 So welcome to the 4D Imaging Lab at Lund University. My name is Stephen Hall and I run the facility here. So first of all we have our Zeiss X-Radar 520 Versa XRM, which is a 3D X-ray microscope for high-resolution 3D imaging of objects from a millimeter scale up to tens of millimeters. Making it at resolutions down to about 700 nanometers in three dimensions. And we can image objects in this machine up to about 55 millimeters and down to capture the full field of view. Inside this machine we have an X-ray generator which goes up to 160 kilovolts. And then we have a CCD camera sitting here on the right with microscope optics allowing us to zoom in on our samples. The samples sit on the stage there, you can see in the middle between the X-ray source and the detector. And then we also have the possibility of doing in-situ tests facilitated by the cable trade you see at the bottom. Then we have our new machine from RX Solutions, an EZ-Tom 150, which has capacity for doing larger sample sizes. And so here we have an X-ray generator of 150 kilovolts and a large flat panel detector. So we can image objects in here of up to about 30 centimeters in diameter and 45 centimeters high. As the name suggests, the lab for the imaging is the major part of what we do. So that may be run in-situ testing under different environmental conditions. These are three of our test cells. For example, on the right you have a freezer allowing you to get down to minus 20 degrees Celsius. In the middle is a pressure-based humidity cell allowing you to get up to very high humidity. And on the left is one of our standard devices for mechanical loading. This device can be placed in the tomograph. We also take it to synchro-tom facilities. In this case it's set up for tensile loading of paperboard. But we also use this for compression and other types of mechanical loading. And then we take images of our samples as they deform in the tomograph. So we have all the software necessary for data acquisition, including complex sample shapes and stitching together large images and doing regional interest zoom scanning into particular parts of specimens. We also have software for visualization of the 3D images, including visualization and 3D rendering to analyze the 3D structure of the materials. We can also do image analysis and segmentation under special arrangements. So now I'd like to focus on the Zeiss X-ray Divide 20 Versa machine that we have and talk a bit about how that works and how we use that for multi-resolution imaging. And also demonstrate the acquisition using this machine. Okay, so this is the machine. This is our X-ray tomograph. In fact, officially we call this a 3D X-ray microscope. And you'll see why when we actually go in to look at it. So you can see, first of all, up here that we have these three lights here. And the fact that it's on green means okay, I can open up the box. When it's on red, it means the X-rays are on and I shouldn't open up the machine. So the first thing is this whole thing is basically a big lead lined box. And so it's radiation safe, the idea so we can work here with X-rays without any issues. So I'll open up quite a lot of stuff. And I'll go through what the different components are to explain what we have. On this side here, we have the X-ray source. Here we have the detector assembly. And here we have our sample, which on this occasion is the star of the show, which is this lever man. On this side, then we have the X-ray tube source. This tube source has high voltage passing between a heated coil. So we have a current going through one side of this heated coil. And on the other side towards the front here, we have a tungsten target and we accelerate electrons from the heated coil towards that tungsten target. And when they approach, when they come into that target, when the electrons interact with the target, then the trajectory is changed by interaction with the tungsten atoms. And when we do that, then we produce X-rays. And so the X-rays we're producing are produced over a small beam, a small electron beam, which has a certain size. And this case is about three microns. So it creates a spot on the tungsten. And that spot is the region over which we create our X-rays. And so we look in here, then this is where this is the little window here where the X-rays come out. And so then we have this, the X-rays produced at this three micro size spot. It's a point source until it is divergent. And so our X-rays come out here and then they form this cone coming out from the window here. Then come through and interact with our little gubba here, our little legroman here, or whatever sample we put on the stage. And then the X-rays that interact with this and they make the shadow image. So that is an image of the number of X-rays or number of photons getting through the material or not. So we get the shadow and the image we get on the backside of that coming through is then a map of the intensity of X-rays transmitted through our object. Okay. And then so then the third component, so that's the X-ray source where we produce the X-rays. Then we have our sample stage here, which we'll come back to in a moment. And then the third component is the detector system. So on this side here, we have a CCD camera charge couple device camera. And that's basically a digital camera, just like the one that, well, just like very similar to the one I'm using here to broadcast to these images. But it's a little bit better quality to say. And so that's just, you could replace this