Advanced XEOL/XAS Experimental Probe at BM08-XAFS/XRF:...

tax XRF uh illuminating of electronic nature of nano uh nanoposphorus. So um uh thank you very much uh Dr. Latif let me introduce you uh to to the audience. Dr. Latin Yulakan is currently working as a pinland scientist in X-ray absorption fine structure and X-ray fluoresence Finland at a sync uh of synretton like of experimental science and applications in the Middle East Cesan Jordan Dr. K earned his PhD um from the institute of chemistry at University of S Paulo Brazil where he explore explored optical luminous and fundamental magnetic phenomena in reduced dimensional rare earth's nanomaterials. He carried out post-docctoral research at the same institution focusing on novel on novel nanospherus h for solid estates whitelight emitting diodes and before joining cesame he worked at a post-docctoral as a post post-docctoral researcher at the CN PM center national energy material Brazil where he gained considerable experience in single radiation x-ray fluoresence mapping and microscoping imaging So please Dr. Live uh the floor is yours preparing looking forward to hearing your presentation. Do you hear as well? I we are needing um permission to allow the the camera and mic from Dr. I'm trying to activate your microphone. Okay, now we hear you. Okay, fantastic. >> Okay, so I think I will share but uh I think my camera is not on yet. Okay. Is okay. That's fine. Um, >> we we hear >> I think I share my screen. Can you >> Yeah. Can you see it >> now? We can. Yes, it's perfect now. >> Yeah. So, uh, thank you Karolina for the introductions. Uh, my name is Latif Lakhan and I'm working as a beam lines at the BMZ Exaf RF beam line source. Uh I will go to present the both X-ray excited optical lumisence and X-ray absorption spectroscopy experimental probes at the BM exaf beam line. U probably you already have this introduction from multiple of speakers about the sesame. Sesame is one of the synotron light source among the 60 light sources in the world and it is located in uh Jordan near Aban about 20 kilometers. We have operational five uh six beam five beam lines and one have also the branch that we have soft X-ray XPS branch. As far as the BMZ beam line is concerned, it is uh the ESRF beam line. It was rob beam line that is donated to the sesame. So it is the old beam line. We have the um silicon, we have the oxford monochrometer that we upgraded already and it has a two crystal silicon 111. It is the bending magnet beam lines. So we have limited by the flax. As far as the specification of the beam lines, we have uh two mirrors, silicon coating and platinum coating. So we are can probe the metal from 4.7 to 30 kilo electron volt. Uh we have two crystals one silicon 111 and silicon 311. So we can maximum go to the 30 kilo electron volt. As far as the in station is concerned, we have the facilities both for the X-ray absorption spectroscopy, X-ray fluorescent and X-ray exerted optic lumic science that I recently developed in collaboration with the IAE that is funded by international atomic energy. We can measure X-ray absorption in both fluesence and transmission mode. So we have two floor cell detector 64 cells uh that is developed with INF and we have the KC SDG detector. In the sample environment we have the cryjet cooling systems. As far as our beam is bigger we are working in the millimeter size beam. So we are not working with the micr focus beam. I will go through the briefly to the X-ray and the uh matter interaction. If we irradiate the matter with the X-ray multiple phenomena could be happens. First of all could be scattering and this scattering could be coherent elastic if it is in the face that if it is absorbed by the by the matter some X-ray it could be inelastic scattering we can measure the component scatterings as far as the X-ray absorption is concerned we can measure the X-ray absorption either in fluoresence mode because if the metal side is absorbed radiation and that's after recombinations for the core hole it's emit excel Florence we can detect this we can a flloresense mode if there is photoeleron emission is could be also there so we can also move the electron yield to electron yield way and there should be also possibility of the o electrons because the surface electron could be removed by self absorption from the im fluesence emission that happen from the metal site well to briefly introduce all the three techniques in the energy level diagrams that could be X-ray absorption, emission and X-ray emission of X-ray excited after lumisence. If we have metal site and that's weird radiate with the X-ray the core electron could be ejected and that could be like a photo electron waves we can measure this absorptions before and after uh the samples and then we can get this X-ray absorption coefficient. We can get the X-ray absorption spectrums. Anyway, this electron could be stay in any firm levels uh conduction band suppose in any energy levels and that could contribute here to the near edge region. It could be also states in some other fmy levels like 3D orital and that could be also contributes to the pre-edge regions uh that is the fingerprint to know about the summit sides. Similarly there should be core hole re combination could be occur from the near uh shell and that's we can measure the ka k beta on the fluorescent spectrum rather than this this if this materials is flicensed and that have cintilation capability the electron could be stay in the defect surface defects inside the latice and that could be recombined to the valance band and there should be exiton re combination or if there is a 4f energy level from the rear ions that could be also undergo intra configuration transition and then we can measure this optical emission that's occurring from the samples under irradiation with the X-ray we call them X-ray excited optical lumoses in order to introduce the X-ray absorption the first X-ray absorption spectrum was just measured introduced by the drogley just after the invention of the X-ray that was by Rton 1895 and a couple of decades less about two less than two decades. He measured the X-ray absorptions. So he designed a simple spectrometer. He put a crystal on the barometer and then uh he put a photographic palate and then when he get monochromatic X-ray beam from the from rotating the crystal by some theta and then from the X-ray tube that developed by the DRO by the Rton