 Great to be here in Lund so to speak. And as you said I'm telling you about our approaches and recapitulating a little bit the methods but then this leads to applications and the applications are in bio-imaging and in particular tissue and so I put up here some reconstruction, rendered reconstruction from brain tissue, human brain tissue, human lung of a COVID-19 patient that's Marina's subject then in the following talk and also small animal heart that's actually an in-house result so we go from laboratory, micro-city all the way to high coherent synchrotron radiation and it's an entire team behind this talk in these efforts and in particular my present in my past PhD students and as a staff scientist in our group Marcus Ostoff is in charge of much of the genics end station that's the beam line or not the beam line the end station that we operate and that we're built for special purposes at the petrary storage ring. As biophysicists, we are interested in looking at functional units from structural point of view and for a lot of scales in biophysics and biology it's true that that structure makes enables function dynamics is important as well but it all starts with figuring out how this works in space and time and resolution is important but it can't be overstressed but from the molecular to the organ there are many scales hidden or let's say intertwined hierarchical length scales from molecules to organelles to cells in imaging the challenges to move that window that you can cover back and forth to zoom in and to zoom out and the resolution is as important as the field of view and of course ideally it's all three-dimensional because our world is three-dimensional and that entails a lot of work also in optics to cover this just one example in getting in here in a consortium of in driven by neuroscience we are interested in understanding the connection between two neurons the synapse that's what the transistor is in your CPU that's in our that's the basic unit where signal transduction and modulation can happen and in the human brain they're tend to the 14 such synapses so ideally one would like to have this as a computable unit in virtual space so extremely crowded extremely many constituents how do we go about it if we do structural biology well we take just constituents and then try to understand something on this protein on this lipid it's already quite an endeavor to say hey let's have a functional synaptic vesicle filled with neurotransmitter can we look at this this is also a lot of modeling involved in biochemical data and then integrating structural biology so also for high resolution methods like x-rays or neutron scattering you work a lot with model systems here we were interested or are interested in the merger between membranes and how can we understand that and find out about curvature and and this is in the framework of this talk it's just to remind ourselves that diffraction is a great thing classical far field diffraction you really get you really get even in aqueous functional units you get really structure on on sub nanometer angstrom scales and yes this is extremely important but there's a but you have to have a model system you have to have quite repetitive structures you cannot look into something that is very heterogeneous and even if if you if you go up the level to the organelle like a synaptic vesicle yes you can still treat something by small angle scattering in solution averaging over many many organelles in in your cubit but you have to model and you lose a lot of traction because of that averaging so that completely breaks down if you go to a cell or even from a cell to a tissue into an organ the human brain is is a tremendous challenge and people use all kinds of different methods to get beyond the anatomical data go and to retrieve the cyto architecture and ideally you can do a lot with cutting and slicing and staining but ideally you would want to do this on a higher number of tissue samples for instance to understand neurodegenerative diseases or to understand connectivity in in the slicing and classical histology or even em techniques it's very difficult to reach isotropic resolution and very difficult to cover large volumes so here's a tremendous opportunity for for x-ray imaging what's the problem then in looking at specific configurations with x-rays well x-ray microscopy is around it it does work it works with lenses but there are challenges loosely put if you consider that classical light microscopy is limited by the fraction in abe formula so it's the numerator lambda makes a problem but you know the denominator the numerical aperture can be quite quite high for x-rays you should win a factor of three orders of magnitude but unfortunately your numerical aperture also goes down so working with zone plate lenses there's progress all the time tremendous progress but but still you are far beyond let's say lambda for hard x-rays and there are other inconvenience like several orders or dose issues so classical x-ray microscopy is limited by not the fraction in the sense of this year but in the sense of this year you cannot easily you know focus x-rays and that that carries through all of the optics you want to do basically the challenge is to do optics in the asymptotic limit of every material that we would like to make an optical component of being almost like vacuum yeah the index of refraction asymptotically you know beyond the resonance going to to one for in a simple oscillator