bin. You place your detector at one of the black peaks from this crystal. Then you rotate your crystal to collect a stack of 2D cuts of the 3D research space. So, this gives you the intensity of the distribution around that particular black peak. And you need to use phase retrieval algorithm to get back to the real space, which would give you a complex electron density quantity. And this with the modulus part encodes the electron density of the crystal, which gives you the morphology information and the phase information encodes the lattice displacement field from where you can extract the strain formation. So, Black Tiger Coffee, you can simply think of it as a scanning version of Black CDR. So, in addition to the rotation scan, you have to introduce a positional scan as well. And you have to make sure your illumination at the two neighboring positions have to be partially overlapped. And this introduce some redundancy in your data, which would help to give you much more robust reconstructions. And because of the scanning, so you can look at much larger crystal instead of small crystals, you can look at with Black CDR. So, the reconstruction, the phase retrieval part would be very similar to the transmission in Tiger Coffee. So, you have to look through different positions and get back and forth between real and detected plane to apply the constraint there. But the only difference here is that with Black Tiger Coffee, the quantities are 3D and in transmission Tiger Coffee, they are 2D. So, here's a rough timeline that how Black Tiger Coffee progressed. So, its first x-ray experiment was done in 2011. It's basically a proof of conceptually demonstrated its working and actually later on people are trying to use its 2D simplification to study some sim films. In this case, you don't need to do the rotation scan. You only need to sit at the black angle and to do the positional scan. So, the experiment is much more simpler and also the reconstruction you can basically directly adapt from transmission Tiger Coffee gives you very good reconstruction as well. But of course, you are limited to 2D sim films, so you can't look at 3D samples. So, later on actually people find out with 3D Black Tiger Coffee, it's actually better to not reconstruct the probe. So, you normally pre-capitalize your probe and fix it and that helps to give you better optimal reconstruction. Then later on you basically add more constraint to your reconstructions. You know your sample has certain thickness, so you apply this constraint. It's a bit similar to the small constraint in Black CDR. So, basically as the reconstruction techniques improves and also the bin line techniques improves. So, slowly you can acquire much more successful data set and Black Tiger Coffee slowly becomes a practical tool. You can really use to tackle real issues. So, more applications start to appear, but despite all these success, Black Tiger Coffee still remains very challenging. It needs a lot of expertise, efforts and sometimes luck is also very important to get a decent data set. So, one of the challenges I found very difficult is the inclined 3D geometry. So, all the quantities here involved are 3D and they have their own natural orthogonal frames like the probe object and detector. They have their own natural 3D frames. You need to convert them between different 3D frames. And also the most important thing is to assemble the data in a correct way. What I mean by that is the detector frame you get from the experiment. You might need to flip left and right or up and down or even transpose. You need to get this correct in your defined frames in order to proceed with the reconstruction. And also the 2D translation, the translation has to be decomposed in this reconstruction frame correctly. And the sign of the motors has to be aligned with the frame you define correctly as well. So, even this work is quite complicated and that still makes a lot of mistakes when doing this. So, another challenge is we cannot really use 2D simplification for 3D case. In transmission typography, you actually can combine with tomography. In that case, you actually can deal with each angle independently as a 2D case. But with black typography, you cannot deal with it in that way. So, angles has to be reconstructed simultaneously. And in this case, your stability has to be maintained through the whole experiment. Not as in the transmission typo tomography, stability only need to be maintained on projection level. So, it makes the experiment of the requirement much higher. And the 3D problem reconstruction is not converging. That's also limiting your objective reconstruction quality. So, another challenge factor is the limited photons that you can get in black geometry compared to transmission geometry. Often it's one order or two orders of magnitude less. So, this means when we do black typography, we have to work at focal plane to gain the flux density. But the problem is at focal plane, your probe are very small and with typography, you need to require overlapping. So, your step size has to be smaller as well. So, your stability requirement has to be better than your step size. So, your stability requirement is much higher as well in this case. And because the photon is less, you need to expose longer to get decent signal to noise ratio. So, and it involves rotation and translation. So, the whole data acquisition really takes very long. We typically design experiment to be around 10 hours longer than this would become very impractical for an experiment. And during these 10 hours and you need to make sure your your being is very