at that time. uh and he gets some blacks uh lines there and then he said that there should be because the photographic plate was irradiating with the X-ray it was glowing like it was fluesence stripes but it was getting absorbed some radiation and that was getting dark and then he said there should is happening some absorption so it was actually the silver broomemide that was the edge of the silver buramine that was getting this black spots black region dark regions As far as the X-ray absorption is concerned, we have two region near H region and extended regions. The near Hit by ICR it's defined that's above the H till 30 electron we consider this near H. And then if we go to the extended regions above 30 we consider the extended regions. So this is the Y square definitions but anyway if we have go the first multiple scing peak is hip earring. If you go behind this, we consider this extended region. So as far as our exual experimental setup is concerned uh it was developed the project grant that was um facilitated donated that was supported by the in 2023. So we have uh we develop in house uh sample environments that is capability to irradiate at the same time with the lasers and with the x-ray and then we have also the lens if we have in the future micro focus we can put here objective lens and then with the camera basel camera so we can also capture the image in the real time rather than this we put here uh collating lens um that's we oceanics we bought from the ocean optics And that was integrated with the uh optical fibers and that's optical fibers are filling the spectrometers that was also from the ocean optics. So we can also if we have the laser and also if you want to cut you can also use a filter. So we have also a multiple here at the end we have also adjustable um uh housings for the filter also. So we can put here the filters and then we can also cut off certain wavelength. So at the same time we can measure the photo limit sense that we can go from the 200 to 920 and at the same time we can also measure the X-ray excited optical sense and also the X-ray absorption spectrum. Uh recently we also worked on the mapping so we can do 2D mappings. So we tried with the with the um X-ray beam but we also also worked with the laser. So we were just moving the sample holders and then uh we were capturing the spectrum at each points and then we were constructing this uh 2D uh map of the the samples. I can show you in the next slide. As far as we needs a different laser so we also inhouse assembled our own laser. We bought a different types of diodes from from different companies and also the lens. So it is focusing by the lens. So it is we can go till less than 100 microns. So we can focus the this laser also there. Uh recently we have this 400 nome laser but we also have some 390 nanometer also we have but we are assembling it. Yeah. Recently we integrate the XCO and we capture a 2D mapping. So it has the capability you can capture the spectrum. Suppose this is the rare earth metals. You can capture the spectrum rather than spectrum. You can also see at the same time the beam by the by the camera if it is after if the material is glowing. So this is a spectrum you can capture and rather than that you can do 2D mapping. Suppose this is a sample and that is with a rare earth sample that is starbium and then you can also we we we make this 10x 10 mm and simple because we don't have micro focus so we we slid down the beam and it was to the minimum size that we could achieve but I don't know it was still in not in 100 of microns but it it is in hundreds of microns maybe millimeters size uh so we are step by step moving the beams uh uh uh sorry the samples and then we also uh map this all samples. So you can see here this is with a turbium some fast force. So you can capture the spectrum at one side and then you can also construct the 2DB uh map. So where is distribution of the rare earth metals there? SRS could be widely explored uh could be widely implied in the rare earths minerals rare earth spectroscopy. So they are rare earth. So I will go through the rare earths principally here there are total 17 element ranging from scandal nutrium from lentanum to luteium and as continue to the name of the rare earths it is much more abandoned than certain metals like silver platinum and gold steel uh the least one is 0.5 ppm it is the thium that is the least abundant there are total 130 million ton matrix reserves worldwide but nowadays the s the ch China is the major contributor contributors of the rare earth minerals to the world and it is reserve is 35%. Bazil is getting recently exploring a huge area metals also it's getting to the second in the to the second in percentage wise earth reserves but these three countries are the third the second that's also sharing the real earth reserves in the world rare earths have wide technological application that have critical roles in the fourth industrial revolution nowadays they have they are implying ranging from magnets to automotive industries to the renewable energies, batteries and to the catalysis and metologicals. These are the certain fast force that's 8.4% 4% that's contributing to the centilation lasers display devices detectors and medical like MRI and also the markers to work with the real they have a huge extensive in intra configuration energy level structure that is arising for the 4F 4 in configuration transition that is extending to the 12 electron volt the lab source is limited by the 5.6 electron volt. If you have very powerful UV source, you can go there till 5.6. So to probe all this energy level, you should have a good synotron facilities equipped with the UV vacuums that you can probe all this metal that you can go to 10 electron volt nearby there. Rather than thus the synotron radiation have high flux, good trainability for high uh brightness and good temporal stability compared to the PS lasers and laboratory source. Here I will give you an example of the fun persistence lumisence. If you want to understand the rare earth persistence lumisence, you need to uh implement a complimentary approach. Suppose you must understand the brand gap or the post lat structure. Suppose you can go get through the XRD UV vacuum bend gate and then you could un you could know the sides the metal sides that you can get from the X-ray absorption and X-ray exited optical lumisence and then you can probe the