model so so that's that's the challenge and so the idea is around since quite a time now to do x-ray microscopy without lenses to use diffraction you know magnification by the wave equation alone the smaller the the structure that you diffract from the larger the cone you capture that and you solve a phase problem and this is a very powerful idea and roughly speaking what we do is a variant of this but most classically you would have plane wave illumination you would have an object plane you would have a detection plane ues you would work in kinematical approximation and the intensity you measure is the transform of your electron density distribution in the plane modulus squared so this is where the phase problem occurs and this is now well understood under which conditions there's a way around it basically in short you have the measurement the data here on the detector but you may also have additional constraints such as simply positive the defect their fitness of electron density a compact support sparsity any such information can help if you consider it for your problem to be uh not so invasive and then using projection algorithms this is their their powerful tools mathematical tools to solve you know what is a high dimensional matrix if you want so but but but many many interesting issues signal theory wise and of course resolution those resolution practical implementation it's it's also a challenge when we started this we work with something that is very powerful and uh uh manuel is here in the audience and others are using this to too much benefit this is the idea of using partial overlap between successive exposures to solve the phase problem you have to create more data than unknowns you have a test structure you go around in a spiral or some scan and look at the diversity of all these uh measured diffraction patterns that's by now an old idea it has been matured it's it's probably the best tool at hand if we go to high resolution and just as a I'm still nostalgic about it this was the first time when we tried this just after the algorithm had and the basic ideas of using simultaneous reconstruction of object and probe came out we tested this and we're already happy about such a reconstruction only two years later that is 10 years back by now wow you could even see a test structure as it should be fair resolution despite the fact that the beam the probe is completely aberrated it has horizontally and vertically different focal widths so uh you have these um these these aberrations and yes you could also use this for biological imaging and for tomography there is an issue and this is why we did not continue along that route and this is for the simple reason that we were in most problems that we were considering with our collaborators interested in very extended samples uh in tissues and uh something that was not compact enough or small enough to really do the lateral rustering and then the the 3d tomography even though this can also be scaled up so let's start at the opposite end let's go to the hospital and take look at a radiography classical absorption based radiography you get almost a um you know a projection image as if it was geometric optics with a shadow mechanism that is given by Lambert's law a central simple linear absorption and then the absorption coefficient varies with composition and that gives you ability to see uh in particular mineralized tissue but for soft tissue it's challenging this is a small animal cochlea but we start to see at this um in this projection recorded at liquid jet source with six micron spot size at at a distance where you have some partial coherence you start to see the edge enhancement and that the fact that the x-rays are actually waves so you can beyond you can go beyond this and you can record um a tomographic scan and then hope for more visibility for soft tissues so what follows then is full field tomography and the fact full field is really this macroscopic benefit of covering in this case the chest of a small patient that's a mouse and here's the signal the variation of the attenuation curve and this is just uh you know recapitulation for those of you who haven't worked much with the tomography you record or let's say you you you you complete your sinogram for each of the lines in the simplest case and see how these these bones so so as we move around while while you turn the specimen and using filters and back projection in the simplest case I had to do it um iteratively and more sophisticated but often this is sufficient and it's also fast um you can reconstruct this to the slice and the similarity is similarity or parallel slices and yes you see bones and also soft tissue you do see the lung tissue here and uh this is um this is not yet exploiting really face contrast but now we want to go better and we will go and want to zoom in we want to see the small soft the tissues as as good as we can for the small animal so here we exploit the fact that the index of refraction is complex valued we have attenuation of course but also phase shift and by self interference of the beam after it leaves after it penetrates the object um it may be attenuated it may be phase shifted so then these this this wave interferes with itself and phase information is converted to measurable intensity in the detection plane detector has to be a little bit far spaced there has to be a room for propagation of the wave field and there has to be partial coherence uh sufficiently high but if you