stable. So, I've already decided challenges now on to share the developments we've done to cope with these issues. I want to demonstrate those developments with these two works. They actually published quite recently and also uploaded the code and data for both of these works. So, if you feel interested, you can try to get a copy and play with it. So, the first work is actually a collaboration with with Felix Hoffman from University of Oxford. So, he's trying to use helium ion implantation to tungsten to mimic the unusual irradiation effect in the fusion reactor at the production layer and see how it affects the mechanical property of tungsten. So, he basically prepared this sample and we did this experiment I do one and which we choose an area that covers a bit of the implantation area and no implantation area and close to a grain boundary. We did this. A practical scan on this area. The first and the first problem we faced is basically how do we do the probe reconstruction. So the bypass is to first to characterize the probe using the transmission geometry. And you fix this probe in the black toggle for reconstruction. And that's our characterized probe and that's the black toggle for object reconstruction. So, the grain boundary has a good indication that this reconstruction is basically successful. But, you can see all these features that it's not quite true. So we think we should consider to reconstruct the probe to improve this. And in the x-rays, because the focusing optics have much more numerical aperture, this means you the depth of focus is much longer. In this particular case we probe get the probe through a few hundreds of microns and you don't really see big changes. And if you compare with the lens that being intersect with the object, it's much longer. So it's safe to say your probe doesn't really change along the propagation direction in this case. So we can basically use this constraint to construct our reconstruction. And indeed it helps with the convergency. We managed to do the probe reconstruction in this case. And you can see the object reconstruction is, it becomes smoother, but it still has a lot of these high frequency features. So if you compare this probe reconstruction with characterised probe, you can see actually you have a lot of details outside the reconstruct window. And we think this affects reconstruct quality. And this window size is actually limited by the angular stop size we used in the experiments. Actually, to get a smaller angular stop size, we borrowed this idea that in transmitting technical, you often have some pixel that's pixel gap on your detector. You can use our word to retrieve those gaps. So it's basically a similar idea. We can virtually put some angular point in between the main ones and use our word in trying to recover those points. And this space effectively give you a smaller angular step size and in return give you a larger reconstruct window size. And indeed probe, you can reconstruct much larger area of the probe. And in return that gives you a larger field of view on the object as well. And then you start to see some small features that you couldn't see before, basically some dislocations. And of course the reconstruction quality is much better. And this part of the illumination we couldn't see is because the grain boundary. So this part of information is not really measured on your black pick. So yeah with this reconstruction with the phase you can extract the lattice property here is a strain. And we can see the depth from the top is about 2.6 microns which matches quite well with the expectation and also the strain value. Even how we prepare the sample. What's nicely here is you can see this little strain enhancement as a top and bottom surface of this slab. And this is actually caused by the by the safe damage during the sample preparation. And what's nice here is then you can when you analyze the strain caused by the healing you can exclude those two layers caused by the fifth. Which this is not it's not possible with other techniques like like my actual microwave or electron, you bestie this sort of techniques. And you can also extract the lattice rotation, which is quite smooth indicates the distortion caused by healing is quite the direction is quite random. So it doesn't have a preferential direction here, and you can also zoom in the to those two dislocations and see see some fallage and the face distribution around it. So the second work is what's done at nanomax so thanks to dinner we got this commissioning been time and nanomax. The good thing that is the current flux is very good. It's about 10 to the 10 12 kV. And we designed this crystalline silicon star structure as a test pattern to to to test the performance of practical thing and also the the been more. So with this good coherence we will manage to to use a very short exposure time in this case 33 microns. 