00:00:00
Valerie Brogden: I'm Valerie. The emcee of this event, keeping us on schedule, introducing the speakers. I do technical marketing here at Covalent. Here's some housekeeping items. I think I already told everybody, but there's coffee and snacks around the corner. We're going to have lunch at noon. There's bathrooms right down the hall this way. Feel free to come and go at any time. And this can really be as informal as you guys want. Feel free to raise your hands and ask questions if anything's unclear. And you may also notice that you have stickers on your badge. That's going to help us split you up into groups for lab tours and demos at the end of the day. Okay, cool. So before we get into it, I just want to tell you a little bit about Covalent, for those of you who might not know. Let's see. Yep. Here we go. Okay, so Covalent is a service lab that's really built around the idea of making advanced materials analysis easier to access and more useful to researchers. So we position ourselves as a partner to our customers, not just a test house. So that means we focus on really responsive support, practical answers, and helping customers move their R<unk>D forward.
00:01:15
Valerie Brogden: Faster. And another really important part of that story is quality. So we are ISO accredited and we support a really broad set of standardized and self-developed methods. So customers can feel really confident in both the technical rigor and the reliability of the work. And we are a very high growth leader in outsourced analytical services with revenue that just keeps skyrocketing year over year. We're growing pretty quickly. It's awesome. And we built that growth on a really solid foundation. So we have an expert team and a very large and diverse customer base across more than 30 industries. And we have really close partnerships with our instrument manufacturers, including Oxford. And we have an extended network of partner labs. There's Debbie, she supports our Covalent Connect program. So that lets us support really specialized or niche techniques that we might not have in-house. So altogether that means customers get the depth of a really strong in-house team, but plus the reach of a really broad ecosystem. So, of course, one of Covalent's biggest strengths, as you'll see on our tours today, is breadth. So, we support all kinds of stuff, including electron microscopy, X-ray characterization, optical imaging and analysis, particle analysis, material property testing, and analytical chemistry. And that broad capability obviously matters because we all know that real problems.
00:02:52
Valerie Brogden: Barely fit neatly into one single technique. So we can help connect structure to composition to surface features and chemistry to just one complete answer of what's going on in your sample. And when a project requires something that is beyond our in-house tools, that's when we call in our partner lab network and that really helps us function as a one-stop shop for all of your materials characterization needs.
00:03:19
Valerie Brogden: Okay, so with that, I would like to welcome to the podium Will Linthicum. He is an application specialist at Oxford, and he's going to do an introduction to Oxford, of course.
00:03:32
Will Linthicum: Alright, hello everyone. Thank you for the lovely introduction and thank you for hosting us today. So I'd like to introduce Oxford Instruments to the folks in the room. I am a AFM and Raman application specialist here, but I'll be putting on my sales cap for this brief introduction here. Probably keep it brief so we can get straight to the good signs. So, Oxford Instruments is a global company comprised of almost 2,000 employees, hence the name. We are primarily based in the UK, but we have. Several different facilities across 23 different countries, including a local demo office in Pleasanton, California, where myself and Mike Hjelmstad presenting today are based with this type of equipment. We additionally have the equipment of Raman and these electron microscopy detectors at this facility as well. We are servicing both academia, industry, and government sites, and we have over 10,000 customers worldwide across our different product lines. So we have a large portfolio of instrumentation here. A large portion is primarily in the material analysis side, but we also have products in the plasma edging deposition side as well. So of course, electron microscopy and Raman microscopy is what we'll be focusing on today. But keep in mind, we have several other types of microscopes, including my specific background of atomic force microscopes, NMR instrumentation.
00:04:58
Will Linthicum: Nanoindentation. A range of fluorescent microscopes for life science applications. A few different types of high-end cameras and spectrometers. These plasma etching and deposition tools I have mentioned, and a robust imaging software, Imaris, that's the best in class in the field for three-dimensional. So we cover a wide range of market segments: physics, life science, chemistry, materials analysis, all of that. But we're kind of focusing today on the semiconductor application, given the location here. So we're able to address the semiconductor market in several ways. So from your compound semiconductors and MEMS devices to advanced packaging, polymers, all of this we have solutions for. And we do this in a few ways, including scalability, as our systems can either be on early stage RD, kind of a research grade, up to large automated throughput in an ease of use fashion as well. You can continue to build your instruments from the ground up, depending on where you are at your stage of research. We also have focus on precision for very high resolution imaging, depending on the product line, into the prefometer sort of space, including like AFMs, of course, but and nanoscale feature as well.
00:06:19
Will Linthicum: And finally, process integration as far as developing workflows to work on your specific space, make it as automated and easy to use as possible. So at a brief glance, we will be looking at our electron microscope analysis packages today. So we have a range of detectors here, kind of a Christmas tree fashion, depending on what sort of analysis you need to do. So we have a unique methodology called BEX that Mike will get into more detail on, as well as the EDS and EBSD folks might be used to do analysis on your samples on the scanning electron microscope. Primarily elemental analysis as well as crystallinity and orientation. And not included here, we can also put our Raman microscopes directly on the scanning electron microscope as well to do correlative then the Raman microscopes at a glance here. So our Raman microscopes are confocal microscopes that specifically focus on resolution, speed, and sensitivity in order to get you very high-quality results, but also enabling you to do full Raman mapping rather than individual point spectroscopy, enabling you to understand about your sample what material might be where. The confocality also allows you to do this in
00:07:37
Will Linthicum: Dimensions. So you're able to go through your sample and see components under the top substrate. Yep, so that'll keep it brief. So if you want to look up more, feel free to either reach out to any of the Oxford folks in the room or find more on our website or social media pages. We're all very active on LinkedIn. So with that, I would love to hand it to our first speaker today.
00:08:06
Valerie Brogden: Alright, so next up we have Kevin Wu from Covalent. He is our electron microscopy metrology manager and he's going to give us a talk about an introduction to EDS and EBSD. So let's welcome Kevin.
00:08:26
Kevin Wu: Lovely introduction. Good morning, and thank you to Oxford Instruments for co-hosting today. For those of you who I have not met, my name is Kevin Wu. I'm in the electron microscopy group here at Covalent. I work primarily on projects requiring scanning electron microscopy, energy dispersal spectroscopy. Focus on the milling. Brought ion B milling and electron back scattered diffraction. So over the next half hour or so, I'll walk you through the fundamentals of two of the detector technologies that you will be hearing about a lot today. The first being energy dispersal X ray spectroscopy, or EDS. Second being electron back scatter refraction. EBSD. My goal here today isn't to make you an expert, but it's to give you a mental model for how these detectors work and how what they're actually measuring. So when Mike from Oxford shows you specialized applications right after this presentation, the why. Behind them makes more sense. We'll cover how each detector works, what kind of data it produces. And a couple of real-world examples from projects that we've run here at Covalent. If you have any questions for me during the presentation, please don't hesitate to interrupt me and we'll try to address them as we can.
00:09:46
Kevin Wu: Okay, so here's a quick road map before we dive into the meat of this presentation. We'll spend we'll spend the first half on EDS. Which will cover the elemental analysis. And then we'll move on to EBSD, which covers crystallographic. Now I'll hand things off to Mike from Oxford. To go over more specialized applications right after this presentation. And if you wanted to save your questions until after this presentation, then that's we'll have time for addressing those as the or you can come find me during lunch and we can we can chat. Okay, so. Let's start with EDS, energy dispersive X-ray spectroscopy. This is probably the technique that most of you are already aware of. It's the most common add-on that we have. Most SEM systems and it's usually the first thing that people think of when they want to know what their sample is made of. Alright, so what's up with the detector? Every EDS measurement starts with the same physics. So if you describes with the four model of the atom and the core shell model of the atom.
00:11:02
Kevin Wu: The primary electrons from the SEM's beam. Interact with your sample. And those primary electrons are energetic enough to knock out an electron from an inner shell. Of the atom. In this schematic that we have here. The primary atom sorry, the primary electron has knocked out a K shell electron from the sample. And this leaves an electron vacancy. And sets. Electrons want to be as energetically stable as possible. An outer shell electron, or more outer shell. This vacancy is like, ah, yes, excellent. I want to be more energetically stable. So I'm going to jump down into that K shell of like. And so it gives off an photon, a discrete amount of energy. The energy band gap between the K shell and the L shell. So the this characteristic characteristic x-ray energy. Then floats out into the ether and is detected by the EDS detector. So since shell energies are set by the atom's structure, the X-ray's energy is
00:12:20
Kevin Wu: What we could call a characteristic of the element. Of that specific shell transition. You'll notice that you have lots of shells of different shell transitions. Some shell transitions happen more often than others. And then we can see from Moseley's law that the X-ray energy detected scales predictably in the with the actual atomic number of the elements that we're detecting. So that's the entire basis of EDS. Every element above a specific light element cutoff, which we'll go into more detail later. Produces its own fingerprint and char characteristic X-ray energy. And so when you measure those energies, it will tell you what's actually in the sample. Okay. Yeah. Almost every modern SEM EDS system uses what is called a silicon drift detector or SDD for short. Catch these x-rays. When a characteristic X ray is absorbed in the silicon, it creates a small cloud of electronic pairs. Critically, the size of that cloud is proportional to the x-ray's energy. So the set a set of concentric ring electrodes creates an electric field that drifts those
00:13:41
Kevin Wu: Electrons across the silicon drift detector chip to a very small collection anode. Because of the the size of the anode is quite tiny, its capacitance capacitance is tiny as well. So each arriving cloud produces a very clean voltage step. Whose height encodes the X ray's energy. Electronics sorts thousands of these pulses per second by the height of the pulse into a spectrum that you'll see on the next slide. So an example of where this technology has evolved is in Oxford's Ultimate Max Detector. Why? They come in a range of sensory areas from 40 square millimeters up to 170 square millimeters. Oxford reports that the same energy resolution across the whole range of sizes here. Gives you like roughly one hundred and twenty electron volts. Or the main maintenance chapel. So these bigger sensor options go up to 100. Millimeters square. Critical because you get a very sharp drop in
00:14:56
Kevin Wu: When you drop the beam holes. To to enable better spatial resolution. Your X-ray output. Drops off s severely. So you need a larger solid angle. To capture those x-rays and get a useful signal. Works well for nanoscale work. Example. This is what it all turns into, an energy spectrum. On the y-axis here we have x-ray counts. And on the x-axis we have the x-ray energy measured in kiloelectron volts. Note that this particular plot is just illustrative. It's been generated to show you the shape of a X-ray spectrum. But it's not actually measured data. It's built to resemble a compound semiconductor. In this particular example, it's gallium nitride on silicon. Material that some of you may be familiar with. The sharp peaks on the spectrum show the characteristic x-rays that we've been talking about.
00:16:12
Kevin Wu: The characteristics that we see here include gallium, nitrogen, silicon, oxygen, carbon. The eighth position tells you the element is, that's the p position is. X free energy. Underneath this the black lines, you can sort of see a kind of light gray area. Which A smoothly decaying background, which is described as the Bremschulung or continuum radiation. This comes from electrons of the primary being generally decelerating within the sample and it's not specifically it's not specific to any particular element. You would typically consider that noise that you subtract out when you're doing calculation. So when you consider the peak height, the area and when you correct through the matrix, you'll result in what's called the chemical constant qualification. And that is usually displayed in weight percent or a con percent for each element that is being measured.