have this then the fact that delta for um small z materials elements is um orders of magnitude higher than beta gives you an advantage so so that's that's that's very good the image formation of phase contrast is is is well understood and it's a nice chapter of you know an optics course you start with Fresnel Kirchhoff integrals you have an object uh variation in absorption or phase and then the whole thing evolves and it's also very nice to see and to understand how you can this can solve this how you can simulate this on a discretized grid by fast Fourier transform and and exploiting the different ways that you can write this these propagation and very easily and very quickly if you have this in the lecture or in exercise the student can calculate um how absorption and phase um varies significantly emiss systematically with the Fresnel number yeah this can be all the way to the holographic regime and the holographic regime at smallest Fresnel number f is is important because here your sensitivity to to small contrast is is highest let's look at it more quantitatively we take a in this case real 3d phantom that's that's nice we have a plane wave illumination we take um indices that would match let's say a protein at at a relevant photon energy and then you can see what this would look like without noise or with noise this is without noise but um as a function of f so credit to Dennis Gabor who who gave us inline holography at his time going back was really hard you took the analog film as a second sample and then let the wave travel backwards but in that case your intensity acts as a phase and your your amplitude is is disregarded and and this gives simply not a very good reconstruction yeah so a challenge is really the inverse problem how do you go from the measured hologram to the sharp exit exit wave and the holographic reconstruction is problematic if you if your object is optically thin you can you have an explicit approximation for the propagation you can express it in Fourier space of the wave in lateral coordinates in other words the Fourier transform of the hologram can be described as the Fourier transform of the exit wave times a multiplicative oscillatory filter so image reconstruction is a structured filter a regularized filter and this is what is known as the contrast transfer function and that works fairly well but not good enough for many samples and and requirements and then we have iterative reconstruction where as in cdi but now just with the different propagator we cycle between the object and the detector and and we just apply our measurements and maybe additional constraints or simply two measurements in two planes and no additional constraints and everybody puts in what she or he has yeah in that sense works beautifully is a bit more involved computationally but gives you really very good very good quality so image reconstruction for a long time in this near field regime has really been focused on one step steams built based on transport of intensity equation for instance the famous peganin approach or and that was already so to speak high end the cdf based approach if you let go this requirement of having a one step operation you can really go base you can really leave linearization and in our view this is important long story back we come back to the mouse we have the absorption contrast image but now we focus in and we measure this with a high resolution detector so that the coherence length is larger than a pixel and here we see face effects and we get a chance to see the alveoli the tiny air sacs in lung that's good but if we don't do any face retrievals a step our reconstruction our reconstructed slice looks still quite noisy and we cannot segment if we adapt at least if we take a little bit of effort of face retrieval we can now distinguish these components quite well so so that's that's the benefit even if what i have telling you was really the idealized face retrieval in the lab of course this is not i mean it's a broad spectrum it's a large source so so just barely enough partial coherence and there are inversion schemes which are optimized for that too so that that then gives us some recreation we can look at something that we have reconstructed it's a movie that that martin krenkel in the group compiled from then from that data and because we will hear more about lung later in marina's talk it's a pleasure to play it and also to remember that laboratory techniques in the meantime are quite powerful and in within a field of interest here in the lung where we used the high resolution detector the higher resolution fiber coupled detector we now reconstruct really the alioli walls in the soft tissue and can do statistical analysis and on this despite the fact that it's an entire mouse yeah so no definitely the entire small animal now what can't we do what where does the lab not allow us to go a little bit further this here is again lung tissue again marine for mouse but here we want to see cells in 3d in green we have macrophages and again we have a zoom in in the center we can zoom in at different magnifications but this is something where we need if we if we leave this you know just barely I must say the resolution that we can achieve with best detectors microscope based detectors now we have to really magnify and we need very small spot and saw spots and we need high coherence and we go to the synchrotron another example three examples where you see this line between laboratory and