33 milliseconds. And that gives us decent signal to noise ratio. And with this short exposure you can basically use fly scan with it. And that's what we did. So, in the end, a complete 3d data set and the total exposure time is about 27 minutes. And actually the experiment itself takes some things like less than two hours, because some overheads and as we were told is the scanning stage at that time wasn't quite optimized, which causing quite a lot of overhead and also cause a lot of the issues for us here. So this is basically the, you can see the star on the different angles it, it works around. So, so this, this causing basically a very bad quality. We'll take the data as what it is. And that's the sort of quality you can you can get. And of course, one of the things we can try here is because we scan slightly larger area compared to this star structure. We can use cross correlation to to roughly align those patterns and reshuffle the data basically. You can see the star states much better than before. And with this already, you can, you can have much better reconstruct quality. But the thing is, and this course correlation only gives you a rough alignment so you still have a lot of residual position arrows. And you can see those effect. And the problem is with the, with the conventional reconstructing technology where we call 3D for transform method 3D FT, which I will introduce a little bit more later. With this method, you cannot basically improve further. So in, in 2017, Stefan who switch from APS together with Virginia in Charlotte. They developed this new reconstruct technique called 3D black project and target. And basically allows to enable the possibility to, to, to improve the this type of data that we can run quality further. The reason the conventional expansion 3D FT couldn't improve further is when we take the data is we look through the angles and on each angle we repeat the type of the scan, the position of scan. But then, then you need to, to assemble all the angular slices for a particular position into a 3D 3D diffraction pattern. And when you do this, you have to, it basically requires this position under all these angles has to be exactly the same. But practically, this will never be true. And that's why 3D for transform will never be able to improve further here. But with the 3D BPP, what it does actually is using the forward for a slide serum to decompose this 3D diffraction. Well, the 3D for the transform into a series of 2D for the transform. Basically, you take, you take only one particular angle, you take one of these slides, then you look through all the positions to do the, to do the reconstruction, then you, then you repeat all this for all the angles. And the benefit of it here is the positions on the different angles can be completely different. And this basically allows the possibility of doing position correction because position correction inevitably create difference between, between different angles. And it actually also allows to do the angular correction if this is necessary. So that's what we did, we coded the algorithm and we tried it again. And that's where we can draw quality we can achieve. It's very obviously much, much better. And then you can basically in the reconstruction trying to recover the, the global drifting angles and also the individual position errors on each angle. So if you plot out the electron density, you actually can see these little features that you can see from this SEM figure. Which is, which is quite surprising. We kind of amazed by, by, when we looked at these small features, we didn't expect a rectangle can, can see, see these small features. So we get a decent resolution and field of view on this data set. And with the face, face result, you can extract the lattice property as well. Since we use fly scan so it, it effectively blows your, your diffraction pattern and that's equivalent to partial current effect. So we here also use the mode reconstruction for this data set, you can see the mode distribution matches quite well with them with the forward forward a transmission geometry. The mode structure. We also actually tried the energy scan here. So any scan basically gives you alternative alternative way to, to, to assess your 3D reciprocal space. So with the rotation you basically translate your detector perpendicular to the, to the scattering, scattering vector. And with the energy scan you translate your detector plane along the scattering vector. But it also gives you a stack of the 3D diffraction pattern. But the benefit here is that you remove the rotation on the sample. It helps to improve the stability of course. If your sample need to be in a complex environment, this would be very beneficial. We, we, well, that's basically the result we got. It's actually, you can see it has a bit of a distortion which, which we couldn't manage to get rid of. I think it's, it's most likely the, the scanning stage doing some weird thing now we don't, we don't know. But this, this result shows the promising of this energy scan. Hopefully we can repeat this experiment at another time. So in, in the last section, I will, I will talk about some ongoing work that I'm working on, I searching that one. So the interesting is actually is, is a two branch line to independent branch line. So one is called imaging branch and another is called coherence. So you need a branch to basically do directly imaging technique like in a line phase contrast or transmission equipment cost. The coherence branch is basically mainly do the current deflection imaging techniques, but mostly type of related techniques. So here are some general specifications about the coherent branch. The mono we use a QCM so it's basically for silicon 111 crystals, but mostly we use just two of them to which is enough for the for the coherence for the coherent requirement. We also have another set of silicon 3 311 crystals. Basically we have a lot of space to, to tune the, the, the energy bandwidth. So the typical energy range is 7 to 24 KV source to sample distances to 220 meters. And that gives a very large spatial coherence lens as a saw plane is about 400 microns. So we for the focusing optics will mainly use a financial plate, which give us 200 microns 200 nanometers for size. We also have a KB bendable KB gives like five microns for size. I never