if you really wanted to with your mobile phone. And just like your mobile phone, this picks up optical photons, so normal light, if not sensitive to X-rays. Around the front here. So you see just just behind our Lego man here. We have this screen here. So this screen here is a scintillator screen. So that's the material, which when X-rays interact with it, it shines. And so the amount it shines is proportional to the number of photons hitting X-ray photons hitting. So that means it creates optical photons. So once you can, you could see with your eye, but also going to be picked up by the camera back here. So that means that X-ray the passing through object here. Some are absorbed from a scattered and some are transmitted. And those that transmitted hit the, the scintillator screen here. So it creates an image on the scintillator screen as it interacts with the X-rays, which can then be detected by the camera. Now, I mentioned earlier that this is a 3D X-ray microscope rather than just a simple tomograph. What's the reason for that? Well, you'll see we have this unit here, but also we have this unit behind. So these are microscope optics. And so actually what we have here, if you have the possibility to after the scintillator, either this scintillator or the ones in the front of each of these optics here, we can magnify the image, the optical image again. So we convert the X-rays to X-ray photons, optical light photons. And then we can just use normal microscope optics to magnify the image. So what we're going to need a bit allows us to get higher resolution without some of the constraints you'd have in a normal tomograph, because a normal X-ray tomograph just relies on this geometrical divergence. Now, at this point, maybe I should share the screen on here. Now, what you can see on the slide is that we have a diagram of how the source system works. We have our spot here. As we said, it's a point source. So we see a divergence of the X-rays or X-ray beam from this point source here. So we make our image here. If the object sits here, then we will see, we'll create the shadow, but then the X-rays continues to diverge. So by the time they get to the scintillator screen of the detector, the image has been magnified, it expands. So that means if we put an object instead back here, small object in here, we will see greater magnification due to this geometrical magnification. So if we put an object close to the detector, we'll get less magnification. It also means if our object is this large and we put it close to the source, we can't see the whole of the object. We just zoom in on the central part. Just because we don't have an infinitely small spot size, which would give this nice cone like this, we have some finite size, in this case about three microns, which means we have the superposition, essentially, of multiple cones. That means that we get some blurring at the edge of the object. So in general, the resolution we can get in a normal tomograph, which just relies on this geometrical magnification, is defined by the amount of magnification we get. So we've got more magnification. We have to move the detector further and further back, and then we get more and more magnification. Then you get bigger and bigger detector. And the other constraint on the resolution, then it's the size of our spot. Now, so in this setup that we have, then our little legroman here, our sample stays put, and we can move the source back and we can move the detector back. So we can change the distance in the source and the sample and the detector in the sample. So we can basically move the object closer to the source or close to the detector or move both outwards. Now, with this particular machine, this 3D X-ray microscope, then we have these optics as well. So that means that even after the detector, we can magnify again. So it means that we can actually maintain sub-micron resolution even when the sample is bigger and we're forced to work further back from the object. So let's go to a little bit how the source works and how the detector works. And then we have our sample here sitting on the sample stage. Now, that sample stage there is actually the rotation stage. And so we can position our sample and we can move it around to adjust the position in the image. The key thing is here that this whole thing rotates around as we'll see in a moment. You'll also see then a few other things that we have here. We've got lots of cables and things down here. And this is a system that's set up. You can see the cables coming from outside. So we can actually put devices on this stage and run experiments without the cables getting tangled up. Imagine if you rotate this thing by 360 degrees, all the cables will get tangled up. So this is designed to allow us to run experiments without getting tangled up. The other thing you can see is this big granite block here. And the reason we have that is for thermal stability. So inside the machine is running at 28 degrees and outside we have about 21 degrees. And the reason is if you want to do measurements down at the resolution we're talking about here, at sub-micron levels, you don't want to have a thermal movement inside the machine. Lots of components here that change shape inside just with a few tenths of a degree variation temperature. So the idea is that this will hold everything stable even with thermal fluctuations. Behind the little man there's a small webcam there. So we'll be able to see inside the machine as we, when we turn it on. So if we now