synchrotron and I must say the face contrast homographies when when third generation synchrotron sources came up in the 90s that quality is we now got much better quality in with laboratory instruments than we had at the time with synchrotron radiation yeah and then the this exelum source the liquid jet technology pioneered by Hans Haertz and his group has played an important role in this so just to have the power density that is required it allowed us quite some time back now to to to use this for an actual important problem and application development of optogenetical implants with the with Tobias Moos and his group here in Goettingen and the challenge was to see nerve fibers and tissues and membranes within the bony you know environment of the cochlea and if you understand your resolution in your instrument you can really go from something okay I'm gonna see soft tissue to something yes now I see it sharp and now I can with this added quality I can really segment automatically a gray value base or whatever you use in addition it all scales with the image quality and we were happy that that worked but how about really taking the same cochlea similar cochlea but now seeing every single cell yeah this is again something where you would need synchrotron radiation this is a result that we got for parallel beam illumination not even any magnification 650 nanometer voxel size but here now you can really zoom in to the functional interesting parts of the cochlea the organ of corti and these are neurons that are important in the filtering of the auditory signal so called spiral neural galleons and from such data here you can segment each and every neuron and here you they are color coded according to size you can you can once you have this 3d data you can compute sphericities and sizes and densities and neighborhood information so there's a lot despite the fact that the resolution is not you know that high but few of you is gigantic it's functional in a sense that that you don't cut to something that you cannot interpret you can see the tonotomic map and that was that's from from a publication that came out yesterday and we're we're quite happy with this this quality but in the lab you can do a third example human tissue that the previous was was durable small animal models for that research project but here's a sample from neuro pathology so it's a biopsy punch it's human hippocampus embedded in paraffin that's the standard that they do in each pathology department and it's a millimeter cross section and here the pathologists and the medical doctors are interested in the process of plaque formation in alzheimer's disease and very very clearly you can you can see the overall um tissue cyto architecture already with the liquid metal jet recording and then that allows us to see vessels and calcified lab vessels and the the granule cell band of the dentus gyratus and maybe again you can hear some more on this in marina's talk so that's possible with a lab source but if it if you wanted to then go after small dendrites and connections and maybe synapses at some point that needs synchrotron radiation that's a reconstruction from synchrotron radiation um which allows you in this case these are quite large cells in human cerebellum so called prokinae cells but you have the cellular resolution and you see a nuclei and you see substructures and dendrites and axons so so that's that's more geared towards also neuroscience how do we do this that's now the next topic that was the motivation we need some magnification here we need magnification and resolution first of all when we talk about image propagation and reconstruction whether it's parallel beam or ideal point beam that's the same thing there's a simple variable transform and the um let's say um wave optics preserves geometric magnification as you may have guessed but the the propagation distance has to be scaled so that's good you read about this in Paganin's book uh perfectly treated um how do we do it experimentally we have to focus and if we can't focus uh uh well enough and that does not only relate to the spot size but also the coherence or the stability yeah if that jitters because your monochromator moves or um if that has a wave front that is not clean one thing that helps that we uh found was was helpful is to filter this by by use of an x-ray optical guide wave guide so we use guided wave optics to select certain modes and to use the exit as a virtual point source that offers really a benefit so i'm now talking about uh wave guide optics for high resolution holography how does that work it's a special approach they're different uh different approaches it's not the only way to to go to the nano scale with holography as as you know um but it has some uh some pros and definitely also some some cons typically we lose photons in our instrument uh we focus to 200 nanometers so we overexpose the wave guide channel and then um we work with 10 to the nine photons per second roughly and the sample is in the defocus and is then you know we record the projected hologram this is a reconstruction of the intensity superimposed with the wave guide channel this is a hundred nanometer uh scaled bar and this is where we do this we have an instrument where the sample and the wave guide is in air this has also pros and cons for complex sample environments it's a challenge but we can also run wet samples like also wet neural tissues in buffer or methanol everything is installed here at the p10 