00:17:30
Kevin Wu: In practice you'll use EDS in one of three modes. The first being point analysis. Park your beam on a single spot on your sample and you collect a full spectrum from that point. So you get the composition of that particular point in space. It's useful for identifying an unknown. Such like editing. Inclusion or cortical. Point analysis can be expanded to encompass a defined two year. The beam is ratcheted across a specific area, whether it's like a square or rectangle, circle, something that you draw 2D with a mouse. On your sample. The beam is drafted across that area and the spectrum is collected of that entire area that was defined. The next level is a line skate. So you step the gr the electron beam. I frosted. A line and flick the spectrum at every step along that line. And then you plot the intensity of the chosen elements versus the position along that line. And the result is useful for looking at, say, like a diffusion profile.
00:18:49
Kevin Wu: Check how strip and interfaces. And generally you can sort of use that as a way to identify if you have a thin film stack. Where the different elements lie along same the sample things. And then the third is an elemental map. So you collect a spectrum at every pixel across. A 2D area that you could find. And then you color code the result based on each element. For each pixel on that map. Elemental maps are the most commonly requested EDS deliverable. Because it turns the composition into an image that anyone can read it at once. Alright, so a few practical considerations here. There are a few practical limits worth knowing when we're talking about EDS. Spatial resolution is set by the interaction volume of the electron beam, not necessarily just the beam size. So X rays are actually generated throughout a teardrop shaped volume under the surface. So even if your beam spot size is only a few nanometers.
00:20:07
Kevin Wu: Your effective EDS resolution at typical SEM voltage is somewhere saying about fifteen, twenty kV or so. Is usually on the order of one to two microns. When you drop the accelerating voltage, that interaction volume shrinks, so you improve your spatial resolution. This is why most nanoscale EDS work is often done at five kV as opposed to twenty kV. There is a trade-off though because as I mentioned before, you pay for it with a steep drop in the X-ray yield. Why the larger silicon drift detector? Checkers. Exists. And you also run into a potential limitation of not exciting the higher energy equivalence rays that may be necessary to accurately identify what elements are present in your sample. So when you have a more complicated material system, the lower energy X-rays for those elements may overlap a bit or it may be a little more difficult. Deconvolute all of these different gates. So one potential compromise here is to run
00:21:23
Kevin Wu: Two different analyses. One with a higher beam energy in order to excite all of the relevant peaks in your material system. And then you go back with a lower K V beam to improve your spatial resolution. And knowing what elements exist based off the high energy analysis, you can then self-select. What elements you're actually looking for? Energy. So there's also a practical detection limit for EDS. The typical detection limit for routine EDS work is about 0.1 weight percent. Roughly one thousand parts per million.
00:22:12
Kevin Wu: I guess. And there's also a light element floor. So hydrogen and helium don't produce charis characteristic X rays at all. And lithium is typically the lithium K alpha x-ray. It's typically too low of an energy to reliably detect. Without advanced windowless detectors. And then for elements of beryllium through fluorine, you typically need a thin window or windowless detector. That are more sensitive to chemical bonding shifts and self-adsorption to be a lot more accurate in content for those elements. Okay, so let's transition to some actual applications of EDS on actual samples. One of the many industries that Covalent supports is the battery industry. Most common type of analysis we do. Battery and electrode cross sections. In many such cases, we're looking to see the uniformity and elemental distribution. Of all the primary materials, the presence of the carbonaceous finer material. And the electrode cake as well as any contaminants that we don't expect to be there.
00:23:33
Kevin Wu: This particular battery cathode was repaired at a cross-section with a broad ion beam millington. The broad ion beam can produce a cross-section on the order of about a millimeter wide. And produce a relatively flat cross-section surface. Yeah. Might see in the secondary electron image there's still some curtaining. Associated with milling through a lot of voids. But in the backscatter image, all that kind of disappears. So we still get a pretty good signal for the set of EDS elemental maps here shows us that the primary constituents are oxygen, iron, phosphorus and carbon. And it looks like there are trace amounts of titanium in cake as well. So given the presence of iron and phosphorus and oxygen, we would expect this cathode battery chemistry to be lithium iron phosphate. You'll notice that we don't have lithium here in the EDS map. That's precisely because our detector that we're using is not a windowless detector, and so all lithium x-ray energies are too low. And are absorbed by the w detector window itself.
00:24:50
Kevin Wu: Fact that we have value cathode, we assume a lithium was there. Okay, so here's another example of EDS used to determine elements present on a deposited gallum nitride fill stack. A focus ion beam was used to cross section through the thin film stack and expose all of the layers present. To mitigate surface damage at the plane of interest from ion beam damage during milling, a protective cap of platinum was deposited. Prior to the milling step. From top to bottom, we see an L E D S elemental map depicting the organometallic plant and cap. Liam and carbon. Followed by the gallium nitride. Then what appears to be an indium tin oxide layer. Silicon oxide and then the silicon wafer substrate. One thing to note here is that according to the individual maps for iridium and tin, we see some X-ray signal coming from the platinum cap. Time away. So these X-ray signals are almost certainly originating from the background for empty longing.
00:26:05
Kevin Wu: That I was mentioning earlier. The X-ray peaks for both indium and tin are relatively short compared to the actual background signal. So on this particular map. It's plotting both the real signal as well as the X ray. Revisal on the background. On the l on the right here we have an EDS line scan. The EDS line scan graph overlays all of the elements onto a single chart. You can see significant spikes and drops at the layer interfaces. But you'll notice that it's not a strictly step function. The intermixing of X-ray signals here of the elements around the interfaces is influenced significantly by the interaction volume. That I was describing earlier from the primary electron beam. Reflecting X-rays from behind the cross-section surface plane and around well so that's why layers would appear thicker if you were up to a top two. Measure layer and this is based off the ES signal. Compared to the actual SEM. Image itself. Okay, so let's switch gears to the EBSD.
00:27:29
Kevin Wu: Where EDS tells you what elements are present. EBSD tells you about the crystal structure and orientation. Of your materials. It's important to stress that EBSD is a technique for crystalline samples. That includes metal alloys, ceramics, geological materials. What kinds of information can we learn from electron back scatter diffraction? Well, EBSD can teach us a lot about a sample. Learn much about the crystallographic orientation of the sample, which includes grain size and distribution. Brain boundary characteristics, the crystallographic texture. Which is used to describe the preferred grain orientation. And that typically arises from the forming and shaping processes of metals. EBSD allows us to determine the distribution of material phase in multi-phase samples. This does require us to know which phases are actually present ahead of time in order to index the data correctly. Or combined with other chemical composition analysis techniques such as EDS to help identify unknown bases. EBSD also allows us to visualize localized deformation and strain within grains.
00:28:50
Kevin Wu: As seen in intragrain misorientation max. Botless. In addition, the four scattered electrons can be detected and imaged, providing enhanced phase orientation contrast. So let me go a little bit more detail each of these outputs that you see here on the screen. On the top left, we have a grain orientation map that color codes the crystallographic orientations in the area of interest. Relative to the directions described by an inverse pole figure. In this particular sample, the inverse pole figure map. Red is associated with the zero zero one family of planes, blue the one one one family of planes, and green the one zero one family planes. We can see if you squint. Real hard. Within each frame I've added a unit cell Q. Whose orientation accurately describes the orientation of the grain. I find this to be a lot more useful than trying to see, oh, this color of blue is slightly more green than that one, so it's slightly more rotated this way.
00:30:08
Kevin Wu: And in the top middle plot here, we have the Euler Angle Orientation Map, which is just another way to interpret the grain orientation. By color coding based on the three boiler angle rotations required to arrive at the final crystal grain orientation. Given the fixed order. So I believe it's rotation around C and then rotation around X and then rotation about the new C. Axis. Quite complicated. Okay, and then the image on the top right is a four-scattered electron image. Due to the large seven degree tilt of the sample, a majority of the electrons that are scattered are scattered towards the bottom of your SEM chamber. So on EBSD tools. Install or scatter electron detectors around the phosphorus screen. So these the four scattered electron signal can be used to display orientation slash channel and contrast. And where the grayscale values of your image correspond to different crystal grains. And then four scattered electrons can also are also highly sensitive to surface features such as scratches, cracks, and height variations, and can be used to view topographic control.
00:31:36
Kevin Wu: This particular four-scatter image does display topographic contrast of our sample surface. And then moving on to the bottom left is the missile orientation map, which essentially displays a heat map. That quantifies the vocalized misorientation and lattice distortions in the crystal structure. It compares the misorientation angle between a central pixel and all of its immediate neighbors. The higher the average misorientation, the higher the local adapting and distortion caused by things such as processing and mechanical load. In the bottom center maps are both the pole figure, which is the stereoscra stereoscopic projection circular on the top there. And the inverse pole figure map. The whole figure maps. The orientation of the measured crystal planes relative to the macroscopic sample axis. It is useful to analyze the overall preferred orientation of the area of interest. The inverse pole figure maps a specific macroscopic sample axis relative to the internal system coordinates.
00:32:55
Kevin Wu: It's commonly used to color code the grain orientation maps as seen in the Okay, alright, so. Setup and sorry the fraction pattern that we see with EBSD. Like X-ray diffraction. Electron back scatter diffraction relies on the constructive constructive interference of elastically scattered radiation. Electrons in this case, to form patterns on a phosphor screen. These patterns are known as Kikuchi patterns, named after Seishi Kikuchi. And these patterns can be indexed. A fraction of the instant electrons that are scattered by the crystalline material satisfy Bragg's law. Where lambda is the wavelength of the electron, d is the spacing of the diffracting planes of atoms, and theta is the angle of incidence from the diffracting plane. These diffracted electrons then form a set of paired large angle cones that correspond to each diffracting plane and land on the electron. Great. Interpretation of Pikachu patterns and how they relate to material phase and orientation relies heavily on knowledge of crystallography and material science.
00:34:15
Kevin Wu: Which admittedly is a little difficult to understand, even for me. But luckily interpretation of the pattern data is handled automatically with today's software. To use a simplified description, the intersection of Kakuchi bands correspond to zone axes in a crystal, the common directions shared by two D crystal planes when they intersect. Position and orientation of the projected Kakuchi bands and intersections are key when indexing patterns and determining the orientation of the crystal plane. That's currently being analyzed. The diagram on the bottom left provides a visual example of the gnomic projection of the fracture bands on the phosphor screen with respect to a few common planes. Cubic crystals. Doubtful. One oh oh 110 and 111. You may s see sort of seeing these projections on the previous slide. On the the IPF grain orientation map that we had in the top left.
00:35:34
Kevin Wu: Okay, so. There are several key considerations before performing an EBSD experiment. The surface of interest must be extremely flat and well polished. For an alloy sample, for example, this typically involves encapsulating the sample in epoxy and mechanically polishing. The service of interest. Standard polishing procedures may still leave residual stresses that were induced during the said polishing steps. And would not accurately reflect the as received sampling. In the ideal case, the sample would then be prepared with a vibratory column. Final step to remove any stresses induced during mechanical polishing steps. As an alternative, a mechanically polished sample can be polished with a low ion a low energy ion beam to remove the surface level material that was in where deformation was already induced. This can be performed on a broad ion beam cross-section polishing colour. Like the one that we have. Actually the two, like three that are capping house. Bulk EBSD experiments typically require a beam instance of 20 degrees to the surface, which translates to a 70 degree tilt relative.