energy levels what is the oxidation state and from there you can get the four energy level structures from the XC all together with optical spectroscopy and from there you can get this all this in uh energy transfer are intra configuration uh intramolecular energy transfer diagrams from there you can prop this emission that it is happening from the intra configuration for transition of the turbium ion. So the goal to predict the fundamental mechanism of the photo emission occur from the net from the samples or phototer side. You should have complimentary approaches from the aol could be uh play exol together with the exhaust could be play a critical roles to to probe the metal side photometer side get information oxidation state and get on the four energy level structures that can give the correct information from where this emission is occurred. So here we studied one uh examples using the same uh probe. We we prop the European 2 and plus three site using complimentary both approaches the X-ray absorption and X-ray exerted opticumance together with the XRD. XRD give us the longest structure while to probe the local for local structure we get through the uh to the uh X-ray absorption X-ray abion fun structures we fit this and we get from there in order to know it is plus 2 + 3 we also acquire the XCO and from the XO we get European plus2 and plus 3. So the photo emission of the European plus 3 is happen because it is selection root forbidden transition for F4F but if it is occupying a sides inside the crystal latice. So by the mixing of the c uh crystal fields of the lians the selection rule is relaxed and then this allowed the force called force electric dipole transitions. So it's allowed so it's give you this emission line. So the European if the European enter the symmetric sides it's occupying the barerium side. So we from X we get this information that European plus 2 N + 3 that's occupying the barium sides in the latice and from there uh that's emission is occur. So it give the emission if it is not occupying the sides so it not give this emissions although this interconiguration of transition it is allowed but the for transition is per give this emission. So we published these articles recently 2025 that was work done with the Jordan University and journal atomic energy. So from here you can clearly see the uh the capability of a fuel. You can see here this is we we get here because our sample have europeium nitrate a lot. It was European nitrate was not fluesence because it is water quenching a lot. There is water there here that's crystal water there. That's a vibrational level. But if you can see here you can have European plus two here and plus three here. So but you can see here European plus2 in the XCO very well and European plus 3 both energy levels you can see here from the XCO. Second example I will give you about the thermography. In the thermography there is a centilator is used that this is a simple to thermography setup that have the x-ray sars and that specimen that specimen uh putting this after the specimen they put the centilator and then the the lenses and also the uh digital camera simosa etc they put the cameras. So these cintilators have the rare earth are these are used aopium that use a critical role to get this captured images because the optical camera work on the basis of the optical light not on basis of the extra. So centilator have a critical role there. So this centilator that is used in the beats beam line in tomography that is given to me by the bit meline scientist fara and then uh I studied using the vacol I get this I I try to understand and we found that this is ginium gallium garate and that's you see this the emissions of this lenard that's very European is very sensitive to the geometry and interat atomic distances of the lians and the uh the uh the the 40 emitting sides. So here the quint this is the magnetic trans transition dipole transition and this is the hyper sensitive transition that's electric dipole transition that is very sensitive to local geometric side. So the symmetric side is centroymmetric in the case of this europium that's occupying here at the gillium side. So that's why it is suppressing but here you can see here we have distortion here in the latice that's this electortization gaining intensity so it is distorted structures we get this symmetric side is destruction due to slightly inversion of symmetry there this is another example we studied by the same setup I I show here only the x-ray excited optical luminance spectrum but I do not see the exact but we also measured for them and then be published also this example this was organic cintillator that is provided by university of s paulalo so it was actually the simarium it is also that have also this cintillating behavior it was a complex as we have the bending magnets and it is non-focusing beam so it doesn't damage this uh complex although I put in the laser and I damaged by the laser it was one watt laser but it didn't damage it so we capture all the four horror configuration initial line so it As X-ray is a very clean energy so there there is no stray scattering or scattering just like in the UV lights. So it doesn't take it but it capture only the photon that actually coming from the defects or energy levels of the metal sides or any photo emission that from the exit combinations that capture purely this and then you can understand actually the metal sides and also you can get the information on the uh photo the the structures there in correlation the absorption spectrosc there was in other works We also did on the same probab. So they have this photo emission sites emn plus2 and we also found that the re earth have critical role are there because it's it's change the local sites geometries local sides when when rare earth eneral which is with big size is bigger atomic radi so there is distortion in the cycle so in the sides of the manganesees both is in the same sides So there should be also crystal changes happening and that's critically contribute to this enhancement in emissions. So if you are changing the metals with the different ionic radi and different intra configuration energy levels that was transferring energy there and that it was also changing the excitation sorry the intensity of the photo emission. We just published this recently in 2025. You can see there on the one side and the X-ray absorption you can get which