close up the box, so now the yellow light is on. That means it's safe to turn on the X-rays. And now we'll come over to this machine here. Okay, so now we're looking at the tomograph machine here. And so what we can see over here, we can see the little, the R object, our Lego man sitting inside there. And then I'll take a step back. This is our Lego man inside the machine. We're looking at the optical camera here. If I just zoom out slightly. So you can see on this side here we have the source. This side here we have the detector. And this red line here indicates the height of the beam. And this indicates the center of rotation. The center of rotation is very important because it's defined whether we can reconstruct the object. So we position our sample to be roughly in the center on the center of rotation here in the center vertically. Then we can go on and then we can now turn on our X-rays. Here you can see the object we started our X-rays so I can now turn on the camera. And we should be now seeing it live. So you see this screen here. This is a live image of the transmission of the X-rays through our object. And so you can see that we have around the side. It's brighter and in the middle where we've got the object is darker. So this is the intensity on the camera. The darker means less X-rays are getting through and lighter means more. So what we can do with play around with now is that if I, for example, move the source closer to our object. Then you'll see it moving in over here on the optical camera. And as we do that you can see this image changing up here. So now we're moving the point of our cone towards our object. Now we can see the bottom of the X-ray window. So now we've zoomed in. So we've moved the object towards the point of the cone. So we're getting more magnification but we're seeing less of the object. Because he's now starting to go outside of the cone of X-rays that we're using. So I'll move the camera out again. This image now is live. So you see we've zoomed in on part of the object here. As we move the camera in, then you see we're zooming in on, so we've moved the camera out. Zooming in on the central part of this, of the little guy here. If you want, I can move this. Move the camera in a little bit more. The source a little bit more. So we'll see that moving in. It won't crush. So now you can see we've zoomed right in the middle here. I don't know if you can read the numbers up here. But here we have a pixel size. So each pixel in this image is 12 microns. And our field of view is about 12 millimeters. So we've zoomed in on a small part of this because we're only looking at a small bit of the cone. And so we're going to image a small region. So now if we move our source back out again. And we will see that we increase the field of view. So we're going to reduce the magnification. So image size there is the same. It's still a thousand by thousand pixel roughly. But the size of each pixel is changing because of the geometrical zooming. So now we've gone down to 33 micron pixel size. And our field of view is about 34 millimeters. So we want to get all of this guy in the field of view that I need to bring the camera back in again. So we're going to go up towards the guy. So bring him back in again. See him here. Okay, so we've almost got him. Let's go a little bit closer. Remember the numbers. We'll just keep going close until we get him all in there. And we can even go out a little bit on the source as well if we want. So we've got the two parameters we can change. And then we can see. There we've got pretty much all of our little man sitting there in the field of view. Now you might wonder why we've got two fields of view. Well, what I can do now is rotate by 90 degrees. You'll see him rotating around here. So now he's looking straight at the camera. And so I can take a picture on this side here. Now what we do with this is we're actually aligning the little man object in the center of the field of view. So we want our area of interest to be centered on field of view. So aligned with this line here and this line here. And if we align it in plus or minus 90 degrees, then that means that they should be aligned. The force of central rotation should be aligned in the center of this image on where we want it in the object. So we align in both directions and that aligns everything. Okay, but so we have here again done with you can read it, but down the bottom here, we can see an intensity value. So this is just the number of counts on your on your detector, which is like the gray scale in the image intensity in the image. And as we said, in the reconstruction, what we're interested in is the transmission. So this isn't really transmission. This is because the beam actually changes its intensity with position you see here brighter here than out here with the profile to this. And also this is just a number of photons recorded. We want the transmission. That's the number of photons recorded divided by the number of instant photographs. The idea is what we want to move the object out of the way and take a picture of just the beam without the object in there. So now you see he's been pushing them out to the side and hopefully moved out of the way far enough. Should check this beforehand. Yeah. Okay, so we ignore this little bit of the side. That's that's slightly unfortunate the six never what we've done now is that we have now. Normalize the image by the by the open beam, what we call it the flat field. So we took an image of the of the beam