used it and I was told stability was quite the issue so it hasn't been used for quite long. But I think this both sides quite good for black city I so I would like to give it a try, but I don't know how difficult to to bring it back though. We also have a different set of a CL as upstream of the of the beam line, which we can use to either call me the being or auto form a secondary source, which also give us some flexibility to to tune the focus size combined with nose on play, which we want to want to try to see whether we can get a larger slightly larger focus force to do the brach CDR. So the techniques here we mainly do it's basically both transmission and brach geometry and transmission is mainly a type of fee and combined with different techniques and mainly here is a tomography, so it gives you high resolution 3d volume image of your sample. And we also have to rub down house and back in part but another detector to do brach geometry. Basically we can do both brach CDR and brach type of fee. So the brach geometry here, the detector can can move to in this quadrant space horizontally up to 30 degrees and vertically 29 degrees. And the robot on hold the vacuum pipe is about it's about two 2.6 meters and and the detector we typically use here it's a diamond in house detector called Excalibur RX 3M is basically use a chip 3 by 8. So, yeah, those are the specifications. So the sample stage has three rotation stage. So, the rotation along the, along the vertical axis is can rotate more than 180 degrees because we need to use it as for the tomography case as well. And now the two rotation stage and goes to from plus minus 15 degrees. So, I also take this second star sample and did one did brach type of fee. Oh, I've been lying. Of course, our flux is much, much less compared to non max. So we have to use a much longer exposure time here, but our focus bodies is larger so we can use much less positions to cover the same area here. So in the end to the, the whole experiment takes about four hours and this includes the overhead this needs basically the whole experiment. It takes and the I think this lens is basically it's not too bad for a third generation source giving especially giving the, the, the reconstruction quality you get on the object. And also here actually deliberately opened up the front lens leads to to get a bit more flux but sacrifice a bit on the coherence and you can see the most structures. And mostly horizontally that's basically a sleaze I opened and then back to coffee did quite a good job to get this most structures out. And I also tried and I'm a crystals with both black CDR and black type of fee. So here it's basically want to the, the ultimate goal for me is to want to show that back to go we can can reconstruct, reconstruct much highly strain crystals, but it's also very interesting to see the black CDR here because both sides it's, it's quite small, it's smaller than the particle so we have to defocus, I've been to to do Brexit you know when you when you defocus you have this face curvature effect. So it's also interesting to see this defocus effect on black CDR. And yeah, basically, black, black type of fee has this benefit that you can deconvolve out the profile in this case. So we won't have this, this face curvature effect on your crystal. And, and another could be beneficial is, is you don't have this string image and I'm great as you, you have in black CDR. So here are some basic the primary to use for the for for the data collection, and three different crystals actually done more but I only show three here. So the, the first one is you, you use to cut a small hole in the, in the, in the crystal and the second one is intact. So we've done nothing to it. And third one is, it's a non-intended, indented crystal. And, and then we can see a little bit more detailed comparison. So the first one, you would take the, the cut through the three orthogonal central planes, you can see the face actually broadly matched with each other, but you definitely see differences. And if you align them and take a face difference, you can see that's the difference you get. And then, as a beginning, I would thought, okay, the face difference would be just the face curve to get from your pro. And here to, to be noted, the, to be noted is, is a scale here. So the crystal is only like one micron. So if you compare with the probe here is only the pretty much the, the central blue area. It's actually quite broadly flat, I would say, has a small curve to them. In ideal case, I would thought the face difference is just, is, is just the face curve to go from your pro, but, but it's not the case here. And the reason I think later on it's them, it's them. When we do the Bruxy guy moment we typically repeat a few working curves to increase dynamic dynamic range. But the thing is when you each time of this working for a moment to your crystal might be slightly moved respect to the bean. So this can cause an average effect. So we don't necessarily see the face difference is just the, the, the curve to from your pro, but it's still, it's still confusing to me that this, why the face difference is like this. So in this second case it's, it's an intact crystal. And actually, to me, this comparison, black tackle seems more reasonable because, because the face is quite flat. That means your, your crystal doesn't really have a lot of strain inside it. But somehow the Bruxy guy gives them, gives the basic issue of some strain in the crystal. If you do a face difference between these two, and you compare with the profile is actually actually matches a slightly if