what is the sides occupying by the uh rare earths in the Z in the quantum DIS and the manganesees both together and from the XAX excited optical luminesence together with the optical spectroscopy if you can also get a photo there you can get this in configuration uh configuration energy transfer structure and from there you can get how this uh intensity is changing with the rare earth tiles. Yeah. Uh this is another work we just did in 2024. So these are the core shell materials and it has a unique properties. It has both MN plus2 emission and thium is the rare earth. It has a both emissions. So one thium is blue and MN plus MN plus2 is greens. And you can see here that um if you irradiate with the near infrared laser it's 98 nanometer near infrared lasers. So upcon conversion is the low energy this is called stock shifts because it is lower energy it's transferred to the high energy atomic energy levels and then this is undergo a series in the shell energy transfer shifting of the energies occur from the core to the shells in a series and you can see here that we get finally the photo happening to the intra configuration d transition of the 14 emittings mn plus2 sides where is also emission occur from toin So this is we call them long life emission and here is the shutter emissions. So you can see there in the photo if you see the blue emission that's happening here from the uh from the thium if I play here and you can see here we appear a green emission that's happening from the MN plus2. So it is energy migration happening between the core interface and in the shells and that's the MN plus2 gain when you radiates with the near infrared lasers and it is under excit due to the excitation of the high energy levels of the rear that transfer energy the DD excited level of the D intra configuration DD transition of MN plus2 and that's give this phosphoricus long life emission The other we work also in the energy materials that is in the collaborations suppose select fuel cells you know stuff of S Paulo also there are also critical roles there. So they are fused in the cathodes lenthan and also in the electrolyte and also you see here the lenoms. So these are the salt oxide fuel cells that they use. We measure here the ex X-ray absorption excitar and then we found that the anodes uh that was from the nickel oxide was poisoning. We focus the beam here on the sides where there is a cathode and there was a posing of the anode is happening and after the few cycles we saw the uh anode was inactivating and you can see here that is changing from nickel oxide to the uh nickel metals and we we did a principal component analysis and then we did a linear combination feed and we found there that anode is completely changing to the nickel metals. So they are working to improve the university. Yeah, this is the work that's doing by the green gas innovation engineering school University of St. Paul. Yeah, we also work in collaboration with also analysis in electrocalysis that is working with the national center of physics. So they have this electron accelerator and it is uh work on iron beam. They have iron beam. So iron beam is a very clean implantation methods that you can implant in the local sites uh ions there in the latice while rather than very clean because you you have no chance of different phases because there is not a chemical methods but it is very pure but you can do with a specific dose and percentage there. So they have this uh iron beam. So they implant the iron beams in the latice and then we want to see if there is they also enlu and different atmosphere and then want to see if there is s surface defect there. X absorption could be very informative for this and also if there is where is the implantation is happen absorption can also give information. So we do this work in many paper but one we published recently this. So we uh probe the defects inside the iron implantation and then recovering of the defects after the enalings uh in the control atmosphere. We also published one paper. We work on the X-ray absorptions trying to implement some machine learning method. But at that time we studied works together to uh we apply the evolutionary algorithms with the reverse mult simulations to simulate this input different with implantation the kubaltide that was implanted with the nicas. Yeah, this is another work we are also working on the photosis material the quantum dots. So quantum dots have unique behavior. The uh photo emissions is depend on the size. If you change the size, you change the bend gap. Suppose from the if you widen the bendway gap it is blue emission. If you narrow the bend gap it is more red and green emissions. So they this nanop particle was prepared in west of Sal. Palo but they have the same size. they they are co material cadmium narin sulfide. So they have the same size uh under the microscope. So how this blue emission there was they found a blue shift in the emissions with the different uh type of particles changing a little bit the synthesis methods. So what they found we found here that there is an exchange here between the core and the shell. So cmium sulfide is a wide bin gap material. Kumide is a narrow band gap. If you wide B gap is mixed with the narrow band gap, so it's threatening the blue shift. So we did this by quantitative X-ray absorption f structure fit analysis. So the first one you can see then we were mixing up the same latice same locals same crystalographic structure there and then we found that there is a mixing of the sulfur is happening there. Yeah. Recently we are also worked to induce some mapping there because this was the demand from the culture heritage and also theological the archaeology and also the geological survey or the geology samples. So to pro some mappings there so this mapping we induced there this paper we published with together with the bishnim culture heritage. So we mapped this there was artifact there and we map these sides and we found that there is a distribution of the iron uh different metals there and that is responsible for this bluish green color. Well so as far as since uh uh 2020 since the corona time we were also working to implement some machine learning models. So I was looking to apply different models. Uh as far as you knows that the latest generation synotron facilities that produce uh huge data in minimum time uh because uh suppose in the inoperando experiments if