without the sample, which up for this little bit of the side, but ignore that the moment. And so now what we have at the bottom here is our transmittance. So this is the percentage of transmitted photo photons really. We also see we don't have this profile look back on this image here we have a profile from bright to dark because of this geometrical effect. Now we flatten the image. We've normalized by the instant profile. And in here then we see that we have up here about 75% transmission out here basis 100% which should be because outside the object and different amounts of transmission different places. So this is this is what we call the flat field. And so this is really what we're working with when we do the reconstruction. Okay, so that's basically how we set the topography and you can see here now this guy is the image we have here then is the transmission, and then we'll go through and reconstruct and make the 3d image of this reconstruct slice by slice so we take each horizontal in the sample and we reconstruct and then we can put those slices together to then make a 3d volume. So in this view here, what we have is a slice through our object in horizontal frame. And then the red line here corresponds to a slice vertically through the object and the green line vertically in north of the direction and you see on these two images your blue line corresponds to this position here. So we reconstruct slice by slice but we put these together into a 3d matrix, not 3d matrix you can just cut it up however you want. Now, usually you cut it up into three orthogonal sections. That makes life easier and it's logical way. But if you want, you could cut it up. We still see all the sections but they don't have to be on the Cartesian direction. So let's just list this guy any way I want so we can get different views on this is just a 3d matrix do whatever you want with it once you've got it and you'll be able to play with these images in the tutorial that is online. These data here the Lego data here. And so what do we see in the image so we see gray. So we just have a single scale of value, and this is the ratio the attenuation coefficient. You also see that you have some areas, which are brighter than others. So brighter means more attenuating. So it was just a complex scattering effect. This would mean denser but there's also a photoelectric effect so there's more complex but it's suggesting that this is the first thing is telling us that this is interacting with x-rays more. And first order assumption is it could be density but there's more complex. Over this side here you see a histogram. On the left hand side I don't know if you can see what analysis. So this is the histogram of the grayscale values across the bottom is the grayscale value in the image of the intensity in the image basically. And so the intensity of every pixel. So then here's just frequency to a number of pixels or voxels with a certain grayscale value. And we have zero hits of black here and white here. So I can change the level of this of what I call black, so that we then increase in the value, what would find a zero. So if we do this we're now saying it's transparent so we can see through it and suddenly a little guy has appeared. You can also change what we call one just to make it brighter so this is just contrasting the intensity. So now this is our 3D reconstruction and or 3D rendering. So you can see that we have the bright part of the hands then the breathing apparatus here is probably the darkest part and this is somewhere in between. So yes we have different plastics in here. At least different density plastics so we can see variations in there in the attenuation coefficient. And so we can play a few more tricks. So this is just intensity value well, I don't really like him being great that's rather boring. Let's make him orange. That's much, much more aesthetically pleasing obviously, but this is just playing around the image all I've done there is just use a different lookup table for the colors. I play with this such I only visualize certain parts. So with this button here I'm making everything below a certain grayscale value transparent. So you'll see that the darker things slowly disappeared increase this here. Until we just left with the densest object in fact I can even make it. So we just see the really dense parts. And this is kind of manual segmentation of this image we were manually breaking up the image into different palm what you can actually do in the images is identify just this this bit to give us a label of like right hand this one label left hand and just extract that part of it. So all kinds of games you can play once you've got the data. Okay, so we're back to looking at the full image. As I said it's a 3d image. So that's the matrix and we can just cut it up as we wish. We can just take part of it and cut his head off legs off. So we can now look inside our object here. Equally we could do that in the slices and move this in and out and zoom through the object and look at different levels. So this is this is the data we get out. So we get these maps is 3d maps of attenuation coefficient, and then it's a matrix that we can then play with. We're going to go back to the machine. Not forgetting obviously to turn off the x-rays. Turn off the x-rays there. And we'll go back to the machine here. And we will change samples. Now we're going to look at the new sample is a sample of the cast iron which is a ferrite matrix and a complex multi-scale structure volume graphite