you if you, if you see this slightly. Well, if you, I don't, I don't know how to describe the colors here in English, sorry, but I mean, it's basically you compare with this. It matches basically this, these slides and this slide particularly. And this slide is through this, it's, it's not quite there. It's, it's basically, I personally think it's quite interesting comparison here. And the last one with another identity crystal, it becomes very interesting and and confusing to me. Because look at the look at the face. So, Bruxy guy actually shows them. I would expect with another identity crystal have this sort of face profile, you know, but black tackle for somehow shows them quite confusing face reconstruction here. But actually if you look at the modulus part. Bruxy do I show this oscillations in the module which I don't quite think that's, that's real and and it also why this part of the crystal is missing. So, yeah, it's still very confusing and the results quite new to me I need to spend a bit more time to, to, to dig into it. But I want to show here and so the audience can give some opinion suggestions to to create a good discussion about them. So to summarize. We, we basically implemented the proper construction with this propagation variance constraint, and that helps to improve the object record on quality. And with the long data collection issue. We can use angular up some plane to to mitigate that and also with four generations control source, we can afford to use default spin, which means your probe can be larger to cover same area you, you, you need use much, much less positions. You can also with four generations in terms of you can use slide scans. And with the stability issue. With this new because I'm in technique or 3dbp you basically can incorporate position correction algorithms, which helps to relax the stability. And with the, with the Brexit and type of free work I'm currently working on I 31. I hope I can bring to you that back to cook for a while on a certain dash one. But, but the comparison between these two on the nano crystals. So the default spin, how much it affects the result and and how it affects the result. I think here, what I need is, is to have one of these crystals to use a larger focus spin to to to do the plan plan with Brexit on to to to have another reference here. So, I think it's good for her for highly deformed crystals, which, which will be struggling for brexit I, and that's what I hope I can demonstrate, probably the examples are showed here it's not very convincing. Maybe I need to probably find a better crystal or do more do more experimental on on these crystals. And with this year. I would like to thank all the other collaborators and colleagues are involved and helped with with the work I show here. And also, thank you for your attention. Thank you so much for this overview. I would say it's almost a tutorial actually in practice. Yeah, it's really, really clear and interesting. And now the floor is open for comments questions. Please don't be shy. You can open your cameras unmute yourself or if you like, you can just write the question in the chat and I will read it. Thank you so much. I think you should unmute yourself. I'm afraid. Yes, I'm muting even helps get making my point. There was a really nice talk, thank you for this. I was wondering do you I mean I was there the break CD I was this break type of comparison was super interesting to me. I also have ideas how to get with like a third method. The ground truth or ideas for deciding a sample, which has such a controlled strain distribution that you know what you're getting. Yeah, that's what I mean towards the conclusion I said. To the summer I said, we need to do a plan with Brexit on it, which we need a lot of folks being to create this plan with condition. I hope that gives you it gives the ground truth, because ground truth comparison basically. But how do you know that is ground truth. I mean, apart from this, and only simulation I guess can can give you the answer to it. Experimentally, I guess you always have the doubt which one is really the ground truth. You're second in line. And you should also mute yourself. Same. I've lost the mouse. Okay. Hello everybody. Hi, I'm super happy to see this resource. I remember during your last talk we are giving a little teaser about promising experiments and it's really cool to see that you have managed to progress quite a lot. I was wondering about the curvature. So I didn't really get where is the curvature in which of the sample so the difference between the two presents a curvature, right, a face curvature. So it could, it could either be that the curvature is in the Bragg's di that I said, or in the typography that I said, right. But, but the thing is with Bragg toggle for you. No, but it could be right. So the bragg toggle decompose the probe out right because I don't know if you remember when Arthur was doing inversion, but he was always retrieving a sample with a curvature and the opposite curvature in the probe. You don't, don't you remember this. Yes, but it won't be the case here. Well, in this case, actually my crystal is, it's a small, small quantity and my being is the extended quantity. So I basically use my crystal as a probe to scan my being in this case. But the thing is, you cannot have an opposite face curvature on the small quantity and have another opposite on your extended quantity. They won't cancel out. I don't understand what I mean. I think I understand what you mean but it's counterintuitive with what I remember from Arthur reconstruction and experiments. Yeah, I don't