you have inoperando experiments you uh in a you you acquire the data in millisecond in microscond and then if you are uh running some experiments you collect a huge number of data to treat all this data the material um uh suppose the software or the product platforms suppose it is um we have used the for pre-processing and normalizing normalizing of the data is etc that have limitation you cannot import a lot of data there so you must be have alternative uh some platform that you can process the data in fastest way and in a huge data in in in efficient to uh to uh to get some results and this is uh slowing the data analysis. Suppose if you are in the beam line and the beam line research scientist suppose the user working there and they get some data. So for the processing that take a long time maybe months to process the data and they go analyze it. So if you have some channel very fast and that should be very fruitful to um while working for several years with the user again this idea it will be much more fruitful to get some result very fast. So we start thinking on this uh and work on this. So then we found that the machine learning could be an efficient way. In the literature there are multiples machine learning methods is available. One is the py fitted that package sold group from southern university Russia and together with Julie from CNRS they have some py fitted package that is python based uh they implement the fdmin also there evolutionary algorithms that also tried and then there is multiple super learning methods that they can get the local structure and acceleration state crystal cluster size etc. The pi fit is focused on the simulation of the near edge regions. So these are the multiple approaches that's already available in the in the literature. The important place is the near h regions that give us a huge informations and here you can get you can probe the density of states and from there you can get the defects there is of frequencies and etc and uh as far as I think there is no very well structured uh some method is available fdm is could be very good but it's very not It's very hard working and it is not very easy to consider to get some result from it. So it's still open. Idopi 50 is also implemented but it is it's very hard. It's not easy to use but it is not working very well for all everything. So there is a still a lot of capacity in this regions to be explored and I think the theoretical physist people could be working these regions very well. Maybe this is it's needed. So if you consider the sense regions in the sense we get uh if we if you if you want to have we have two different type of ages we have the LH and we have the K ages. If the electron is going to some localized final states the behavior if it is going the electro suppose it is allowed electric dipole transitions dipole transition is allowed so it give uh high intensity here the pre the the edge lines white lines speaks uh because it is localized going to the 3D orital if in the case of suppose in the LH because it is allowed dipole transition in the KH go to the 4p 4p is non-loized states and it's give this u this u near age white line peaks broadens in difference because it is highly this is localized state line localized states and much more um distribution is longer so it's a you can you can say the how can we say to the lifetime of the uh coreole core width width of the stage is higher so we can get we can have this widen different isolation near w different FE feature of the white line peaks to consider this. This is also extremely not easy but FDNS is implemented very well. the two different different approaches for this multi- electronic approach or multi-leon field theories and also the DFTs and that consider this very wells uh for this but it is not easy to get uh quantitative information from the age still need much more work on this but how it is very well developed we have multiple uh lectures on it and it is available on YouTube and one can work on it easily and they can get some As far as the selection rules in the this is concerned it in the case of the X-rays in the hard X-ray you work in the above the 4,000 electron volt suppose you work in the above 4.5 kilo electron volt. So in this um um in each type of transition the electro dipole transition is much more important and that is dominant is highly dominant in X-ray absorption. But if it is working in the hard access regions so the quad voltage region can be also efficient and can play a role. So you should also consider this here. But FT M&S uh is already implemented uh the pitted package is both is implemented by default the electric transition but you can manually include the uh quad transitions. So we also implement this py fitted package in the python based uh in on our servers and uh we tried to get some local structure on it um together with the fleminus we implemented. So it is a huge different python scripts. Uh this already paper is published in computer physics communications by Andrea Machini and the Salato group from uh SA University. So we get uh some 3D structures on it. So these are the different we tried on the complexes because the rare earth have um is not easy to crystallize. So uh to have some crystal structure to get some from single crystal from the x-rays. So if we have some model there and we try a different model getting from some DFT simulations and etc. And then you have this um you you you can apply this near edge or some um uh some fitted approach there to get some local structure from there. So these are the complexes. It was different europium, simarium and gadelinium that was uh provided by the University of St. Paul institute of chemistry and we tried to get the near edge measure the near edge spectrum from it the LEGS and then we try to simulate by this pipeted approach that we implement on our servers. We published this article in 2023 in organic chemistry. So these are the complexes that use as a light converting molecular devices. So how it work as a molecular converter device? So it absorb it has the h shifts it [clears throat] because it absorb the higher energy and emit the lower energy. So if you radiate with the US the liant is excite the liant can act as antenna and then the liant undergo interystem crossing to the triplet level and the triplet level is transfer energy to the emitting site emitting level the gate G1 the intrafor configuration exerted