00:36:54
Kevin Wu: Wars on plane. Such an extreme angle limits the physical dimensions of samples that we can fit into an SEM chamber. Unless we run into the risk of hitting the whole piece or the EBSD detector itself. In order to obtain enough signal to form an indexable Cakuchi pattern, bulk EBSD typically uses an accelerating voltage of 20 KeV and beam currents above three nano amps. Such a high beam current is necessary. So sample conductivity is also a critical factor to consider. Steps must be taken to enhance sample conductivity and reduce sample drift. Because any sample drift present during an EBSD data collection will result in an inaccurate representation of the material in question. The beam spot size, or beam current, and accelerating voltage are the two biggest influences on the spatial resolution of bulk EBSD. A lower voltage and current will improve the resolution to tens of meters nanometers laterally. Though the signal to noise ratio typically suffers as a result. And dramatically increases the time required for data collection.
00:38:11
Kevin Wu: Under standard electron beam conditions, it becomes difficult to accurately resolve crystal and thin films. Where grain sizes can be below 100 ampere. Can be done, it's just a little bit. Okay, so now let's discuss a few practical applications. One relatively straightforward type of EBSD analysis we have completed regularly at Covalent. Has been on copper foils. However, foils are intended to be used as a key component in battery manufacturing, and it's important to understand how the foil pouring processes impact grain size distribution. The preferred crystallographic orientation and a few other brain structure properties. The copper foils have been prepared in cross-section to be able to brought iron view from milling. Of a brought ion beam can produce relatively flat cross-section surfaces free from induced mechanical stresses. In order to accelerate data acquisition for such a large quantity of samples, a wide area of the foil is analyzed, but with a relatively coarse fixing size. So on this particular example, about a 430 micron wide area was measured in EBSD.
00:39:28
Kevin Wu: Using a pixel or step size of about five hundred nanometers. A sub size must be sufficiently small in order to detect changes in the crystallographic orientation associated with grain boundary. In this particular sample, we can see that the area-weighted mean grade size is on the order of 230 square microns. If you count twinned regions as separate grains. This does not appear to have a significant amount of localized strain across the cross-section, as indicated by the misorientation map. Second mouth. Huh. And the pole figures and inverse pole figures appear to indicate that there's some preference towards the one one one family of directions. When looking from the z direction. Okay, and my second EBSD example circles back to the gallium nitride film sample that we saw earlier. Yes. So regions of this particular sample exhibited bump-like defects and pits. That can be seen from the top down SEM image here.
00:40:44
Kevin Wu: On the bottom left. EBSD was conducted to characterize the grain orientation and the regions of lowest grain within this gallium nitride film. So for this analysis a forty by thirty micron area was measured. With EBSD using a pixel step size of 100 nanometers. And we can see that the entire region that was analyzed is oriented along the hexagonal crystal structure C axis, which is the zero zero zero one direction. As indicated by the gray orientation map, the pole figure, and inverse pole figure. Actually a little warm. All one colour. When we look at the kernel average disorientation map, we see zero misorientation along all of the flat roots but we see an increased amount of disorientation and strain associated with bulk defects and bits. There appears to be increased strain in the lattice between the flat regions of the abuffs to account for the perhaps dislocation concentration and the height. This magic here.
00:41:59
Kevin Wu: And that is the fundamentals of EDS and EBSD. Mike from Oxford is going to build directly on this foundation and show you what these same two detector families, as well as Oxford's BEX detector, combines backscattered electron imaging and X-ray signals into a single fast measurement. What those can do for specialized applications. I can take a few questions now if you have anything in particular you would like to ask about, or I can hand it over to Mike.
00:42:41
Kevin Wu: When you print the it does for PMA, there is no alteration. Direction then we need to put that in the direction when you see that the nomine direction or this is the trend boss in the direction, something like that, you need even you fully think was perfect. That's why I just wanted to ask about those kind of perfect nominee direction. Right, yeah. So when you start collecting data you do have to tell the system. Here, this direction is the normal direction, this may be the transverse direction, this is the rolling direction. Which may be relevant when you're dealing with formed materials that go through going processes and that sort of thing. So yes, you do need to define that ahead of time. And when it displays the IPF yeah. The directions that it describes can relate to the normal direction transverse. Etcetera that you describe at the start of a project. I have some questions. Yes. So you were talking about how you can see these lighter elements with EDS when you have no window. So I'm curious: like, what does having a window gain you? Like, why have a window at all?
00:44:09
Kevin Wu: Right. So the window itself is primarily as a protection feature, I I'm pretty sure. So most windows I believe are rillium. Not Berlin? Okay, that's that's my old college education talking. Anyway. So the Windows our my understanding is primarily to protect the actual protector of the bank. From possible damage from samples like particles flying up into the detector. And also helps cut down some of the low energy background noise. Bully? That's that's why I think that's a just to just to build on that for Kevin, it's it's it's cleanliness as well. So not just like protecting the sensor from damage, but keeping everything at all dry vacuum behind it. Because EDS detectors they go on all kinds of sets, they don't just go on the highest end things, you know, they'll go on.
00:45:27
Kevin Wu: Might not be a post-code as vacuum you've ever seen. That would be a better example so they came in and technically. Mm-hmm. Well, you lose the vacuum behind it. So, you know, usually at the factory they're pumped down ultra high vacuum 10 to the negative 10 type tour. You know, when you're working with a real high end thing, maybe you gave a negative seven, maybe a negative eighth. You're actually working in a worse environment coming up with being window. But we can I mean, I I can talk about lithium all day too. I know Thomas is going to talk about it from a Raman standpoint, but EDS with lithium is very difficult in the sense of creating bands right in the beginning. So do you oxidize your lithium? Probably not something you're going to see from the next one. That's one of the advantages of having the Raman on the electron microscope is you can do correlative work where you can't see the lithium signal on the Raman side of it and then do all the rest of the EDS and EBSD work. On the same system. Idea to pop a window up, but I've not getting to it. It's a tricky part too. What about like a fun?
00:46:44
Kevin Wu: Maybe. Do we have any other?
00:47:00
Kevin Wu: I couldn't see but so in the middle or the picture. It does have the green or where we see SEM images. So, does that just mean that they have a different orientation than? So the map that we're looking at here is the integrated disorientation map, the strain map. It might be a little difficult to see here, but this is basically a heat network. Blue is close to zero and red in this case is about five degrees or so. So the green here is basically just showing us that we have A enhanced amount of misorientation, which we ascribe to like strain in the system. So those the strain that we're seeing is a thing associated with the kind of hill like so there's literal hills growing out of our flat plain. There's going to be some strain associated with. That growth into Christmas.
00:48:16
Kevin Wu: So the one on the right is our grave orientation. So if you look on the corner there, we have a color coding based off of the crystal orientation of the entire system. So we're dealing with a hexagonal. Crystal system here. But suffice to say, red corresponds to the z central axis looking up. Math is you imagine a hexagonal. Cylinder. And this little unit cell figure here shows you what face we're looking at. When we're looking at this crystal system. So the fact that it's all red. Means that presumably it's all a single crystal. Based on the way that this the gallon nitride system was grown. And it's all oriented relative to
00:49:30
Kevin Wu: This is really interesting because if you look at that this is not centered. That's why it's limited. But if you look at the most original black dark means is the Quiche pattern is hazed because it is not crystallized in a detect. That's why this orientation showing here some separate color of a grain, and that's why you don't new two to get the other side this is some artifact. And that's likewise is very great to for understanding this orientation. That's why in the micro tension it's it's not hope vector aligned to the under your product. Yes. Just slightly. Slightly misaligned. And I'd look at the center graph as more of a gradient under orientation. So how fast it's changing in one? Any more questions? I have one more. Yes. Sorry. Okay, so I know backscatter electrons hit the nucleus, come right back. Secondary electrons get absorbed into the electron cloud, different one ejected. What is a force scatter electron? So force scatter electron in this case is sort of like a secondary electron, in which it does kind of interact with the
00:50:47
Kevin Wu: Actually, no. I guess it would be more akin to backscan electronic. So instead of going. Directly back towards the beam direction because of the angle the sample is kind of skipping forward. Sort of in the center of change in the same direction as all the electrons that are and so yes, because of the extreme angle it's scattering forward. Towards the bottom of the chamber. And we can still collect those backscatter electrons to form an image. That makes sense. Yeah. It's it's a primary beam electron that goes down. Okay. Cool. Any other questions? Okay, let's see first two.
00:51:41
Valerie Brogden: So let's welcome to the podium Mike Hjelmstad. He is an application specialist at Oxford Instruments, and he's going to give us some special applications for EDS and EBSD.
00:51:48
Mike Hjelmstad: Yeah, thanks for coming everybody today. My name is Mike Hjelmstad. I'm the senior application specialist for the EM products at Oxford. Glad to see everybody today and I wanted to thank Covalent for you know putting on the workshop. Appreciate the invite out. I gotta say it's been a little while since I've been here. I kind of make it here in stages. And every single time they show up, it's it's more equipment, newer building. I feel like when I come back in like three months, you're going to have another building that you've taken over, another 100 employees or something like that. Keep up the good work is all I could say. But yeah, I mean, I appreciate Kevin's presentation. He did a nice job of laying the groundwork for EDS and EBSD. I can build off of that. And I wanted to focus on. I've heard there's an industry in semiconductor around here maybe. So I thought that would be a little bit more relevant to the folks here. So I want to kind of talk about some of the challenges, some of the more key applications we've found. What I really want to do is break it down my my three headers on the left. Those are three different detectors that we're going to really talk about. We're going to talk about EDS and elemental characterization. And then I wanted to kind of introduce this BEX that's kind of getting a little bit of a hype train going today. But the idea of full sample characterization. Then we'll talk about the microstructure characterization, build off of what we've talked about with EBSD. And just on the right side, some of the
00:53:19
Mike Hjelmstad: Materials, some of the parts that we're going to kind of go through in the presentation here. When I think about semiconductor, when I think about microanalysis with semiconductor, I think resolution, resolution, resolution. It's small stuff. I mean, it's it's stuff that we can barely see with the electron microscope as is. We have a bigger challenge with X-ray analysis because that interaction volume that we saw Kevin shows. The X-ray volume's a little bit bigger than the electron portion of it. So it's tougher for us to get kind of the resolution that you might from an electron image. This is kind of modeling like some of the key data that we've been able to achieve with this here. And I kind of always show that one of the problems that we have in EDS is sometimes you can have multiple key elements, this is indicating copper, hafnium, tannum. They can all be in kind of underlying the same peak in EDS. And sometimes separating it out can be a trick. So like if we look at like a regular map of hafthium in this vinfet part, we kind of see that these via's getting mapped to S. Hafnium and he kind of I thought it was the copper that was part of that, or I thought it was something
00:54:36
Mike Hjelmstad: Yeah, other material, but what we do now in mapping is we we create an algorithm that it's been out for, gosh, 10, maybe 15 years at this point. But what we do is at every pixel we really remove background, that Bremström portion of the spectrum, and we do peak deconvolution. So we look at the shapes of this and go, okay, this is kind of at this point, it's slanted more towards the copper. This might actually be copper, not hafnium. The the liner here is actually the hafnium. And when we do the TrueMap we get something that looks like this. And when you look at the size of that liner in this FinPed, this is 3.5 nanometers. That we scale this up and measure this at. The other thing that we've been able to do with the TrueMap side of this is we've been able to map arsenic and actually see a little bit of a dopant profile in this sample. No. Well one thing I said is this represents the cream of the crowd. What I didn't tell you right off the bat is take it to the T yeah. So, we do do TEM detectors for X-ray analysis. It kind of represents the high end of what's possible. I really wanted to kind of focus on STEM a little bit more because TEM's great, but everybody I meet goes,
00:55:52
Mike Hjelmstad: Is there a way to do this in the SEM? How can I do it? Because basically, time-to-analysis comparison here: if I look at it, that's TEM on the top, SEM on the bottom. You know you can take these numbers to heart, you know. Somebody in here might go, when I work over on the JEOL TEM, I can knock it down to 30 minutes. Cool, I bet we could knock down the sem time o as well. Kind of point is seven hours this should be a quicker turnaround time. It is a question of can you achieve what we just saw on the previous page? You know, when I think about different kind of modes of operation the set. The one on the left is what we've used for years for EDS. So just high kV, crank it up, you know. 10 to 15 millimeter working distance. And really the the cost of resolution is what that probe size is. Generally, you'd say, okay, you could probably get to. Maybe a sack. 50 nanometer kind of resolution for I'm talking about for EDS map. You know, and I'm kind of being generic on what element I'm talking up to. It doesn't necessarily mean silicon versus boron or something like that. That's all fine, great. That's not getting the job done for
00:57:08
Mike Hjelmstad: Or something. So the other options are the two on the right. And the one that I'm kind of interested in more is located. Kevin did the presentation, he said, yeah, when we do this nano analysis we gotta get down to five k V. And even 5 kV historically there's a lot of people going I never did EDS at five K meeting. That's gonna be tricky. I'm gonna show you the detector that I'd like to do even better. I'd like to do this at like 3 kV. I'd like to do it at maybe 1 or 2 kV if it really presents itself. When I do something at that low at accelerating voltage, I'm shrinking that interaction volume and I'm kind of getting down to maybe ten nanometer resolution in a seven. And then the last one is to effectively simulate your TEM. You know, why not make a lamello? You know, do we really need 200 kV accelerating voltage? Can I use chances are you can get resolution down into the 5 nanometer range. That's usually electron imaging and you can probably do about the same with the X-ray map at that point, just because you almost have no interaction volume. So coming back a little bit on the locating. This is an NN flash device here. But what it's kind of showing in this scale kind of going down here is kind of identifying three different detectors here. And Kevin did a great job of kind of saying, you know, Oxford, we make these different sized detectors.