see why the comparison is not valid. One of the possibility I explained is because when we do Bragg CDR will repeat a few working curves and something together to give you more signal. No, I understand you found an explanation to explain curvature in the Bragg CDR that I said but I was wondering whether the curvature could not be in the Bragg dichography that I said instead. And then I don't understand why you are ruling this out 100%. But okay. Okay, it's maybe too technical for the audience. Okay, thank you. Oh yeah, was the next online. Thanks for the nice talk. I have a question related to what Tillman and Virginia said, seeing this comparison between Bragg CDR and dichography doesn't that compromise this as technique to determine the strain for one or for the other because you have a huge differences there. I guess it's, it's determined which in which conditions you get the same results. It's difficult to know which one is correct regarding the determination of strain. Can you comment on that please. Well, I mean, I, I always have the struggles to, when I see a paper, I always have the struggles to see when I see the result. The question is, can we trust this result. I always have this doubt. And that's kind of the reason I want to, to compare these two. But of course, Bragg dichography in this case is not guaranteed that gives you the correct result. So it's a bit difficult to draw conclusion here from those results. But the thing is, if doing simulations, and on the other hand, it's always give you the correct reconstructions and it's quite different from the real experiment because you have all sorts of issues. I don't have like a conclusive how do I say, conclusions here but but I hope I'm trying to find out the answer here. Yes. Thanks. Thanks, Matt, I was next. Yes, it's about the source you use a continuous source, and if I understand well, I mean a constant intensity. Is it possible to use a pulse at the x-ray for this technique, or if you have experience about this. Not too much experience, but as long as the being, each time you measure the diffraction pattern of the pulse, as long as there being three different pulses, they are consistent. It's basically similar to the continuous case, I guess. But I guess in reality, the pulses are quite different from each other. So this would cause issue because black tackle relying on your probe being constant through different positions. There are there are algorithm trying to deal with it in transmission type of fee. I guess that can be tried. But to what level the the the final quality would be like I cannot guarantee basically. I hope that answers your question. Okay. Nick. Hi, I'm really cool to see the updated results from the comparison of the Bragg CDI and Bragg Tyco work at 13. I was wondering if you'd compare the crystal that you've got on the screen now. Have you compared this with the APS data from the same crystal. Like, I guess, with the view of looking at a different probe. Yes, while I'm preparing this talk, I realized I should should get that result and trying to compare, but I haven't done that. But yeah, that's definitely. I mean, it is the same crystal, right? I mean, it looks very, very familiar to me. Yes, it is the same. And I realize, yes, at APS, you basically is a plain way with Bragg CDI. Yeah, yeah, you should be able to get that data, no worries. Yes, yes, I'll definitely do that. Cool. I had a second question. When did you look at the, if you don't average the, like, you did eight scans or something, and then some of them, if you just, is there enough, is a reconstruction high enough quality if you only reconstruct a single scan or not? I haven't tried, but I would definitely try to get rid of this weather. But then another question occurred to me is, even during a rocking curve, your, your crystal can move slightly with respect to the being. Certainly, yeah. Yeah, so you didn't really answer the question here, but I definitely tried that. Yes. Cool. Thanks, Bang. Yeah, thank you. Thank you for giving the word to other people who would like to ask. I have a question and a comment about this is obviously very intriguing. This comparison with results, so you were expecting to raise some noise here. I think that in this case, simulation would help a lot. Obviously, you're using the particle in a de-focus beam and obviously you don't know which part of the beam the particle is, is hitting. So this is, this is one, I think this is, in my opinion, intuitively this is the main factor here. The fact that over the three, over the 3D scan, you are not sure that your particle is actually illuminated by the same part of the beam. And then another factor here, and, but I can't say how this is affecting, is that the crystal is actually a filter. So while with the tachograph, you're actually using a parallel beam with the CDI, you're using a divergent beam. So would you need a larger rocking curve to pick up all the component of the phase? You see what I mean? Am I being too... I'm not sure I understand the second question here. If you need to have a larger rocking curve with the CDI just to try to pick up more of the divergence of the beam, or the divergence of the beam is smaller than the particle. No, it is larger. Maybe, maybe this is a bit too technical for this, but... Sorry? Sorry, we can discuss this offline. I'm not sure I completely understand your question there. Are there any more questions or comments on this talk? I have one quite general. Oh, Virginia, I think you're... Yeah, maybe I finished the question. What about energy scans for the 3D? This is an open question. I have