state of the suppose in this case is europium but rare earth ions and from there it's emit it's undergo radiative decay to the fundamental forest energy levels and that's give this emission. So it give this red emission. So that use in the huge in the in lot of display devices materials uh detectors etc. there are in this way that's give this emission. So we implement this pi fitted and we simulate the near edge there. The beauty of this py fitted that it's uh you can adjust the uh core hole width and also the excited widths because I show the near edge region is depend on the excited bits there where the uh this um um electron go to the localized state non localized state. So you can adjust this there and then you can adjust this uh um um near. So what we did we we actually uh we get this C files and then we um we did by FDNS and then compare with the FDNS we also did by PI fitted in and PI fitted give a very good result to us. So in order to validate the structures we get from this pip fitted approach we use exaf analysis and we get rid of fit on its and quantitative fits and also it give reasonable result to us. Then we explore these structures of the two selective complexes. And then as in the rare spectroscopy the europium is used as a probe because on this you can probe up easily the local satic sides because the it is less than half field. So if you go here the quantity is zero. So the stock levels and this is separate f0. So, so it is the ST level is very well defined there and from there you can get a very good informations on the symmetry sides because it is multiple symmetries are there one symmetries there and so you can get a very good information there. So we give this to the University of S Paulo and they have developed uh they measure the emission spectrum from this. So there are two you can see can D0 separate F hyper sensitive transition and separate F4 sensitive transition that depends on the local geometry change distortion in the geometries if there is happening happening and plus uh and this one is depend the distance between the uh for the lians and the uh ligan atoms and also the metal centers for emitting centers. So they calculate the intensity parameter for it using the jet parameters. Uh they calculate both uh using the both complexities. So and they use the structure from theory. They also develop uh online joy spectra web platforms uh that take as input file this 3D structure crystalraphic crystalraphic file or you can say this files or XYZ files uh and then uh they can calculate this um applying this data flat par theories to calculate the parameters omega 2 omega 4 omega 2 is for the hypers sensitive transition omega 4 for the sensor transitions and from there we found that the both values were bitted and with the experimental one it was experimental they get from the experimental spectrums from the photo emission spectrum and the bi structure they get from the theories and then from there they compare both it was closely so the local structures was closely the sides was it was low symmetric side was occupying C2 symmetry side so it was similar to the uh to the one that experimental So the pi fit it because I showed you the near edge region is very informative and it's give a very good information about the local geometry and also about the symmetry sides rather than that about the uh there's defects there that is desired vacancies there sometime there it give very good information so it's need a lot of work they are huge and maybe theoretical people will work a lot in this to improve this regions information the other region is the extended region and actually it is much more easy. So as we are not very theoretical physist so working with the user we get some ideas if we have this extended regions and it's behave you can see here it's behave like a photoeleron waves electron as you know if you radiate the metal sides with the x-ray the core electron go and then you can it's look like um uh photoeleron waves that photo electron waves you can see here is spreading things and that could be scattered from the nearest neighbors and that's undergo the interference. You can see here on there constructive and destructive interferation and that's give these photo electron waves. So this is much more easy to implement some model there and then some apply apply some uh machine learning methods there to get a very fast local structures as for as I said before that's we need some alternative methods that because if mostly users are coming from different disciplines and they are not very well trained with the X-ray absorption so they want to use for their um own research. So if you have some platform that can give them easily very very fast some prediction of the structure it will much more informative for them because they use they take a lot of time to process their data and then get some result especially if they are very untrained users. So in this case uh we have we measure this extreme absorptions and then experimentally we can calculate this K very easily. K for this photo electron waves vector. We can calculate very easily from the energy vectors. We can get this K we can vector and then K wave function are the photo electron waves and then exafillation we can get easily. Similarly from the exaf equations that is already implemented by the fif has a very huge work on it. So it is already available platform there. So you can calculate the sky uh initial from the fifth. So you can then you have both and then you have both for you can choose the model because there are two problems to choose the models for prediction of the structure. Number one is the input data because the model need a huge number of input data. how you design the input data and number of seconds um problem choose of the model selection of the model how you properly select the models. So we are trying to explore different models but we found some model that is good work working well but you can if you are very well working with exper with some computer scientists they can also choose some good models they can have idea to uh to get some good predictions. So this is the region the extended regions that we can treat and it is much more easy and we can implement some model to have some fast predictions. Uh thanks to the X-ray large community the smat develop a very good library package. So we have pythons source python language and we have very well packages