00:58:37
Mike Hjelmstad: Because when we do okay the analysis just not going to get a ton of x-rays out of it. I'm searching for resolution, so when I go low kV. If I crank my beam current back up, all of a sudden I make my probe bigger again, and I kind of defeated my purpose here. What this is trying to show is kind of a common beam current that's being used. And to get some data with a 65 millimeter detector. You know, we're running at seven K V and we're starting to see some of the functionality here. But if we have a larger detector. We can use the same beam current brought at a lower accelerating voltage and still get even better resolutions. Maybe arguable it's getting a little bit better. And at at one hundred seventy we can start to run in the sub five kV range. And now we're really starting to see structure, especially on the silicon, we're starting to see structure on the nitrogen here. Another kind of good example to look at is you know if you go from the original 10mm detectors way back when you might need a really big beam current. But now you can go lower and lower beam currents, and you can see the resolution improves. Yeah. One I kind of think is interesting is this is what we call extreme ultimate extreme infinity is what it's called now, but you know this is the one that's now kind of challenging you know five to ten K t five to ten nanometer resolution here. This is the same device, again, kind of same beam current.
01:00:08
Mike Hjelmstad: Now we operate at 1.5 kV. Now we start to see the surface structure here. The silicon's really, really well pronounced here, but the nitrogen's really, really well pronounced there. Nitrogen's really tough to do. Yes, especially with a window detector. Another example of this here, this is this is a SRAM device here, but this is again kind of similar to that fin fet in a little ways, where there's a liner material here, and that's the blue that we see. This is tantalum. And this is something that was done. I don't remember if it's set on the screen. I believe it was done at 3 kV. It might have been even lower than that. But again, t at really low short working distances. It takes maybe a little longer to capture all these images and whatnot, but... This is an example of 10 nanometer resolution in the set. Then so kind of just to give you a little bit of do that? Like, what would you change? The goal is kinda threefold in my mind. You know, I don't want to get too salesy here, but like we wanted to change the front end of the detector. So this is kind of a typical EDS detector right here. Kind of have this thing, if anybody's familiar, it's called electron track. It's basically to keep backscatter electrons out of the
01:01:27
Mike Hjelmstad: Out of the of the detector itself. We don't want it impinging on our detector. But this is the part that kinda gets big and clunky and when you stick it next to the actual electron column you can't get this magnet too close, otherwise your beam's not going to actually go where you think it's going to go. So standard EDS tends to have this kind of higher takeoff angle. It tends to have a front end that doesn't let you get as close to the sample. As you want. So we change the front end. We also changed the shape of the detector itself. It's not a circular detector, it's more elongated. And that again lets us just push this detector even closer to the sample. And the last thing was in Valerie asked a question. Hit the window. It's amazing how much more signal improvement you get on some of these low element on low energy lines, on like light element lines, you know. All of a sudden, you know, nitrogen having about three X imp improvement in signal. That opens doors to really seeing these fine materials and really doing nanoscale analysis. The other thing that's really important to this is. We're working at short working distances too. That's kind of your key at low kV. You've got to have short working distances if you're going to get real resolution.
01:02:42
Mike Hjelmstad: This is the detector I feel like we built to kind of say, hey, my cosmos, you know, you want us to work at low K V? Alright, we'll we'll meet you there. No more, no more just cranking the the amplifier to 11. This is this is another part here. It's the same, I think it's the same SRAM if I'm not mistaken, but top down. But the example on the left little bit lower resolution if you see it, it wasn't meant to necessarily be that pixelated, it just blow up and on a screen there. At three kV. You can pretty much work with every single element. And people kind of go, well, wait a second, why do I gotta go to 20 kV? It's easier to get all your X-ray lines at 20 kV, but there is almost an X-ray line for every single element at 3 kV. I think we say 3.5. Definitely everything, but three, pretty much everything still, and especially in the semiconductor room. I think we'll be fine. But we can go all the way to one K V. You know, and you sit there and go, Do you have enough overvoltages? What line series are you using? You know, there's gonna be limitations at one K V, but there's X ray lines for tungsten, there's X ray lines for titanium. You know, now we're using titanium L, we're not just using the titanium K.
01:03:58
Mike Hjelmstad: Okay. Same for the silicon. And then the nitrogen and the oxygen are in that range too. So all of a sudden you can really see this kind of 10 nanometer spatial resolution open up. For something like that. Again, like I said, we could do the other mode, we could work in stem mode. And this is a pretty good example of this is actually not getting away from the window list intention. This is just a regular standard window. Create a lamella, and we're using the thickness to kind of help dictate the resolution that we have here. So, this is actually a gallium nitride transistor. We bake the lamella, we put it in a little holder like this. That's just our little holder for lamella. Everybody's got one. I think, you know, I'll admit, I've probably seen some of the ones on the Vibs here. Probably even better than that. But you know, I want to correlate. Know kind of the higher energy lines. I want to kind of look at the light elements and see this separation here. So I can still use. Things like my titanium K series or my nickel K series, because this one's at 30 kV. I'm generating enough signal. The trick to step. And one of the things that's important, and if you if you kind of look at this, they always say line series and series and series, it's like, why is it saying series?
01:05:17
Mike Hjelmstad: That's amplifying this TrueMap that I've been kind of talking about earlier. But it's important because. You're gonna get signal on a stem lamella. You're gonna get it from everywhere. Beam goes through, it's going to hit some things on the back side. It's going to, you know, fluoresce, it's going to give off different elements in there. So we got to be careful about using the Trumac feature. So that background effects don't kind of lead us down a wrong path. And that's been a good algorithm here. You know, it kind of defines this nickel and the and the sorry the titanium. And the nickel series is right kind of down there at the foot of the of the well that's the gate on the left there, but you know, that's that's something that you can do with staff. I always like kind of showing this one. This one's a really good example of a failure analysis work that was done. I believe this was done with a windowless detector. I'm not 100% sure, but. It's a it's a power device failure. This is basically a nickel or sorry an aluminum core right here. And you can see in the electron image you're like something happened. This does not look like I wanted this to look in my imagery. What happened? And you start to look into it, you know, again, with the TrueMap being your your assistants here. But what's happened here is this is a pa this is kind of a coating here over your aluminum power component. And you can see there's a break in the line with the titanium.
01:06:43
Mike Hjelmstad: And what really happened in here is during the patterning process, you put your photoresist down as largely a fluorinated compound. And find that the flooring can kind of, at least when you look at a smart map, and again, that's kind of the old school mapping I'm kind of describing. See this kind of infiltration, you kind of see flooring in there. A little bit kind of misleading. When we run the TrueMap, we see that it's kind of gotten over the top. But it looks like it's a little bit aggregated kind of near that crack there. Really when we look at the chlorine, that's kind of another like signal to us, because this was a chlorine etch, plasma etch after the fact. What really happened here, and what led us to kind of figure this out was: yeah, the break in the titanium, but the actual location of the fluoride. Kind of told us that the fluorine really infiltrated into the aluminum cavity. During the high temperature processing, basically it started to corrode the aluminum and that's what led to your failure there. Right, so one of the other things I feel like's really relevant that we do for the semiconductor group is working with the EDS signal to try and compute layer thickness.
01:07:59
Mike Hjelmstad: And a lot of people kind of go, wait, you can do that with EDS. You can do it. It takes a little bit of work. But what we kind of look at as comparative tools is something like ellipsometry. Something like Fib, something like XRF. Ellipsometry is not a cheap tool to have around. I mean it's a great tool for determining thickness. It's not necessarily destructive like a fib might be, but you know, it's going to take a little bit of time to do it. How localized you can measure these film thicknesses is, I mean, it's not going to be on the nanoscale, you know, with the with the way it's done. So what we do is we really put a model up. Kind of say, okay, here's here's the film stack I think I have, and there's kind of a good example there. Come up with film thickness estimates. We go, I think it's this thick, I think it's this thick, but I need to check it. And this is the kind of thing where we expect you're looking at the same materials over and over, so the starting guess is probably the same routinely. But then we look at the EDS spectrum and we can simulate kind of here's the model, and here's what the model should read. Here's what the spectrum is telling us. It's a comparison of something called K ratios, if you really wanted to dive into this. But this is a pretty good example of, you know, there's a metalization layer on top of a silicon device here. You can see it's kind of a it's a silicon oxide with a tight nitride on top of it and then an aluminum coating.
01:09:26
Mike Hjelmstad: And in the number two location, that's where, you know, this is where we've done our EDS map and we've kind of colored our our aluminum in the reddish color there. So we can kind of tell that we're getting an aluminum signal for sure. Versus one, we don't see the signal at all. And when we actually put this through the algorithm, we find that, yeah, we've got about nearly a 300 nanometer aluminum coating in here. The ti-nitride's maybe a little thinner at that point than it is at point one. I'll admit that this takes a little bit of development to get this right. You have to kind of dial in the densities of your film. That's really where the big source of air comes from. But it's the type of thing where you do a little bit of work up front and it can be very, very repeatable for for s kind of QA, QC type applications. This is another case I think about this for anybody who's doing lamella. Or trying to d determine thicknesses here, kind of the end stop of like when do I when do I say, okay, enough's enough on the milling. You could you could do lamella thickness with this. And I and I know it's a bit misleading because it goes, well, what's your substrate here? We trick it, we'd tell it burly it was something very light, you know, like the electrons. But we can use the same model here and start to figure out: like, okay, I've milled down and my gold contact here is only about 300 nanometers thick.