not seen many results, even with the Brock CDI. Is there a fundamental issue, you think, for the retrieval of energy scans? Well, I'm not surprised to see more results from energy scans, Brock CDI. I don't know what caused that, but certainly the reconstruction is slightly more complicated. I don't know whether there is an existing package that can do that and open-sourced. I guess that may be one of the limiting factors. So I remember a couple of years ago, Alexander Bierling from Nanomarks and Jasper Valentin, I think, from the university, they were trying to create an acquisition mode in which they moved, for every energy they moved slightly, the detector away, or closer to the beam, to adjust for the pixel size. But I don't know if they ever converge to any... But I think, in principle, you can do this computationally. You don't have to move detector to compensate for that. Yeah. Okay. Yes, if I may comment to this one, I think we have a paper with Stefan Ruskevich, where we introduce or where we use the back projection in order to compensate for the change of the pixel size during the scanning in energy. I don't know why it is not used that much, but maybe it's because with the energy scan, you are moving only along the QZ direction. And so you are not maybe probing the reciprocal space in a very efficient way, could be the reason. I don't know, I'm not very familiar with that. So I wanted to ask a question. Actually, it was your question, Dina, about simulation. I think Dina was right when she mentioned that you can solve a lot of things by making different simulations in order to understand the origin of this curvature, in one of the other data set. Did you try already to introduce, I don't know if you are, if your sample is slightly moving during your working curve, or if you have a limited amount of photons, or if you do not know exactly where you are with respect to probe these kind of things. Are you planning to do? I'm planning, yes, I'm planning, because looking at this result, seems like I cannot avoid simulation basically, because though the result, we cannot draw a conclusion from them. So yeah, I have to do it. Yes, I'm planning to do it. Okay, thanks very much. Thanks. Oh yeah, next. Yeah, related to the energy scans, one of the issues you can have, even if the energy scan is not very large, that the beam will move slightly depending on the offset of your mono. That means that if you move by one micron, two microns and you are illuminating with another part of your beam, that can cause a problem. And that might be one of the reasons this is not used so often. Yeah, in that case you need a larger focus beam to basically avoid this kind of issue. Even so, you will be illuminating a slightly different part of your monochromator and then changing things slightly. So and it will shift vertically in a typical monochromator if you don't change the offset, it will change by one micron. So yeah, maybe with a very wide beam, you can get away with it. I see. I would like to bring to your attention that Marie Ingrid who is in the audience. Hi Marie Ingrid. I would like to just put on the comment chat a link to a paper to which she just called her, I believe, about the comparison so this is something that is there for people interested. Yes, because that's her. Yeah, sorry I'm doing. I'm at the same time, but yes, in this paper, so at ID one, we are doing energy scans while doing BCDI, so this is a work. We have done, yes. Okay. Thank you for that. So is 1633 I officially close the talk, I would like everybody. So I would like to thank thank but we are all obviously everyone is welcome to stay longer and to contribute. What. What. Sorry, sorry to contribute to the conversation into this to the discussion. I thought I had closed the chat. What are you thinking of some special no no don't say that I have a question for you. Were you thinking of some special system for your post experiment. Yes, we had the fun, not me but one of my colleague has a financial support for a free electron laser. And he is, yes, he's planning to build some device for pulse at the X ray beam. Okay, and here in Torbergata, he asked me some experiment related to two dimensional materials and what we can do with two dimensional material connected to with the pulse at the X ray beam. So the question was just about to understand if pictography was possible with that source. Yeah, is it crystal samples or, or you do it in my case they are films very very thin films to dimensional materials. Yeah, but you want to do brag or transmission with them. In general, he's acquiring the source and then he can use with the every kind of material with can be crystal single crystals thin films what he wants that we can just exploring which kind of materials are also biological different kinds which are suitable for that kind of analysis, which kind of analysis we can do with that facility. So this is a laboratory source this means that it's a low energy. You know there are a large amount of money because of the financial support for European community after after COVID. So every group has a plan led to increase the the facility of the laboratory and one of the group here had that opportunity. Yeah, I think I can maybe send some. Yes, yes, yes, probably your suggestion will be very useful for me. Yes. So Pang, what I was saying before it's, it's seen that the beam is curved, you have you basically with your crystal you need to probe the whole beam to reconstruct it in city I don't you need a larger angular scan for that. With respect to the typography, am I being too, too naïve or too