b-line of petri three of that storage ring and uh yes this is where we work we are proud of this we like to take photos of this configuration in that configuration a while ago we also temporarily installed a stead microscope we have here in Göttingen but designs so we can take it to to Hamburg and side by side take X-ray and and optical recordings of cells so so that that that works and and this is is promising but now let's focus on the the wave guide optics you can couple in a beam from a side from the side within cladding and here we use front coupling really we really focus on the entrance and then we have formation of modes we can simulate this with finite differences and study precisely what length and what lateral diameters we want these are the simple schemes we can we can test them against analytical schemes and by now we even have time propagation we can even could even see for for short femtosecond or autosecond pulses how how modes propagate differently in time and that that can also be done for the wave guides we use them as simple channels for imaging but we could do more complex geometries we can think of splitters we could also started to fabricate splitters we can curve channels we could delay channels we could design structured illuminations in that sense for instance to have an additional benefit from off-axis holography encoding also absolute phase shift or from tilting the beam towards high angles and having heterodyne phasing at high numerical aperture these things can be done and experimentally we showed that we can have wave guide exit at something like 20 degrees if you simulate it really even a circle would be possible with 30 percent transmission an extreme whispering gallery mode then in in in that case the first millimeter of you know propagation along the path would look like this the last like this for straight and for curved the mode is a little bit you know pressed it's it's like in the car by centrifugal force against the outer cladding yeah so to speak that that's that's that's nice optics of curved wave guides and maybe there's some interesting applications to come up with with the free electron lasers but here it's all about filtering filtering means that okay we have different modes of the source but it also be a mirror or monochromator so your ideal focus that you simulate is not really what you have you would have be looking at something like this and this filtering means that different input modes you know you lose photons but in the end this becomes much more stable so we studied analytically and numerically this propagation of coherence and it looked attractive to us we spent quite some effort to do the nanostructuring and to build wave guides from layered sequences and then crossing two such planar wave guides or from the photography can be done and this is then the exit wave intensity that we record on a detector five meters away compared to the mirror of course we may not have the best mirror in in in town so to speak with the more height errors but if you go to an x-ray beam line ask for the probe let's always have the probe because you can divide out but you will always be limited also in in in the probe and it's this quality that then really gives us very decent very ideal holograms that have fringes all the way out to the detector and that's just a simple example a test structure comparing the holographic reconstruction with twin image problems and the CTF reconstruction that you know makes some assumptions which are well warranted here actually the it is a weak object but the iterative reconstruction with an automated support finding a step gives you really the best the better reconstruction now when we used that for imaging we found that we could get reconstructions at surprisingly low dose compared to our own let's say pure graphic tomography reconstructions but if you study that problem there shouldn't be such a difference so it may probably be more in the experimental determination and if you are in a defocus position and it's it is it's in a sense it helps not to overly waste photons that that you would have in the focus I mean in the focus you should be able to move faster and and scan but you have to critically interrogate your focusing optics also in view of having too many modes maybe modes that don't contribute so for wave guys you always come from from from lower fluence or flux density and you have all the intermediate steps it's a while in in far field cdm tomography it sometimes you have the feeling if it works you get high resolution but you also have high dose and it's how how continuous can you can you go up and down that that's that's still an open question here it worked beautifully to reconstruct material cells or eukaryotic larger cells macrophages we can compare different reconstruction techniques so yes and and and this is an advantage you still can you can put the wave guide in and take a defocus image but you can also go to the focus and do a raster scan and look at the far field diffraction for purpose of tomography or scanning socks of fluorescence you can do this on the same specimen those resolution relationships issues that we studied numerically and for for cdi we found benefit for holography but then the group of chris Jacobson compared tomography and holography and found equal dose in in practice in in a numerical model so that that is of course these are important issues but what we what we then do is is and what really facilitates let's