spheroids and carbide carbides. And this material is often used for a truck engine box for example. We're looking at a mini tensile test specimen which has already been deformed in tensile loading. And the central part has a diameter of about 1.4 millimeters, which is where we're going to be concentrating in this example. So now I can come back to this screen here. Remember at the start we had this screen here. So we're going to line our sample approximately with the beam. So we're going to put it up here and zoom in so you can see it's a bit better now. So we want to align the middle of our sample on the cross. We'll just move in the motors. So that's one angle and I'm going to rotate by 90 degrees and make sure we line it. So the leg amount was a little bit off-center. So there we've roughly aligned our sample. It's a much smaller sample. And we want to zoom in on this. We're going to change the big optic here to one of the microscope optics behind. So I'm going to change what we call the 4x optic. So with any luck, now you see the camera coming out to the back, moving across. And then we're putting in the 4x objective, four times objective. So the previous one was a 0.4x. So that means actually we take this image of this large detector and then we de-magnify it basically to fit an optical detector. And now we're taking the 4x optic so we can magnify the image. Okay, because it's small, we want to move the source in here. So now we're going to get most of the cone onto the sample, most of the photon onto the sample. Try to do it without crashing, obviously. It's always a little bit of a sensitive moment this. Okay, so then now we can turn on our x-rays. And we'll just see if we get an image. Yeah, so there's our sample. And so we need to align it to the center of rotation. So we'll have the x-rays on here. This is the live image. And we'll just double-click in a part of the image that I want to move to the center. Then it will move it to the center. And we want to align this approximately so it's centered in the beam, where the beam should be aligned with this cross here. So you can see on this image that this is the gauge volume area. This is the thinner area in the sample. Then it's thicker at the top so it's darker up here. So the less photons getting through up here than there are down here. You can also see there's some texture in here. You can see the white lighter bits. So we said lighter means more transmission. Darker is lower transmission. So there's clearly some heterogeneity in the sample. So this could be kind of interesting to reconstruct and look at in 3D. So we've aligned it in that direction. Now we will align it in the other direction. So it should be rotating now. You should see it in the image up here rotating. So now we're rotating by 90 degrees. We can take a picture of it again. So we can just move across slightly just to align it here. Now you can see that we have a little bit of space still for the sample. So I can be a little bit braver on this. And I can actually move the source in to get as many of the photons onto the sample as possible. So if we look at this now, then we have a pixel size of 1.5 microns. And our field of view is just over 1.5 millimeters. Now to make this go slightly quicker, I'm going to do a few different things. So our image here, so the detector we have is a roughly 2000 by 2000 pixel detector. Now that means that the field of view in full resolution is 2000 times the pixel size. Does that make sense? We have a certain dimension depending on the zoom and then we have 2000 of them. Then the field of view is 2000 times that. And we usually use it in binning two, which means we've been together two by two pixels. So it becomes effectively 1000 by 1000. Why do we do that? Well, it's because then we get more photons per effective pixel. So we can go faster. So we see at the moment we have counts of this green value at the bottom here. We have counts of about 1600. So and ideally the instructions say we shovel about at least at least 5000 projection. 5000 counts. So this is the intensity on the detector. So one thing I'm going to do, I'm actually going to bin the camera four by four. So each pixel now in this in the image we're going to have is going to be effectively four real pixel by four real pixels square. So now we're just by doing that, then we see that we have. Intensity now of 6000. This is this is great. But now I've reduced the resolution down to three microns per effective pixel. And the size of the image is now just 500 by 500. So I've reduced the resolution and in most cases we wouldn't do this, but we want to to get a faster image here. Now the other parameters that we have, we have the exposure time here. So if I put this to two seconds, then if we just hit at one point here, we've got seven and a half thousand counts there. Now we're up to 15,000. So doubling the exposure time just leads to doubling number counts. And so then it's a trade off between how long it takes to acquire the image versus the signal to noise. And we know that roughly 6000 counts is enough to get a good image. So we don't need to go much higher than that. We also have to find our objective here. We define the binning. Then we have the voltage. So this is controlled acceleration voltage of the of the electrons and defines the energy of our X-ray beam. So 140 kV means the maximum energy of X-ray to have the 140 kilo electron volts. And we can make this lower. So 80, that means the maximum energy of 80 kilo electron volts. Higher