01:10:46
Mike Hjelmstad: But the other passivation layers or the quantum wells in this and this is a pixel here. They're still relatively thick, so I can't quite do them at the same rate. I need to kind of look at this a little bit differently. But it would give you the idea, like. When to stop, when to continue, that kind of thing. Yeah. I always like to correlate this. Somebody always kind of said, yeah, but how good is the analysis? Like, how far of a step can you take this? And so you know, I I never knew I never knew quite how much the AFM community loves and they love to use those flakes because very repeatable atomic layers here, but we did it for an example case here of the layer probe and just kind of two monolayers on in one sense. So like I think it's this region that's the two monolayer. No, sorry, that's the single monolayer. This is the this is the two monolayer. When you look at this in EDS, you can actually see a pronounced change in the peak. At the molybdenum and sulfur peaks. And it's an overlap, but we can still sort that out and we can actually compute a roughly 2x difference between the thickness of it. And it gets correlated by the AFM. So kind of a
01:12:02
Mike Hjelmstad: You know, if you check the model, if you dial in the density, this is really can be an accurate tool that you can use repeatedly. Alright, and so kind of moving on to oh, I got ahead of myself, full scale, I I want to talk about this detector that we call BEX. Backscatter electron imaging and x-ray. But it's a detector that's gonna go in from the side, so it's gonna go in under the pole piece. This is something that combines really two different detectors. So if I was a sample kind of looking up at the at the at the detector itself. That's what it looks like. It's an X-ray sensor, it's EDS sensor, I should that's why I should put on the screen there. It's it's about a thirty square millimeter detector and there's two of them. The the basically the dark space we've made into a custom shaped backscatter detector. So we didn't want to get rid of your backscatter detector by sticking an EDS in there. We wanted to make both. What we we're gonna grab is we're gonna grab both signals from the actual interaction volume. Well, why do we want the X-ray signal there? This is the highest collection efficiency for the x-ray right there. Where you have an EDS detector that might be sitting off here at this angle.
01:13:19
Mike Hjelmstad: Could be one of the larger detectors. It might collect, say, 150,000 counts per second. This X-ray detector, these two combined are gonna get about a million counts per second. So it really starts to open the door for us. So it's combining our our backscatter signal. Combine an X-ray signal. People look at that and they go, that's just the X-ray image? Yeah, that's just the X-ray image. But we're going to combine it and we're going to show it off like that. I think this has been on everybody's slides so far, but you know, we're really starting to see something where if we wanted to drive around in real time. And see things change. Long movement. We want to see it move in real time, we can kind of evaluate things, not just on grayscale, but now we're finding things that really where our c the color's gonna identify and our eyes are really gonna catch. On something. It allows us to really go into higher detail. There's a lot of people I meet that just do these kind of maps and just capture an image like this, but this is really an EDS map. They just move on and move on. That's maybe all the analysis they do. You can come back and dig deeper on this for sure. This is meant to serve as a little bit more of a starting point. For full sample analysis. Yeah, this is kind of a couple examples.
01:14:36
Mike Hjelmstad: It in kind of the semiconductor microelectronics phase, like there's nothing preventing you from working with topographical features here. This is obviously a chip right on a board itself. The detector is right above the sample. There's no shadowing going down. So you can start to look down holes and actually get EDS signal down there. Or you can get signal off the edges of something that's built right on the board. Sure you can do the ICs, this is a great chip, great example here, but I always like this one here where it's kind of showing you, you know, there's your grayscale image with a backscatter. You don't necessarily see all the chemistry until you start to put the colors on it and really give yourself an the next slide, yeah, this is a this is actual real example of this here. So I've kind of always liked this one. This is looking at solder bumps. You can see the actual material right here. So we turned it on live, we're looking at an actual solder joint. You gotta have the elements that are gonna come on here. So what you're really doing in this case is, okay, I'm gonna start to examine some of these solder bumps. I wanna make sure I don't have any defects. You start to scan around and you know, oh, woo oh, I always kinda say, why don't we come back, come back? Yeah. Yeah, we got the controls. Oh, that looked interesting.
01:15:55
Mike Hjelmstad: There's some kind of silicon defect going on with that right there. And you kind of go, well, what the heck? You know it's silicon already. You need to dig in a little bit deeper? Yeah, we can stop right there. But what I want to do is I want to scale the whole thing. So I'm just going to take my reference image. I'm gonna draw a box over it. I'll come back to this if I really care about it, but I'm gonna get through this quick. And so I'm just gonna scale this up. I forgot if the video I hope the video speeds up here. I think it does about seven frames. But this is the kind of speed that we're doing this at. So we're trying to get over large area samples in minutes, maybe hours, but we're not trying to take days or overnight or anything like that. Yeah, there's the end of it. 44 fields in that example right there, two minutes and seven seconds. So that gets you an idea of what's possible. So I think of really, really small things in setting detector over huge areas. And this is a pretty good example of how to tackle. I'll get through the character. So I'm moving on to EBSD here. I'll try and hustle it up on this one. Couple kind of fun examples here. This was taken by, I think it was our counterparts in China. Did a wire bond here. Now, what you're actually seeing is this is the orientation.
01:17:11
Mike Hjelmstad: And it's relative to the actual z direction, the x direction in this case. So don't get too caught up on it. What I really want you to see is that as the wire comes from one source, it's all kind of reddish. It's kind of following this direction horizontally. And as it turns, it comes down to the bond itself. You can see an orientation. Can see an orientation change, and it actually kind of goes from this. Preferred through the wire, and it kind of twists and it kind of ends up in this orientation as it comes down the wire. Basically, what you see is one zero zero direction starts to align with the along axis of the wire. That's something that you can do. These are really two different maps that were acquired, but then kind of stitched together. The other thing that you can look at, and this is again something Kevin did a great job of kind of exploring and richly is these this is that we call it kernel average misorientation, but kind of that intra-grain strain kind of thing. You see, like as the shape changes here, as it turns. You start to see elevated stra a straight. And obviously near the bond pad itself, you're gonna see that elevated strain. The image on the bottom is actually kind of which grains themselves have the highest distortion, is grain orientation spread if you want to look at it that way.
01:18:30
Mike Hjelmstad: But that's a good example of kind of highlighting it. I had to go to the kind of the plint chip packaging because this is a really really interesting one at least in my opinion. One of the things that happens in this and I think yeah, it's the nickel here. So this is your EDS map. These are the different phases that form. And you get a lot of innermetallics that kind of form in kind of the. The tin kind of silver regions there. But what happens is the grain structure is wildly different between all these different layers here. And the one that's really tough to refine is mickel. Nickel ends up becoming really, really deformed in this case. And it actually becomes the point where we really struggle to get good patterns out of this. There's an app note to this that we've actually improved the data analysis a little bit using pattern matching software. I won't dive into it now, but I know we can do a little bit better than what we had here. But the fact that we had zero solutions at all is it's because it's a higher higher deformed region of the sample in there. This was a technique that was brought to me by the plasma group and they kind of said, we're looking to figure out if we've removed contamination on this is a cross section of silicon carbide in this case, but did we remove this thin film of what we figure is resist, you know, basically something non crystalline?
01:19:52
Mike Hjelmstad: And if you have a little bit of a coding, you know, Kevin kinda mentioned it's really important to polish EBSD samples really well. The patterns start to degrade really, really quickly with mean like nanometers worth of coding. This is what it looks like where you actually have a little bit of coating that's remaining, this massed area that's just a little bit fuzzier. And we can kind of measure the metric. We can kind of measure the band fuzziness if you want to look at it that way. But the one on the right, that's the pristine silicon carbide. And you can go, I mean, we're clear. Of our coding in that exam example there. I love the fact that it kind of did a gallium nitride example because this is another one that we did, but again, with these films, you start to see these little kind of topographic areas in here. And we were able to get patterns in a lot of these cases, if it was measured right. And what we attribute it to is threading dislocation. Sample. And one of the things that we didn't go through the whole routine of EBSD and we might do that for the demos, but you know, you're kind of taking a pattern and breaking it down into a a separate transform and trying to measure all these inner band angles.
01:21:11
Mike Hjelmstad: There's a degree of accuracy that gets lost in it. You know, this starts to just look like a wash of dislocations in here. It's really tough to pick it out. And what we can actually do is we can do this pattern matching algorithm and kind of improve the orientation accuracy and really kind of pinpoint exactly where these dislocation centers are. And then we've run a routine called weighted Burgers vector. Now again, getting away in the weeds, but I can tell you based on the orientation data here and the relative orientation data route, I can tell you whether these dislocations are into the plane or out of the plane or whether or not they're actually edge dislocations, not even screw dislocations themselves. So you can do a really thorough analysis of your thin films if you wanted to. And this is maybe the last one. One of the things that we didn't talk about earlier was EBSD is done at 70 degree tilt. When the electron beam comes down, it spreads a longer spot. So effectively, you limited resolution. If you do it traditionally at 70-degree tilt, your resolution limits probably around 150 nanometers. There's a technique that was developed, I think we're going on, almost fifteen years now, called TKD.
01:22:29
Mike Hjelmstad: And the idea was create a lamella. And have the beam go through the lamella. And actually kind of put the diffraction signal onto the phosphor screen. Using all the same all the same hardware. Didn't change anything in the hardware here, so we can actually do what in this case we call op access DKD. But now we can start to resolve grains that are even smaller. I think in this example, we're probably on the order of maybe. 20 nanometer grain size here, but this is this is silicon channels in kind of a tungsten matrix here, but you can actually start to resolve these smaller features. You can see the scale bar here at 200 nanometers. I mean, we're stepping the pixels out at that. Two and a half nanometers to kind of decipher that. But just wanted to kind of point out that this is another technique that's been invented, something that really pushes resolution a lot. Kind of the last thing that we've kind of thought about is is there anything we can do with the hardware to maybe make this more efficient? One of the things that we just recently introduced was why not? Well angle the actual phosphor. Now we can put our lamella in dead flat, because previously we kind of tilted it back to get a little bit more signal.
01:23:46
Mike Hjelmstad: Now we're going to put it dead flat. We're going to get as much signal as we possibly can onto the sample there. And we can really do something I've I think we've shown like four nanometer resolutions. That's kind of the end of my talk. I just wanted to kind of get to the fact that, you know, EDS EDS, EBSD is there for your elemental ch characterization, there for your microstructural characters. Mm-hmm. We're going nanier scale sub 10 nanometers.
01:24:20
Mike Hjelmstad: Okay. Do we have any questions from the audience? Hey? Very nice. Course.
01:24:50
Mike Hjelmstad: It it's where it's it's where I like to say we do that for you. So the idea, yeah, at three K V as Dilro was saying, you're going to have way more overlaps. You know, that's the threat. You know, now all the X-ray energies are comb just packed into this region where. Resolution of the detector might not be good enough. You rely on good detector characterization. And and that's where, you know, I again, it's going to be a pitch for me in a sense, but it's you gotta make sure that the peak shapes that you believe you have on your detector are very, very accurate. We don't take them just from theory. It's been a lot of experimental analysis. It's involved work on synchrotrons to make sure that hey, you know. The silicon L peak when it goes through our our chain. It does spew it a little bit, or it does suppress it a little bit. The shape is more like this, not that, you know, and trying to get the shape. Theoretically as accurate as possible. And then relying on the least forest fit to do the work. And so, really, the only way to say, did you do a good job is find some known standards. Test it, and then you kind of go, yeah, it looks pretty good. And that's one of the focuses of the new detector that we did was to have real individual calibrations for each detector.