say life is to have this full field for a tomographic reconstruction this is a cell which is quite large it's a marine caliomyoside and so with rather quickly with i'm not sure how many angles we used in that case that you know single cell can be it can be reconstructed and it can be analyzed into hondria and and so forth but a single cell we also must keep in mind that a single cell is really something that can be looked at with so many other powerful techniques where x-rays are really unique is a single cell in the entity of the entire organ and that's then really the direction that we would like to move on that we are moving right now and this is also where many questions arise so you saw the single heart cell but how about this is human tissue um how about uh caliomyocytes in um in functional heart tissue like from a biopsy or an autopsy here we were interested in the vasculature and again in changes and pathologies of COVID-19 which is subject of the next talk by Marina so okay uh how do we go from here uh of course the three dimensionality is important the field of view is important but the other end the high resolution is also important and it bothered us that typically we hit the wall just below 50 or 30 nanometers with the standard approach so here's a sign I guess this is um also showing uh referring to the uh a few more minutes so that's fine so how do we how can we how can we use the benefits of holography at the same time increase our resolution well what we uh did most recently we recorded with the waveguide beam but we go to a pixel detector and we don't divide by the empty beam we take the data as it is and we can reconstruct both from the holographic signal and the diffracted signal and this is is really taking up the theme of so-called keyhole cdi but with all the benefit of everything that happened in between in particular holographic probe reconstruction and it's this scheme that some of you may know exploiting a very particular optics namely uh and a plane in the waveguide axis plane where the probe is highly compact we can use this and then we have maximum curvature in the object not maximum but high curvature in the opposite of uh object plane and high compactness in in the source plane and this is is uh an excellent match and here goes uh here goes really my presentation um but I have to finish it um let me see no I can I can't move anymore now Dina did you do this to cut me uh to cut me uh let me let me give it a promise I didn't you didn't you didn't let me try it once more um that I can uh I have to go out of the and and then I go in again so that uh gives you maybe a break um and then I have to start anew and here we are so there's we we use more or less the predetermined uh uh probe as a as a um constraint and we use simply um the Fourier propagator but putting in you know the very specific phase front that we have in the object which is a plane which is very very very very clean in this case we reach 11 nanometer resolution at a few of you in a single shot that is much higher than what you would get from the sampling constraint of the detector of your pixel size 27 the 75 micron for the iga simply because the low spatial frequencies are encoded in the uh holographic signal so that is is is very good we get very good uh single acquisition data it also works for for cells we're trying to put this to use now for for tissues and for this end at petra 4 we want to build an optimized uh instrument we still go for a wavefront filtering to really emancipate ourselves from everything that happens upstream and to have this very compact plane where we can say if we can use a very powerful constraint no stray radiation no orders just uh you know 50 nanometer um um support constraint or something else and we will use pixel detectors but now at uh we go to 25 meters uh to to really uh use pixel detection and to have a reasonable uh few the few yeah because we cannot always go to to close to the waveguide so that's all along sure is stored we must make we are pressed by our own competition if we go to the lab also in the lab here this is work by marina you can see beautifully biopsy reconstruction single cells so uh the synchrotron has has to move also forward and then we can cover all length scales and open up the window of what we are seeing so that is here uh the team i showed you example reconstruction work by mavaeke tepevin who is now working with xylone uh company tomography tomography specialist in industry marios reichert still in the lab as a senior phd student and so is the marina ekamann and many others that are involved in our collaborators um in hamburg and here in the university clinic in particular um thanks thank you for your attention thank you very much team for this overview really dense of information so the session is open for questions and comments we have already won bansal faber nicolam we asked for the keyhole tomography are you still using a near field propagator that that's an excellent question actually we we don't because typically uh the igar is five meters away and our sample is let's say 10 millimeters away from the source and um so yes you could convert to an equivalent parallel beam scheme take the curvature of the input beam away but if we if we uh if we adjust our sampling we can beautifully do it with the Fourier transform so we take we take the detector data we do an inverse Fourier transform we go to the plane of where the sample is and then we use a phase and