voltage means higher energy means better penetration through the object. Then we have power here. And the power then, so you've got voltage times current is equal to power. Here we control the power and the voltage. So by changing the power, we basically changing the current. So that's just changing the number of photons. So increase this value here from five to 10, then I would double the number of photons. So again, I could double the number of counts in our image. We've set up our sample. We've got three micron resolution size. We've aligned our sample in the middle of the field of view. This is important because when it rotates, we don't want it to go outside the field of view. Otherwise, it wouldn't image the full thing. It rotates and goes outside. Then we lose part of the sample. I can also get rid of this image from the recipe point again by just rotating back to that orientation and taking a picture. Not too much, but then we have the same parameters. Okay, so then there's another factor here. So we have this 140 kilovolts. So that means the maximum energy that we have being produced, the XAMG is 140 kiloelectron volts. And then we have a spectrum down to, I don't know, towards zero, but you'll see maybe some cut off around 10, 20. Okay, so we have this spectrum, this broad spectrum of up to 140 kV and down to, say, some like 10 kV. And it's this spectrum of energy. A lot of those low-energy X-rays, particularly with a dense sample, aren't going to get through. A lot of these photons are just not going to get through material, particularly in the middle, but they'll get through more easily out here. And this led to an effect called beam hardening that we mentioned. And one of the best ways to solve that is to get rid of them. And so what we can do, I've just realized I have to move the source back slightly because I'm going to crash in the moment with the next step. So I'll just move the source back. I'll just check now what we have for a pixel size here. Move the source back. So now our pixel size has gone down to 3.45. So if I want to maintain the same pixel size, then I have to move the detector back to compensate for the loss of magnification on the source side. I have to move the detector back to recover the magnification on that side. So there we go. We go back, we zoom back in. Now we're roughly three microns again. Okay, now you'll see why I had to do that. Because on the front here, I forgot to point it out when we're looking at the device, but there's this wheel here which has filters on it. And these filters allow us to filter out the low energy x-rays. And the procedure for doing that is that we have our picture here. Now we're going to take our reference. I can move downwards. So we'll take a reference image. So move the sample out of the way. The machine behaves itself, which it's not at the moment. Oh, there you go. Yep. So we've now taken an image correctly without the sample still in the beam. So we took an image of the beam without the sample. We saw it move down out of the way. We took an image and now we've normalized by that. And so we've got rid of this variation in the intensity of the beam across the image. And also now, you look here, we have transmittance. Rather than over on this side, we've got intensity. So we see we have about 15%, 16% transmission. This is the proportion of the photon making it through the sample. So we look at this table here. And you see here, we're 140 kV. And then we look here, this is the transmission. We have between 12% and 20% transmission. And that tells us that we use this HE2 filter. This is a filter that's going to cut out the low energy x-rays. So we apply now on that. And you'll see the source move back. We rotate the wheel here and remove the filtering that we want. And this is why I had to move back the source because you see there's some extra thickness there. And it would have crashed if I hadn't moved it. Sometimes you do remember to do things and not embarrass yourself. And so now we filtered the x-rays to get rid of the low energy. This should hopefully stop the beam hardening artifact from being an issue. So we take an image now of our sample again. Now our counts have gone down slightly. It could be removed a bunch of x-rays to the filter. So we take the reference image, moving our sample out the way again. You see it's gone down that way. We take the reference image. And now we see we're around about 30% transmission, which is not bad. We generally want to do about 20% and 35%. Why do you want that value? Well, you want to basically get enough photons going through so you have a signal. We block all the photons. It's got a black picture. But if all the photons go through, you have a white picture. So 20 to 35% transmission means you can interact with the object, but you're still getting a signal. So you're actually going to see something in that. So you can see that this image has a bit more character than this image here. OK, so now we've chewed our parameters a bit. We've positioned our sample. We've selected the filter to stop beam hardening. We've aligned everything. And now we're kind of ready to go. That's basically how you set up a demography. So we can go on to the next page. And here we have a few parameters. We're going to tell it to collect a reference image where we take the background image just once, because you do it multiple times. But before it starts to scan, it's going to move the sample out the way, take an image of the beam, and then use that to correct