01:26:11
Mike Hjelmstad: In case one's just a little bit different, you know, we're trying to build the highest end, but you know things age. Is it gonna cause a tailing artifact or is it gonna suppress this one a little bit? So there's calibrations built in now for individual detectors. But actually the lady question. So you mentioned that historically yeah, TrueMap. Yes, math is. Well we've all we've always had kind of the original smart map, if you will, where it's an energy window and then the element. Yeah,'cause I wonder if you could kinda speak to intervention there's and then also is that purely a software difference? Hardware. Yeah, it's a great question. So first part is pure software. Okay, so it's a matter of interpreting the spectrum. And you have regular mapping. And I always kinda tell people, regular mapping was built for computers like thirty years ago. Where it was like, yeah, if I want to display this in real time. I don't have a computer that's extraordinary. So, what people would do is they'd just create energy when known as spectrum. You know, you just say,
01:27:27
Mike Hjelmstad: I know I have silicon, okay from like 1.6 to 1.8 it's gonna be silicon. Anything that falls in there, silicon. Problem is there's overlaps and all of a sudden tail can start falling in there, tungsten can start falling there. So that was the original method. TrueMap. Goes to each pixel, each pixel has a spectrum associated with it, and it just does two things that are associated with quantum analysis. It filters the background out. It does the peak team convolution. And you go, that's it? Well that's something that we can do in real time. So when you map a TrueMap you can actually see the map come through in real time. Can go a step further. The third thing would be quant map. And quant map would be: okay, you've done the filtering, you've done the peak deep evolution. Now compare what how many accounts you have left to a standard. And do the matrix corrections for the other elements that are present. So obviously building a quant map. Little more computer intensive. I think I did the math in my head. If you have like a 512 resolution map. Three hundred thousand pixels or something like that. So you're not gonna spit those pix you know, that that quantification out in real time. So it's more of a post-processing technique. At the same time, if you wanted to, you could bin down the map resolution and display it.
01:28:48
Mike Hjelmstad: Those are really the three maps that are out there. Yeah. Yeah. Yeah. Wait, which one?
01:29:11
Mike Hjelmstad: So much. That one of yeah, where alternate at the end. But in a bottom side, yeah, yeah.
01:29:31
Mike Hjelmstad: Multiple? That's what we felt it was. Like, the patterns I've seen of that area, really, really fuzzy, fuzzy nickel patterns, which is really strange. Like, if you deal with, like, a nickel part. It's very typical that it doesn't take much of an incurred to get really nice sharp patterns. So we kind of attributed our difficulty there. To probably retain strain in the material there. Could be the prop. I mean, it's not to put that outside of the possibility, but. Also greying now different some but it's two to the growth fast and then that is might be related to the sum could very well be. I mean, you don't see it much in the EDS on the left. But I don't know how long that that was actually captured, if it's going to be enough to really show you diffusion kind of at that at that level. So I I think I think your analysis is pretty good, you know. Now the second question is if PSL provides some pace identification with the EBSD together. And if you also thought you're in a getting together then you pace the identification with the EBSD ESD together.
01:30:50
Mike Hjelmstad: So this one, yeah, this one probably used probably used the EDS to help influence the phase identification for sure. Something we call true faiths, but it's just we're capturing the EDS. Might as well use it as a tiebreaker because you're hitting on one thing with EBSD, it's going to tell you a crystal structure, not necessarily a chemical phase. But the EDS can help you kind of reinforce what phase you have. Unless one in the back one, unless you want.
01:31:30
Mike Hjelmstad: Yeah. I I j I joke with a sales guy, like it grows so fast, like why aren't she up there all the time? Yeah. I was really blown away by the speed. You didn't talk about it. Experiments. While I need for spraying our yeah, but saying how it's easier to do in the GEM. Yeah, or not at the same time. Are you thinking of it from EDS or both? Or both EDS, VSD, EDS? Yeah, I think there's a lot more interest on the UBSD side lately of what we can do and some of the things we've added. So it it'd be an interesting thing to find out. Lab like ours, we do a lot of gold room. Look for s a solder joint with silicon, right, or some defect or crack. How close are you guys to developing algorithms? Can do where you teach it, to say how good is? Are you going to give me my statistical analysis on everything that looks like this? More of a computer vision or machine learning type. Good question and I've been asked by a lot of people and and unfortunately I kind of get the the crickets from the factory sometimes.
01:32:49
Mike Hjelmstad: One of those like you guys aren't working on it, right? You kinda go you know, it it it's I I'm I'm honest to God, I I they're they're playing a good poker game with me. I don't know if they maybe they're thinking about it, maybe they're not. Great. Everybody
01:33:25
Mike Hjelmstad: You showed with the this is particular interesting from a Raman or photoluminist sense point of view because when I do the P L measurements on gallium nitride I see these same kind of patterns where the P L shifts in intensity and in energy. So there there we can learn a lot about the the the bar gap properties and doing some correlative measurements here is again opening up some some really new perspective on the examples. I like the idea, yeah. I mean, I honestly think that that example probably was taken before there was a real merger of the Raman. Lyle has this gallium nitride, this Frank-Read source example. There you see this a similar pattern in the P L image where you see like points where you have stronger intensity and stuff. So this is a very interesting sort of again the correlative stuff good, yeah, good example. I mean I w I didn't think of that one.
01:34:42
Mike Hjelmstad: Question. Okay, I love the bugs detector. Yeah. So great. I've always loved the bags scatter detectors that go in between the sample and the whole piece. But the problem is always comes up when you want to tilt the stage inside of a cross-section. So wouldn't it be cool if there was like, you know, maybe two detectors, one that comes in flat and then one that comes in tilted while it's tilted? Or maybe the detector itself could tilt the side. You ever thought about anything like that? I haven't. Has a factory? Probably. And I think they think about it more from like maybe a stem detector type. Mm-hmm. 'Cause you also kinda have this, you know, I kinda had the TEM example. It's like or the st I guess the stem example is a better example, but yeah, your primary beam goes through, doesn't a lot of your X-ray signal kind of go down as well? Couple. I like the idea, but can't pay your royalty on it. I have two questions. Where is the factory actually? The factory is in in a town called High Wickham in England. It's it's roughly halfway between Oxford and London.
01:35:59
Mike Hjelmstad: And then a lot of grant I understand is semiconductor focus and application seminar. Victor and I are from the automotive industry. Could you speak a little bit? So, a lot of what we do is more like macro scale. Can you speak a little bit about, say, like PBST for like macro scale? You know, what's kind of the maximum resolution in like terms. Look at aluminum castings. I mean it it d it depends on the time that you got available to you. So, you know, one of the things that's happening with EBSD is build a faster and faster EBSD detector, okay? The examples I've had here are not going to go at the screening edge of what what EBSD is capable of, but if you were looking at like a just an aluminum casting set, you can probably drive 4,000 patterns per second, maybe 5,000 patterns per second out of it. You can do smaller regions in minute or two and start to scale it up and tile it. That's really where the limitation would be in terms of overall dimension. So, how much can you move around in the summit at 70 degrees tilt? You know, and how much time do you have? But like a good example for me was I have a lot of geologists that have sections, thin sections that are one inch by two inch.
01:37:16
Mike Hjelmstad: And they go, yeah, you know, EBSD, I used to get like three patterns per second. Now you guys can do like. Hundred patterns per second. This opened up a whole new world to me on you know, it's thirty times what they used to do. They do the whole they do the whole thing overnight. It'll do something that can learn. Up windows in that room, but you know, can I see like a a six-inch wafer over there? You'd probably find a region of interest and do, you know, a small little window. Yeah.
01:37:50
Valerie Brogden: Any other questions for Mike?
01:37:58
Valerie Brogden: Now I'd like to welcome to the podium Covalent's very own Lyle Gordon. He is our director of materials and surface chemistry and also innovation. And he is here to talk about the Raman side of things. Really exciting. We just got a brand new Raman tool from Oxford. So yeah, he'll cover kind of the basics. The theory of Raman spectroscopy and give some examples of common applications that we see at Covalent.
01:38:30
Lyle Gordon: Thanks a lot. All right, so I'm going to give an intro into a little bit of the background of Raman spectroscopy, how it works, what we can learn, and then a little practical information, kind of how measurements go, what can go wrong, a little bit how you can fix it, a couple examples, sort of what Covalent sees day to day in terms of problems for our customers, and how we try to solve them and how we use Raman to do that. We're a compliment. Some other techniques we'll talk about. This is pretty informal. Start with a little bit of physics. If something doesn't make sense, please stop me and interrupt. I think I kept it pretty short, so we should have time. And then I'll point out a few times where Thomas from Oxford is going to come and really go in depth a little more into the instrumentation and some of the examples. So that's pretty much what we'll do. So just brief outline, covering again, so how Raman works, it helps understand a little bit about the physics of what's going on, that will help you understand what the information is, what it tells you about crystals or molecules, and then that really helps you understand how to interpret that data. Then how do we actually go from, okay, we have some physics, how does the measurement actually work? What kind of data can we get? That's kind of like some of the SEM that Kevin talked about, where we can do maps or points, and then some of the challenges. And then the specific applications we'll talk about. Tried to focus a little bit on the semiconductors, but not just semiconductor materials, but the whole ecosystem. So one of the most important applications we use Raman for is really contamination. So people are processing something, they have some defect, they found a particle.
01:40:06
Lyle Gordon: Or something that's not supposed to be there, where did it come from, what is it? And Raman is really powerful for that, especially when we combine with EDS, because we can identify relatively small particles, so other people might be using FTIR and you're limited to bigger things. And then we'll talk a little bit about semiconductor materials and then Thomas will go into that a little bit. So how Robin works. So little physics lesson for everyone, so we're all on the same page. We're gonna start with a pretty simple example, a molecule probably everyone is familiar with. This is just a water molecule. And so all molecules that are above absolute zero, as well as crystals, they have vibrational modes that are specific to different ways the molecule can be formed. So, water is pretty simple, just two hydrogens and an oxygen. So, it has like a bending mode so that it can go like this, or those bonds can stretch, so you can have like symmetric stretches or asymmetric stretches where they're opposite. So, water is pretty simple. That's really all you need to think about. Obviously, you can imagine more complicated molecules will be more complicated, they'll have many modes, especially if they have different atoms. Those atoms are heavier, they vibrate at different frequencies. And then we'll talk a little bit about the end. So this is obviously more kind of chemistry focused, this simple example. Many people here probably are thinking about like crystalline materials. These same things apply to the phonon modes in crystals and you can measure those vibrational modes in the same way.
01:41:35
Lyle Gordon: So all materials have these characteristic vibrations. So now what happens when we shoot photons of a certain energy at these molecules? Most of what happens is sort of uninteresting. They just elastically scatter off. This is called Rayleigh scattering. And this doesn't really tell you anything about the material. There's just light scattered off of it. You can see this around you or if you're doing measurements in the lab, you shine a green laser. The sample just is really brightly scattering green light. There's not much information here. And then if you look A very small proportion of these incident photons, maybe around one in a million. Will actually scatter inelastically and it will transfer energy into these vibrational modes and lose a bit of energy in that. So you'll get a photon out that has lower energy, or we'll talk about after you can actually get higher energy. And this is really important because these ones actually sort of encode that information about the vibration of the molecule so we can see what that molecule is and what how what those energies are. So how does this happen? So this is a little bit if you haven't seen any sort of energy level diagrams in molecules, this might be a little bit opaque. So we'll try to explain it a little bit, but we only have a few tens of minutes. And so if you don't understand, we can talk later, but hopefully you'll get enough that you can get the practical aspects of understanding all the physics.