amplitude of the probe as a constraint and it works it works really beautifully so you use a far field propagator and yep yep yes yes yes it's a far field propagator but we can show in in simulation that the fact that we have a curved a highly curved um wave front is is is really a tremendous benefit it's a tremendous benefit because our field of view in if it was in cdi we could not it would be not properly sampled if we we could not uh you know we if we take the oversampling to turn to five meters and um 75 microns that would give us a field of view i forgot what it is it was three microns i think that was for the i guess it was a little bit larger but we have a factor of two at least that we can uh that we can face stably with this uh curved um uh illumination and we we we think about it and we talk about and speculate about it why this is the case so one interesting fact is that you know the far field uh has this freedel symmetry you know but in the hollow in the holographic cone this is not true so um you actually get less you also get a for for quite a number of pixels you get more information because you don't have this uh you know for the pure fast if i for the pure far field case every second pixel is you know redundant with the with the different one and um yes that's that's that's a scheme i think that is very promising okay thank you thank you i've seen before lots of hands raising and then for some reason they disappear but manuel is there please manuel thanks team for the great talk it's very good to see what you've been up to uh what one question i had was uh if you have tried with you showed this this example that is inspired on this idea of the keyhole etc have you tried to to to measure with a few different positions similar to what they do for the near field dichography in in esrf at the cloutons and here we we we didn't explore that much i mean we we um i think it's it's a very good direction i mean typically you do want to enlarge the field of you literally anyhow so why don't you take a little bit of data there the same for the tomographic scan again you can have very beneficial aspects of you know two projections not being completely independent for the low spatial frequencies so far we really did the bare minimum we um compared uh reconstructions that just have one empty beam that also works nice or take the sample and do a small geographic prescan just to reconstruct the probe and then you do single acquisitions there for all rotations or so so we played a little bit with this but uh not not much much more but um um yeah i think it's really interesting it's interesting also because you're using you end up using pixel detectors yeah with respect to the holographic comparison this is very interesting and we could we can't really use the same um detector for standard holography dividing by the empty beam because our pixel size is too large yeah it's it's really it's from from the holographic point of view it's it's really even if you don't not interested in this super resolution i say yeah i'm not really uh i'm happy with the previous resolution you you you realize that with the fiber coupled ccd or with the uh a syn microscope you have small pixels um but need these small pixels and the the the the the pixel detector doesn't do it but if you if you if you beat this because you have an added uh smaller sampling in the object plane by the high angle that works so that that's also a good advantage and if i can follow up quickly since i don't see i'm taking too much time from others let's put um i was now thinking that you were mentioning it i mean have you done if you if you do tomography with the wave guide by doing this holographic that you remove the probe from the influence of the object etc have you tried to do it in instead of doing every projection in 2d separate the probe from the object and then do tomography have you tried to do it like as a one problem yes yes that is that that is much better that is much better for a number of reasons we we we worked on this already a while ago and we called this irp iterative reprojection algorithm and there are many ways to do it and there are many works but it really is is fantastic it's computationally more expensive but um in particular in propagation imaging the low spatial frequencies are poorly included and this is where you have all you know for your slice theorem all these redundants are very close intersecting so it stabilizes tremendously and of course you can use much better constraints in 3d uh like on better delta than you can do from a projection so it's only and only uh limited by um you know how how good you are in implementing this on a cpu we we seem to do this or we did we we we published something in in 2014 on this first author is uh ike runand and then we decided yeah yeah we can do it for 500 times 500 pixels but it's not sufficient we should look at it again and one do do it serious data science approach and i think it would would fly uh thanks team i was interested in in your comparison between synchrotron sources and laboratory sources i think it's very very very very good point so um i understand that the main reason to go to synchrotron is the resolution that you can achieve yeah also quantitative contrast of course or a wet specimen like if we if the if we um if you have a tissue in in in water not in ethanol but in water not in paraffin or resin and not stained and metallized metallized then uh even if you relax the resolution