simplistic, you know, with the with the typography you are working in the focus beam. So you're, you're, you have a plain way. There's no divergent stuff. There's no curvature. You can relate your curvature into the angular spread of your beam, which is my opinion what it is. You're basically having a number of incidents of incident beams in your, in your right. So you mean, you need a larger rocking scan to probe all. I may be saying a very stupid thing, you know, because it's. I don't think that's true. Because an often when you do a rocking curve you tend to extend a little bit more. So, you really don't see too much photons towards the end of the rocking curves. I think the fact of this convergency should be quite little, I would think. Because anyone in the audience disagree with that. Well, that's my, my thought, I'm not sure that's correct though. But the data that Nick was referring to taking an APS are they taking with a large beam with a parallel beam or with a focus beam? It's a larger focused beam with KB. It's basically a typical set up ID 34C. I think I think our comparison with those will tell you a lot because I have some questions again. So how sure you are that the sample is the same between the APS experiments and even between the two by 13 experiments so much sure you are that there is no did you try to do. Back CDI take over and back to break CDI, for instance, to see the evolution of the sample just trying to understand where it could come from. So you mean the radiation damage. Or maybe the also the reliability of the, the, the, the reconstructions in terms of data acquisition, noise, everything. I mean, you know, if you have two datasets acquired in very similar conditions, but they express the same kind of difference that you have, it's just the, the intrinsic limits of the experiment. I don't know, I'm just thinking. When we talk about radiation damage, I guess the fact it's quite little because it's gold, it stays in the beam quite well, I would think, and it's a third generation source it's not, it's not very strong flux on the crystal. So I would think. But you didn't. No, I didn't. Can we discuss again this parabolic way front, the one I was referring to is a without you. I didn't understand your answer. Yes, and the thing is that if you want to curve on your pro and off the curve on your on your object to cancel it out right. The thing is you need to move your probe on extended object, like to different positions, which was the case of actual rights, which one was doing. Yeah, but you cannot have those opposite curves on different positions on your sample to get out right. I see, I see, I guess I see the point. Yeah, it's not consistent. So what you say is that you have only, no, I mean, I just listen but I don't understand. So you have only a curvature when the probe is smaller than the object. But, but, but I don't understand why. Why, why if the probe. No, no, no, I think, no, I think in either case you don't have this ambiguity in your type of reconstruction. But Arthur had. Yes, I think I think now as if I think about I think what I remember the isn't it's the. And I think it's so in a paper of John Rodin look I think he mentioned it we found a reference to that problem. Yeah, I think the problem is not what you think it's I think it's a new field and far field difference. Yeah. So new field do you have extra phase curves you need. Yes, yes, yes, you're right. You're right. It was it was the it was the near field far field limit. Yes, yes, right. And you don't have anything like this. So here you are. Well, I think. Well, I would have it. I would detect distance is like three meters. I would. I think that's quite safe to. Yeah, but maybe it's it's worth calculating or large was the party cut. One micro one micro. It's not too large. Okay, and what happens if you add an artificial phase curvature to your probe during the reconstruction can you check if you get this kind of curve a child. You try to offset your reconstruction to check if you would be theoretically able to get this kind of artifact out of your beam. What do you mean you mean with the real data you with the real with the real data, you artificially add a face curvature on the probe. Yes, and see if this and if the algorithm is stabilizing without this curvature of the curvature is getting transferred to your object. It's just a simulator. Right. Yeah, I guess you learn to live to the simulation. Yeah. Another question. Sorry. May I just want you to say goodbye to pain. It's done. Bye bye. And by the rest of course. Okay, thank you till my that's what I was saying only to bang that's no nice. Goodbye everybody. Thank you. Thank you. How did you estimate the beam profile at the sample position in the break CDI. Could it be wrong here could could you have this little additional curvature. In this case, in the in the extra milestone is basically the difference between Brack City and Brack type. They are sitting at the same position. That's why I have a default spin and small crystal to scan the bin. Okay, okay. Sitting on the same plane. Yes. So I don't know. I have no idea. But it's really cool. Yeah, that's a super. So, thank I hope you're happy to you sparked quite a conversation here. Yeah, I'm happy. I hope you guys. Learn something or take away some important information from this talk. Yeah, that would be great. Yes. Hey, we're looking forward to the next one. Okay, I have to leave you I have some data to look at. Fantastic. Great. Okay, thanks. Thank you everyone. And yeah, see you in a couple of weeks again in this format. Thank you. Thanks everyone. Bye bye.