01:43:02
Lyle Gordon: So basically what happens is we have sort of two possible things. Yeah. So this is the boring case, this is the Rayleigh scattering. So the photon in, you can think of this sort of virtual energy state, and then it just scatters out and it comes out with the same energy. And that's uninteresting. You don't learn anything there really. But another thing that can happen is you can float in and then come out and you lose a little bit of energy, and that's by exciting that vibrational mode. So that energy difference, that's the energy of the vibration of the molecule. Difference between these two energies, that's called the Raman shift, and that's what we're really trying to measure here. And that tells us about the energy in the molecule. And if we then look at that in terms of this spectra, we can see this difference. So we have this big peak here, which we mostly don't look at. That's from the Rayleigh scattering. And we have this smaller peak here, which is that shifted energy. And this is only showing that bending. There's going to be other modes here, and overtones, and other complicated stuff, which you'll see in the spectra, but this example is really simplified, though the stretching modes are not. And then one thing which is of some interest, and I think Thomas, if he has time, will get to some examples of this, is that since the molecules above absolute zero are all vibrating, you can actually put this photon in and get a new photon out that has more energy. So the vibrational mode of the atom or the molecule can transfer back into that photon. And this is called anti-Stokes Raman scattering. And with more specialized instruments,
01:44:36
Lyle Gordon: You can measure those as well, and they can tell you interesting stuff about like the temperature of the sample. And that's Thomas will hopefully go into that, and that's pretty exciting as well. So this is basically what happens in Raman scattering. And so what does that look like when you put that together and actually take a measurement? So this is a measurement we did the other day. We just put a drop of water on the instrument, shine the laser on it, and record the range of Raman shifts. So this is in energy units of inverse centimeters. So this is a lower energy shift, so closer to the incident wavelength, and this is a higher energy shift. So this little peak here around 1600, that's that bending mode that we showed in the little toy example. So you can really see really nicely resolved the molecules bending there. And then there's two peaks here. That's the oxygen and hydrogen bonds stretching either symmetrically or asymmetrically. And you can see that they overlap. So you can see there is a shoulder here and there's actually two peaks buried there. But because the waters are in different environments, there's all the hydrogen bonding. Those peaks are actually broad, so it's not the resolution of the instrument like in the EDS detector. These are actually broad peaks that cannot be resolved. So we'll talk a little bit more how we can fit these peaks and do similar deconvolution to the EDS to really find the position of them and learn about the material from that.
01:45:57
Lyle Gordon: And then yeah, so that that's water. Does anyone have any questions so far? I think that's the intro to the basics. So hopefully that makes sense so we can then go on to how measurements go. Alright, we'll keep going. So I talked a little bit in the beginning that real molecules, like if you think about complicated polymers like epoxies, they have lots of different modes. So water was really simple. Everyone's familiar with it. And so it's a nice example to understand. Obviously there's way more modes, especially if you think about the backbone of polymers. You might be working on like photoresist. They'll have lots of like CH2 or CH3 groups on the backbone. And those can have lots of different modes. Spectra get a lot more complicated. But that's really useful because you can use these to fingerprint the different types of molecules that. And then I mentioned as well that we can have phonons in materials. So crystalline materials will have specific lattice vibrations. We'll show some examples of like gallium nitrite. This is just a little animation of one of the optical phonons of a diamond lattice vibrating. And you'll similarly be able to measure those vibrational modes with Raman scattering. And just like the organic molecules, those phonon energies will be specific to different materials. So you can use that to identify contamination, or we'll see that those phonons would shift with temperature. So if the sample's warmer or colder, the phonons could soften. And then also, if there's strain, especially localized strain, we can see those phonons change as well.
01:47:27
Lyle Gordon: And that's how we can go in and probe and really learn about the materials. When we looked at the modic world we cannot I didn't realise we still had some of the thank you. Okay, so now practical measurements. How do we actually collect the data? So we'll start with like what is Raman actually good for? So in this case we can identify materials and the advantage over some of the other techniques like if you're looking at like TEM where you need to make thin sections almost all the Raman we do it really involves no sample prep. We just put it on the stage and it's primarily non-destructive. We'll talk a little bit sometimes the laser can damage the material but mostly we try to avoid that and that will only be a small little so mostly we're not destroying the sample. You can give it back, maybe do other characterization if you need to. You could go in and make a TEM sample afterwards. And you can measure basically anything from solids, liquids, even like gases and bubbles and materials, any sort of matter you can measure. And like in the SEM, you can do point spectra. So if you just have one interesting feature you want to measure, you can extend that to line scans, 2D. And it's really interesting in Raman as well. Since we're using light, if the material is transparent, like a glass or many plastics, we can actually do 3D volumetric imaging, or you can do 2D cross-sections into materials without actually having to cut it. And that, again, goes back to being non-destructive. We might be able to see.
01:48:57
Lyle Gordon: How thick is a layer or are the buried defects without having to prepare a cross section. So that's not only like you can give the customer the sample back or do other analysis, also saves you a lot of time. The measurement could take less time than it would have taken just to make a cross section like a fib or mechanically. So that's really powerful. But we'll talk about some of the challenges where certain samples might have fluorescence and that can overwhelm your Raman measurements. So you can use different lasers. In our toy example with water, it was shown with just the one of the most common, just using a green laser at 532 nanometers. That's what we use for most things. But you can use other wavelengths like UV or near IR, and that might help with certain materials that either have resonances that you can get extra signal or probe different bonds or also avoid some fluorescence. So what doesn't work? Basically some met like materials like metals don't have Raman scattering. There's also high symmetry materials like magnesium oxide for example that don't have Raman scattering either. There's all those transitions are forbidden. And I mentioned a little bit, some Raman scattering is very weak, so maybe one in a million photons. If you have strongly fluorescent or photoluminescent samples, that signal can overwhelm it, but there are some strategies to get around that. And Thomas will talk more. Sometimes that photoluminescence is actually very interesting, like to study defects in semiconductors or dopants, like rare earths and materials. So in some cases, if you're just studying Raman, it gets in the way and it might actually
01:50:28
Lyle Gordon: But in other cases, it provides a wealth of information that's really useful. So you do have to manage your laser power, kind of like the electron beam dose in the SEM. You can damage more sensitive materials. We'll show some examples where materials just get destroyed, especially if they're dark colored, you could just see that they're going to absorb light, that they may be more likely to get damaged. And Raman, like the SEM, is measuring pretty small area, usually like the spots around a micron. So if you have a heterogeneous material, if you were to measure with X or D, you might measure quite a lot, like many milligrams of material, and get a good statistical average. With Raman, a few points, if your material is variable, that might not give you enough information. But with the modern tools, we can do really fast mapping and collect really good statistics. And we do that a lot for like graphite, for batteries, where the statistical average is more important than any few points. Alright, so how do we actually do metroman? So early instruments were really slow. Basically, they were serial. So they had a monochromer measuring each Raman shift serially. And it could take days or like hours to days to collect a spectra, use a lot of samples. It was presumably just like chemists studying molecules, figuring out the theory. And it wasn't really a practical approach to materials analysis like we can use now. So modern instruments use a CCD so you have an array of detectors.
01:51:53
Lyle Gordon: Basically, and gratings disperse all those energies on the C C D. So you can measure them all together and you can move the gradings, you can switch gradings to get higher resolution or measure different range of shifts for different materials. So they're really powerful. And they're very sensitive and you can use large amounts of laser power if your sample supports it to do measurements quickly. So even like milliseconds is possible for certain materials. And that translates to very fast mapping and getting data in minutes. Of like overnight, and so you can learn about things really. So this is just a picture of the instrument in our lab. You can really only see the Rayleigh scattering, but in there as well is all those other wavelengths that the instrument will be picking up. And this is just a quick schematic and Thomas is really going to go more in detail about the specific Oxford instruments and some of the advantages. And then we'll talk a little bit about how we extract information from that spectra. In the example with water, we saw that some of the peaks were too broad to individually resolve. So how do we go in and extract some of that information using software? Alright, so now this is kind of what real data looks like. The water was quite simple. Most spectra from materials that we're studying, especially organic materials and more lower symmetry in organics, will have many peaks. This can make
01:53:12
Lyle Gordon: Intuitive understanding a bit more complicated if you're familiar with like infrared spectroscopy, these vibrational modes may look familiar. These are like carbon-hydrogen bonds, these are like CH2 modes. So the selection rules are different for Raman, but the intuitive understanding is simpler. And then for more complicated inorganics, it's not that intuitive. You won't know which phonon modes these are. You'll have to go look in the out. But you can match these to databases which are available and really quickly identify stuff. And this is powerful because different polymorphs of the same material, and we'll show some examples, will have really different phonons. So they'll have a different fingerprint and you can tell them apart. Whereas if you're just using EDS or some elemental analysis, like all the SiO two will basically look the same, but they could be very different. Alright. So one important aspect which we'll show a little bit and Thomas will talk more about is how do we extract quantitative information. So we saw we can fingerprint things and just identify what they are, but what if we want to learn more about it? Like is one part of it strained or are there different defects? What we can do is actually go in and do peak deconvolution. In this case, this example is really easy. These peaks are all nicely with
01:54:30
Lyle Gordon: Resolve. They're relatively broad, so we can measure them without really high resolution. So we can just go fit relevant line shapes. So these correspond to like different bending and vibrational modes in the quartz lattice. And then we can determine accurately their position and their width. So they might be broader indicating more disorder or they might shift indicating like different densities or different strain in the material. And not really going into it, but polarization state can also tell you there's going to be different intensities and the instruments can control that as well. Right. Back to like what we can kind of measure here. We talked a little bit about this. So the simplest is points, and Kevin had a similar slide, so we won't spend too much time. We often do line scans, especially if material has relatively weak Raman scattering or might have a lot of fluorescence that you need to acquire for long. Sometimes a line scan is more efficient. And then if we want, we can do 2D and 3D maps, especially when we're thinking about semiconductor wafers, then 2D maps become so some of the challenges in terms of practical measurement, we want to be able to extract the relevant information from the material. So here we're really focusing on getting the Raman spectra, so understanding those vibrational modes.
01:55:50
Lyle Gordon: So it's pretty weak. And modern instruments are very sensitive, so we can do fast measurements. And I mentioned there's some samples that won't give any signal at all. This can be good if you have like a powder you can put it on a metal, like just aluminum foil, and not worry they'll get too much background. And we mentioned that the laser can change your material. Here is actually just like a pair of tweezers that had a black coating on them. So I tried to measure that just an example. And you can see the laser just like blew up a big crater. This is really exaggerated, so you can see it on the slide, but sometimes you can see changes or your spectra is changing. And you can see the result here if you're familiar. These are just the D and G bands of disordered carbon. So we turned this polymer. Which we know is on that Tweezer coating, and just turned it into carbon. And we know this measurement's wrong, because we just basically destroyed it with a this is really exaggerated, but you do see these kind of things on sensitive materials. You can also heat materials. This is like not the best example of this because GaN doesn't absorb very much. But you can see a really small peak shift as we go to much higher laser power. And I was just measuring GaN. But if you do this on something like silicon, where we can see with our eyes that silicon is going to absorb, you can heat it like 200 degrees with like 50 milliwatts and the peak will shift a huge amount. And then you're obviously not measuring the sample itself. You're just measuring how much you heated it.