you have fifth there and also material implement the fifth and also implement the xlr. You can have both I fifth there and fifth there and then you can combine both machine learning there and you can get some good prediction very fast. So neural network neural network u the computer scientists inspired by the brains functions they designed some neural networks and connecting they have different input layer and hidden layers uh and also the output layers. So combining these different types of neuron that work very very well compared to the many platform then that can give you a very some good precise predictions. So as far as I told you to select the input data and also there is a multiple platform available but you need to design input data and also the selection of the models. I will show you here one model. TensorFlow is from the Google. They have implemented a different model and is freely available package that is uh in the Python based packages available that have new convolution neural network. Different types of neural network is available physical for network and different sort of neural networks available. So you can easily use this entry in the model. What we found we use two model we use uh physic informed and CNN model convention that is from the tensorflow was free models we available data requirement you must have for both a good number of data requirement how you process the data and how you prepare the input clean data the problems with the uh x-ray um spectrum that it is noisy usually so to tackle this wise the model is not in prediction is not very well take this nice. So what we did we pre-normalized the data. So you must pre-renormalize the experimental data and also how you prepare the input data to train the model. You prepare the input data and the labels the label should be experimental in data should be your database. How you prepare the database? I will show you some ideas. Maybe you can work on this. I we are also working on it but we still working as in building it. The second yeah the first thing is the data requirement. The sec the much more challenging in this is the generalization because we don't want for the specific we want for the generalized but here we work for the specific and we take a specific metals suppose that is using in catalysis in renewable energies and etc and then we make a database for it. So this is the model that we trained uh both convolutional level network and physics inform network we get uh different materials uh CI files we download from the crystalography database camera that we have access there. So for the difference and we measure experimental spectrum for the same materials and from there we already wrote a Python script for it together with the X-ray large and we built a input database. So you put there you can just give folder there like suppose you for the cobalt you put cobalt oxide cobalt in different zinc zinc oxide zinc chloride etc or lithium cobalt lithium phosphate it's lithium cobalt phosphate or some something like this and then you can put input folders and output folders input folder you put crystalography output folder uh label sorry label you put the experimental data for each one and we wrote a Python script so it automatically pick all the folders and you put their zip files there for for each one and then it's automatically pick it and it also we all apply the X-ray large there in the Python and then it's calculate the theory and it's fit and also by if fit base it minimize the difference and then we get there the theoretical database in this way we produce the theoretical database this kind from there from the crystalraphic structure using this gif and excel both together in the same script then we clean this data which was normalized, we clean it, process it, scale it and from here we measure and from there we use the same um X-ray large and uh we normalize it and from normalization we normalize the this uh experimental spectrum and then experimentally calculate the KK from there and from there we apply on this u neural network using the minimization Adam optimizer for optimizing the learning rate and we put their data loss the difference between experimental uh the label and the input data and we also implement an other loss function that we implement there the fifth together with X-ray large that fifth fit that is fitted and then from there it's taking care the exact loss and equation so it's validated there so we calculate all these losses from theory and from the difference between the database and also the the label. Yeah. So this is the certain model that's we are still working we are developing but um if you can work also there on it and this is one of the platform that we are trying to do XLR. So we first train the model if when we train the model after that we test on unknown data. These are all the input layers dense layers etc. uh we calculate the last functions and also the we we calculate the uh efficiency and everything uh mean and the physics formulas. So then you after that you predict it. So you put the data but this is working on the as I told you with it for the certain database it not to be generalized but it should be generalized but we depend on the database how my database you provide it it will give prediction but if it is not in the database it's not give a good prediction. So we get this on the unknown data test prediction and then from there you can put a C file and from C file each side because it calculate for each side individually if there is metal have two sides it calculate for S. So you can see here uh with the we get both together with the fifth and the prediction both. So the red is the prediction and the blue green is the fifth and we can get this 3D structure. So here you can see here. So we can get from the large fit experimental and then we can predict. So we can get if you have some structure there and you put there you have already structure from the database it pick from there and it predict for you also and then it give all the data with it. So this is a one way but we are still working on it and we are trying to optimize this very well. I have some multiple contexts with some theoretical physics groups. We are working on it still. uh this exa a zol and exas prop uh we just started this since 2024 last year and then we published in this year multiple paper of its u maybe five six paper using this b in different journals. uh