you ask for a resolution that is possible in the lab you still won't get good results in in in the lab so in in in that case it's your resolution your synchrotron will be the only opportunity but um it's it's shifting all the way and there's there's contrast luckily for compact uh for compact small instruments that is also substantial they detect the the sources the control and that's very good so on both worlds we have we have a lot of progress okay and in terms of optics your choice yeah say that again uh were you continuing or no no i'm i'm done sorry yeah so in terms of optics your choice for the waveguides is really about the quality of the veneer get yeah yeah it's all about the world the quality and also the size it's it's our work is is very driven by the fact that we have unstained a tissue and um so the contrast is very low and it's very of course uh it's biological tissue and you don't want um any damage so we we ask for an imaging scheme that goes with extended samples that covers low contrast at moderate resolution say and that works at kilo gray yeah for for so so that's this is in for that for for that parameter window it's it's it's a quite powerful approach and then also for samples that are really uh hydrated yeah which are difficult in to bring in uh into the vacuum of course if you have a cryo workflow that's again different but large tissues you cannot easily vitrify also by cryogenic techniques okay um i have a question in the chat yes tuna zoo says do you think it is possible to do a biography with nano tube or yeah i mean uh we we do face retriever recordings with the nano tube uh the last image that i showed you from marinas uh publication uh two years ago that was uh using uh um a reconstruction a face retriever reconstruction and then you can also get uh you know let's say three nanometer voxel sizes and you see something that so we we typically don't call it holography because the Fresnel number is not small enough it's just edge enhancement and and the coherence is also um not rigorous enough to say oh this is really how we you know the holographic forward problem so but it's it's um it's still um it can still you can take compromises and you can still see let's say the enhanced density of a cell nucleus with respect to its environment so um to be more holographic maybe if you have a detector that singles out just a smaller bandwidth and uh like we we're trying to work with the men's detector now um and yes maybe we're not so far away from more holographic signals also in the lab but yeah it's it's still still still fundamentally limited i would say okay so are there any other questions for team for this first part of the the micro nano symposium otherwise i would just leave the field to marina team thanks so much and uh we will be here so yeah either don't you serve don't you see don't you serve coffee in between yes it's gonna arrive like in a couple of very very good so i don't know is does anyone want a couple of minutes leg stretch or can we just continue nobody raises the hand so i think it's okay marina i just realized that my laptop made an update so i need to give the permission that i can share the screen so let me see if i really no no i need to do this with my laptops i see okay um and there is a then there is a good minute and a half for for leg stretch for everyone yeah yeah it is ah that's too bad team there's actually a question for this minute yeah that's okay that's good um did you did you think about using like some because you mentioned the money using some kind of sub-pixel registration if you have an this kind of detectors that you that you could break down your pixel size a bit more yeah yeah in with the we mentioned we got from another master we got a prototype and that worked fairly well it was a bit of programming but i think it's very promising it it we had to um uh uh it's anyway the incoming beam though a little bit yeah because it can only work at at low uh flux conditions but i mean any progress in that field of course will be highly and depreciated to which kind of virtual pixel size did you get at which accuracy of your charge cloud determination i think was a factor five with respect to the physical pixel size of 25 microns so so that was already quite good oh that's great cool thank you team i had another sorry yeah i'm just i just want to say i'm sorry for the delay it's okay just let me know when you're ready and you first and i i was wondering about soft x-rays i mean all the work you're shown is with hard x-rays and what can be you know gained using soft x-rays in this field i think for single cell um uh coherent diffractive imaging a holography or geography uh there would be uh quite some interest yeah i mean if we if we started to think about 8 kv and then consider the phase shift of an organelle and the phase i mean of course in a in a tissue and hydrated larger environment you don't have a choice but ideally i think in the tender regime if you would have the right detectors and high numerical aperture on the geometry you would probably get a very good sensitive instrument for 3d reconstructions i mean there's there's of course the classical water window x-ray microscopy that is quite mature and also i think you you see the the the benefits um but then i think uh three four five kv ish two to four kv tender that has been around people talk about it but i don't see a single instrument that can serve it and it could be it could be it could be of interest