all the images that we acquire. We're acquiring for one second exposure time. We are going to rotate over full 360 degrees. Now, the number of projections. This is the number of angles over 360 degrees to measure forward that we use this rule roughly of number of pixels across detector multiplied by pi by 2. So we have 500 pixels across the detector with this binning mode. So 800 and one is about about that. So you can see then we have now have scanned time at 36 minutes. Fortunately, I've already done this, so we don't have to go through and wait for this time. But you'll see that that if I go up to 1600 images here, then we go up to an hour, an hour and 12 minutes. If I were to change the exposure time to two seconds, then our scan time also increases. It's not being linear because of lots of other movements. So this is why we're trying to tune the parameters in terms of exposure time number of projections to not get too long a scan. Usually our scans are in order an hour to an hour and a half, but we have some scans which could take 24 hours. Maybe we're doing really high resolution. The higher the resolution, the lower the accounts we generally have. Okay, so then we set up our system and now we can click on start. So start to the measurement. The sample has moved out of the way. It's moved down away from the beam. So now it's taking a picture of the beam without the sample there. So this is going to give us our beam profile, which as we say, we hope is the same as all of the projections we acquire later on. Now we're moving back up again. So it's now back in the beam and then we should see that I didn't align the sample perfectly. Wow, okay. So ideally that wouldn't be so close to the edge. I didn't set up perfectly, but now what we see is requiring our data. And there's one strange thing you notice here that each of these images is now able to rotate the sample. You can't necessarily see it rotating. What you can see is it dancing around. And that's kind of maybe a bit weird. Why would the sample dance around? Well, it's essential the sample is stable and its position relative to the central rotation is always the same or that we know how it moves. And what we're doing here is something called dithering. And so the sample is being deliberately moved around so that each image is at a slightly different position on the detector. We know the shift there. So we can correct that and move it back again to do an artificial displacement. And the reason we're doing that gets back to this bad pixel issue. If there's a bad pixel on the detector, then we don't want it always to be in the same place because that would lead to the ring-out effect. But by moving around, we're just blurring out that effect. So we're going to acquire now over the full 360 degrees. And that will give us the data like this. So the shift, this dancing around, has been corrected for. So now we just got the rotation of the sample. And now if you look at any, you look at this region here, for example, as I move the slider, you can see it looks like it's rotating around. So now we've got the projection which is corrected for the dithering, the movement, and corrected for the flat field. And so this is a transmission image. White means high transmission, dark means low transmission. Thicker part of the sample, much darker, no sample, much lighter. You can also now, even on this, see features which look kind of interesting. That could, for example, be fracture because a fracture is void. And so this is where you can let more x-rays through. It's going to interact less than the far-right matrix. And so that is possibly indicating open-nose. These other regions are probably lower density material. In this case, graphite. And so again, where you've got lots of those, then you see more transmission. So we have now our data here. So the next step is to reconstruct. So we go to the reconstruction program. And one of the first things we do is we check the center of rotation. So what I'm going to do, just ignore this for a moment, we have our data here. Rotate around. We're going to reconstruct a single slice here with a guess the center of rotation. Remember, we said the center of rotation is really important because with the back projection, we have to project back through the image space always through the correct center of rotation. Now we've aligned our sample and we have a good center of rotation defined in the mechanics of this instrument. But it's always some slight shift. It's not always perfect. So we can actually now check that by just testing different centers of rotation. So I'm going to run through a range of centers of rotation and we'll see what the effect of that is. So here I'm going from a center of rotation off to one side sweeping through and going out to the other side. And so what you can see is when we get the best center of rotation, which is going to be about here, we get the sharpest image. This is where all the back projections are the best across a center of rotation to project across. If we go too far to one side, so we have a center of rotation which is shifted to one side of the detector, then you see that things don't align when you back project them and you get these ring-like effects defocused. If we go too far displacing it the other way, then we get the same sort of blurring. So to define the best center of rotation, we hope it's perfect in the machine anyway, but yeah, it's usually some offsets, some misfit, and so we actually do a test. There's an algorithm that