the important things that we also working on the local community and uh we develop here uh together with the Jordan atomic energy and also with the Jordan University local communities using this XCO and also working on the rare earth spectroscopy because here we develop some small groups that know we are working on the rare earth spectroscopy that is a very good critical minerals but rare earth mineral that will have good importance ing the fourth industrial revolutions. So we already developed this group here. Thanks to all of you. >> Thank you so much Dr. Lad for a highly detailed uh presentation in how to unlock the unknown using uh XR and XRF. Uh we have two questions in the chat. I deeply apologize we couldn't activate your uh camera but let me make the the two questions and then we have one hand raised. Ahmed is asking are these machine learning driven analysis codes precise and do they produce proper results results of analysis because this uh AA models sometimes can give wrong analysis and he's also asking if how could you measure a lantern luminous without antheina molecule and have you used any organic solvent as antheina? uh which solvent have you used for aobium and turbium? Uh that is that is the question. >> Yeah. So the first part of the question is the machine learning and uh as I also mentioned in my presentation the generalization is still we are facing a huge problem for the generalization and our data is noisy. You know the experimental data if you measure from the X-ray absorption it is noisy data. If you work practically with the machine learning you will get a lot of problems there and the nice the machine learning doesn't the model doesn't predict very well the nice data. So for this what we did we pre-normalized the data we have pre-processing and normalizing of the experimental data and the second point I explained there already that we have a limitation there suppose nowadays we are only putting in the database the very the hard material like renewable energy catalysis like suppose plardadin plat catalytic material platinum pladium etc uh using the peroscar and other sides of the in the solar cells and also in the renewable energy in the batteries cobalt phosphates etc. uh the cathode material. So we are focusing on this to develop develop library. If you have trained already models on this local structure sides that as I explained that how we calculate the local structure from the based on the theory and you put for them then it's give you correct but as I told you if you unknown model if you unknown samples if you do on other samples probably will not give you a correct information but for this what we did we already implemented a fifit and we also give an external and other C file for this for validation. So you put in the directory a C file that you know that my structure will be this and also they are from the database and then you validate on it. So we already although I by the fifth and X-ray large we also have validation on it there. So this is the things we are still working on this and we want to do it how we can make easy regarding to the uh European and the other rare earths [snorts] uh in the case of solution well there is I think one good paper is published on it in the nature I I don't know this but it is published from Aragon national I think they studied the I think the radioactive Prometheium is radioactive. They they uh they studied their sides local structure by X-ray absorption using a solution based. So the rare earth complex unit if you solve it it is stable in the solution. If the molecule and you hit by X-ray there. So as you know to get a good X-ray absorption uh spectrum there it is also challenging because if you have solution based and you hit with the X-ray there's a brownian motions and the molecule can move there. So there it induce nice to the to the spectrum if you acquire it. But if you have good plugs there and then uh you maybe you can cool it the samples and then you can uh in the solution you can get a very good absorption spectrum there and from there from the absorption spectrum if you have X-ray could be a very efficient method. Same in the same way if you get X-ray excited optical or X photoense and you get the emissions lines and at the other side you get the X-ray absorption structure. So it could be a very good to probe the local structure if you have some quantitative approach you apply fitting and then you get this there the structure local and you can also validate it by it should be not europium it should be other but it is photomiss and it is centilating behavior so it can give you can also pro it the sites very easily. So as I told you solution could be challenging but it is possible and there is already multiple paper published on it. You can probe it the local side the structure using both technique at the same time. Thank you very much. with the with the with the liquid because I don't know because with the X-ray because with a laser or with a UV source you can study very well very vacuum you can study but with the X-ray it is with a hard X-ray it is very hard because um you will have lower number of molecule per area and the photon coming UV visible photon coming from the sample will be very low intensity but the how the detector can differentiate from the So it will be a little challenging by Z. >> We received several messages in the chat acknowledging your wonderful presentation. Dr. Latip, congratulations. Uh we can sacrifice our five minutes break if Samine is who is raising hand can keep the question very short and the answer as well please. Okay. I think he put the the question in the chat and Sakaria is asking out of context question anytime estimate when the Turkish X-ray photoelectron spectroscopy beam line will be operational at SEM. >> Yes. So the TPS uh you can say it is already inaugurated in 9th December 2025. So it has a one local source um UV and X-ray both. So it is XPS and UPS both is there. So with the local X-ray and local UV source uh that is working already but with the X-ray they are still optimizing. I think in the next call it will be it will be open for the XPS X-ray photos. this call they didn't open but from the next call because this call is already finished in next call it will be soon I think February in the March it will [clears throat] be open I think >> fantastic thank you so much that gave us couple of minutes for uh very short Break.