can do this automatically but we can do it by eye and we choose the one which gives us the best, the sharpest image. So there we define our center of rotation. This is a key parameter. Okay, so that's our center of rotation. We've got a nice sharp image. We've found the right center of rotation for our back projection. We put the filter in there to try and stop the soft X-rays which are going to cause us problems. But I also said that sometimes we can give this a helping heart. If we haven't corrected for that fully with the hardware filter, we might need to do something else afterwards. So what I'm going to do now is I'm going to run through a set of beam hardening parameters just testing for this, this slice. And then we'll look at what sort of beam hardening we have. So I'm going through processing some projections with different correction factors and then we'll look at what the effect of those are. So the ideal thing is you correct it all in the acquisition so that you don't actually acquire those soft X-rays which aren't helping you. And that filter we put in should have done a good job but if we didn't acquire the best one, maybe we should have actually used a stronger filter to knock out more soft X-rays then we will still get artifacts. So we'll go back here. So without any beam hardening correction, this is just basically reconstructing the image as we acquired it. And you can see that we have this profile across the sample which you might think, oh, it means it's lower attenuation in the middle than on the outside. But that's just an artifact of the beam hardening. So this is a classic telltale sign you have beam hardening. Now if I take the scroll by here, then we'll apply a beam hardening correction which is basically working to flatten this curve and then we're aiming to get a value which gives us a flat profile across the sample. Unless we have some other information tells us that we have this gradient which is not usually the case, we're looking for flat profile. We go too far, you see just at the end there it's starting to bend up, we're over correcting. So our best is probably somewhere around about there. So this is where we flatten the profile. This is just a post-processing procedure and ideally we should have probably had a slightly thicker filter to remove more of the soft X-rays. But we would take that parameter there and so now we define our center of rotation and our beam hardening parameter. And so now we can just run the reconstruction. So we will just press go on this and it will do the back projection, the filtered back projection to produce our 3D image which is basically like the slice you see here but it can do for all slices in the sample. Now, again, I can just switch to one that I prepared earlier. First thing you do with any data is to explore them and get an idea of what you have. And so you can start to see through, okay, we've got darker things here, we've got brighter things here, we've got these funny shapes here. But then you can also visualize it in different planes actually in this direction. Look, there's some interesting features like this which may be something slightly different. You can see these white shapes here in different sections. And again, in this direction here, you can see something kind of cool there. And remember that this has been deformed. So some of these features are going to be features from the materials and some are going to be from the deformation. But then remember you have this histogram here. So I can then define, I can change the contrast in here, I can set what is zero. So basically I can make everything below a certain value over here, black. And likewise, this time, everything below a certain value, white. So one of the things you'd be able to do when you're doing image analysis is to find upper and lower thresholds like this which allow you to make the image black and white. And you'll be able to then separate out the dense stuff from the low density stuff or the low attenuating stuff from high density stuff. So here we've made almost a binary image. So we separated two of the phases. Equally, we can go in the other direction and make everything black except for the really white stuff. And there we're separating out another phase. Now I'm just doing this in the program here, but you can actually do it by, with MATLAB for example, to actually process this properly. You can also make 3D renderings of your object which allows you then to explore the internal structure a bit more. And then so we can play around with this again, change parameters. We want to look at the bright stuff, everything else black. Then now we can visualize the 3D structure of the white phase. You can see that's kind of a cool shape to it. So maybe I'll make it slightly brighter, bring that down here, bring that back up there. There you go. You can see that. So you can see these structures are quite interesting 3D shapes. And we're going to try and characterize those shapes to understand the material. And we will do much more with proper software. And then you can see some features on the outside there. Some of these shapes are due to the machining of the sample. Some of them are due to deformation. So that's our material. We can cut it up however we want. You just look a little cube of it for example like this, which is kind of nice. So then you can just zoom in over a smaller volume and maybe understand a little bit better the shapes that you have that form. So there we have our end result. Thank you very much.