01:57:13
Lyle Gordon: Like not very useful. And here's a nice example. This is another polymorph of SiO2 crustabolite. And this is a natural geological sample. I just borrowed it from a database. So you can see the Raman spectra is still reasonably resolved, but there's this really broad fluorescence just out at a bit longer wavelengths. So in some cases, that fluorescence can be so strong that you can't see any basically, it's impossible to collect data. You can try other wavelengths. And there's other strategies to try and control that. So hopefully you can get data and then you can subtract that fluorescence background. And then just reminding again, like the volume we probe is really small compared to a lot of other techniques. Unless you're a TEM person, then it might feel really big. But for lots of materials, this won't really give you that average. And so it's important to consider that if it's a heterogeneous thing, just collect lots of data, especially if that's feasible. If it's something uniform, you might not need to. All right, I'm gonna talk a little bit about some of the examples. These are really focused on things we kind of see here at Covalent, like what we actually use this for in sort of the semiconductor ecosystem. And these are pretty simple and Thomas will go into more detail in some.
01:58:29
Lyle Gordon: So probably the biggest application we see in general for vibrational spectroscopy, so IR and Raman, is contamination or like foreign items. If it's a big thing, we can really quickly do it on the FTIR. And if certain things can be done on the FTIR microscope, but Raman is really flexible, even down to like few micron-sized contamination particles. And so that's where it becomes relevant to like semiconductors. Stuff. So you have a lot of different contamination on wafers. You might have like molecules which you would meet with GC. Then you have sort of elemental contamination at very low levels. There's some analytical chemistry you can do. But if you have particles on a wafer or other sort of semiconductor surfaces or circuit boards, and you want to figure out what they are, then Raman becomes really powerful because, going back to those fingerprinting, here we can see some measurements of some polymers. This is just polyethylene and polystyrene. We can see that they have very different fingerprints. So we can go in, measure those contamination particles people found maybe by like light scattering equipment on their wafers or associated with defects, and then go in and figure out what they are, and then hopefully track them back to different. Processing equipment or where it would come from, so you can actually fix the root cause. And Raman is really good at that and complimentary to like SEM-EDS to get compositional information.
01:59:49
Lyle Gordon: And uniquely with Raman, we showed you can detect the phonon modes of the inorganic crystalline materials. That's a lot better than FTIR, so we're able to see those inorganic materials and identify them. And I'll show a little bit more about that. So this is maybe a bit more like geology relevant, but a nice example that illustrates what we can see. So, like the organic materials, slight changes to the molecule can really dramatically change the spectra we measure so we can tell what they are. And in this case, we have the exact same composition in all these samples. And so, we have three crystalline polymorphs of SiO2, and then one amorphous, so just amorphous SiO2 like oxide on a wave. And you can see they have really different Raman spectra. So we can go in and identify those like you would do an XRD. But here you only need a tiny particle of material and we can tell them. So really useful if you're trying to figure out where some particle came from, especially if you're looking at different polymorphs. And this type of information is not really available by FTIR, so it's really powerful. Alright, the next application will talk a little bit about semiconductor materials themselves, so not just the whole ecosystem, but the semiconductors themselves.
02:01:04
Lyle Gordon: So here is a measurement. So I just did this the other day here. This is just a simple epi GaN on a sapphire substrate. So we measured the Raman spectra of the Gan and you can see there's the different phonon modes we can easily measure. And then Thomas will go into detail on this. But you can see the peak from the sapphire when we're focused on the surface is really, really, really small, even though this is all transparent. So we're really good at selectively measuring only the layers we want. And just neat example. So these peaks will be sensitive to like stress or disorder, and we'll show a nice example that Oxford has on looking at that. And then what we can do here is go and basically move the sample relative to the focal point. So in this case, we're just moving the sample towards the objective and so moving it up. And then the beam goes through the GaN. So we can see the peak area of this GaN phonon mode goes up and then it goes down as we get through. And then we see the sapphire phonon mode. You can't really see, but it would get big. And then back as we go into the substrate. Sapphire has a smaller Brahmin. Cross-section, so the peaks are inherently much weaker than GaN, and that's why this is really nice and not noisy, and this is a bit noisier, even though they both are about like kind of similar density, similar amount of material there.
02:02:27
Lyle Gordon: And that's illustrates what we can do in transparent materials. You could do those depth profiles. We could have extended this to two D or three D maps. It's kind of boring here. It's just a single crystal of can on a single crystal of sapphire. But if you had defects you could see more. And this is showing the example where we can actually there's a defect site. This is a dislocation source on a gallium nitride film. And then if you map that peak position around it and you go into every spectra and fit them all to see exactly where that peak is with very good precision, you can then generate a histogram of where those peaks are and generate maps of tensile and compressive stress. So this is one of the applications we would do commonly on like silicon, people looking at defects or stress in wafers and other materials like gan or silicon carbide. All right, I hope that gave a quick intro. People have an idea of what's going on. So happy to take some questions and then Thomas will go into more detail.
02:03:34
Lyle Gordon: In case of L sensitive material, the case in a few L time. Order, which is enough. You selected some spatial glasses. Yeah, if you're using visible light, then like green, regular glass will work. It has relatively weak Raman scattering and you can just focus through it. You just need to make sure like your objectives have enough working distance. But yeah, you can look like in situ Raman measurements either for air-free applications or we've done like in liquids or even like high pressure in diamond cells and at higher temperature is often done like in air free conditions as well. So that's really practical. Basically, you just need things that are transparent to light. In if you're looking at IR and UV, then it's a bit more tricky, then you have to make sure your windows are right. But for the standard like green laser glass diamond windows work really well. Building out a materials characterization lab and you had to choose something from it or negative bad decision. Thank you. I think from like our uses of it, if you're mostly measuring like bulk polymers, the FTIR is fast. The other nice thing, well, the FTIR is also significantly cheaper. So if your budget is important, that makes a difference. It's also easier to train people. There's less knobs to turn on an FTIR. You pretty much just measure modern ones that are ATR. You kind of push a button. You're not worried about detectors. You're not willy.
02:05:09
Lyle Gordon: Worried about gradings and resolution, it's almost always just like kind of hit it and push a button and go. But if you care about inorganic crystals or you need microscopic resolution, that's where Raman becomes really useful. They're definitely complementary. We often, like, customers will send us stuff and we have both, so we don't know what we're going to get. We know that one of them is going to solve the problem because there's some sort of material there that we don't know. It is, but it's hard to know upfront. So we kind of just try one, if it doesn't work, try the other and then get the customer the answer. So it's hard to make a decision, but definitely like the budget and complexity is a big factor.
02:05:52
Valerie Brogden: Okay, and now I'd like to welcome to the podium again Thomas Meyer from Oxford. He is a senior application scientist and he's going to go more in depth into Raman applications for us.
02:06:06
Thomas Meyer: Okay. Yeah. Thanks for the invitation and getting me over the Atlantic Ocean to give this presentation here. It's very nice. We had already yesterday a pleasant day and had a look at the system, so I also invite you to the demo sessions in the afternoon. I'm going to talk about the confocal Raman microscopy for advanced semiconductor metrology. From microstructures up to full wafers. Okay, let me give you a brief overview. So the first part I will give you a short short information about some backgrounds, about confocality and photoluminescence in in particular, and then we will go through some examples from the whole semiconductor production chain yeah and including also a sample of polymer. Because at the end of the day when you have your chip finalized, you want to maybe glue it into a a credit card and then of course the polymerization also becomes of interest.
02:07:24
Thomas Meyer: So, yeah, what do we wish for? So, we first of all we want a chemical selectivity, so investigating polymers and crystals, all is available for Raman. Nice thing about Raman is we are non-destructive and non-invasive. It's fast or we need the fast and confocal acquisition to really make the 3D images, for example. And this makes Raman accessible to different types of samples. So not only semiconductors, but also chemistry, biophysics, geology and of course we can have a look at various sizes so we can go up to three hundred millimeter wafers with our system what we have here, but of course we can also look at the sub micron features. So what do we need? We need a high resolution in all spatial dimensions. We need a high sensitivity, a high speed and of course the data become pretty large at some point. So by the end of the day, it's a trinity between resolution, speed and sensitivity, all
02:08:41
Thomas Meyer: Going on simultaneously. So about the Raman effect, the structural properties we learned already from Lyle. And I will skip this quickly. Okay. Photo look in this sense, don't be scared by this car. Rather complicated diagram. Photoluminescence actually is pretty easy. You need some energy to get an electron from the valence band into the conductive band and then. Depending on the type of semiconductor, you have more or less instant. Transition again to the valence band and then a photon is emitted, or you need to wait some time for some interaction, this is the so called indirect semiconductors, and then the electron goes down. The difference between those two types, direct and indirect semiconductors, is the lifetime basically. Not here. The direct part cap materials are gallium nitrite, gallium arsenide, and they are used for light emitting devices mainly.
02:10:00
Thomas Meyer: So this is what shines at you when you look at your smartphone. They have a high yield, so basically when you excite them they shine like a Christmas tree. The other bond gap materials are the indirect bond gap materials like silicon or germanium. Here it takes a while, then the photon can drop down. Basically, you need to have an incoming phonon. For the conservation of momentum. These materials don't shine, they are used for integrated circuits like. Microprocessors. Yep. So yeah, here we have a s example of silicon carbide. 4H silicon carbide has its bond gap around 390 electron volts, and the 3C silicon carbide around 530. So, and this difference in nanometers, of course, by the end of the day is a difference in energy. So with the we get the bond gap properties. So, Raman structural properties.
02:11:16
Thomas Meyer: PL band gap properties more or less in general. What is very important for us or one of our key features is the high confocality. A high confocality. Is achieved by the so-called pinhole. And basically we are only interested in the light coming from the focal plane and not from the out focal plane, out of focus plane. So when the light comes from here, you see that the light is focussed again back at the pinhole, and this light can nicely pass to to the detector. To the detector. If the light comes from out of plane, when we followed impath, we see the back focuses here. Focus and then the light is blocked by the pinhole. And this helps us when we compare this fluorescence image, a classical image, versus confocal. We see A higher or a much smaller background, a higher resolution. Here we even see the substructures in the nucleus. And of course with this free D information is accessible, which is not in a classical microscope.
02:12:40
Thomas Meyer: Of course, one can talk a lot about confocality and this stuff, but we'd like to prove our statements. So I want to show you two examples on how we measure our confocality. The point here is we need a sample which is very small compared to the the system function. So the physicist would say it has to resemble a delta function. That we can measure our samples. And here for the xy resolution, so the in plate, the lateral resolution. We measure a carbon nanotube. This is a quick image. And then we do a high quality measurement along this line. So, as you learned before for the EDS measurements, we can also do. Single single point measurements. Light measurements and the two D or three D measurements. And when we measure along this line and just apply apply a filter for the deep end cheap for the carbon nanotube, we get this distribution, and out of the full bit of half maximum, we can nicely determine the resolution in the lateral plane. So with the pre-laser, we are around 300 nanometers, or better to say below 300 nanometers.
02:14:07
Thomas Meyer: In the Z plane, we do a similar experiment. We have a silicon and then on the silicon, you see it here, there is something else on top. This is graphene. And it's a basically freestanding graphene. And when we measure there, we see also the typical spectrum for graphene. And then we do a line scan, but not in the horizontal plane, but from top to bottom. And then again we can just put a filter on the or sitting on the graffiti signal, G band or to D band. And then we get this distribution, and we see the full bit of half maximum. So here is our convocality. So another point what is important is the the sensitivity. And here we have a nice experiment. We look at the fourth order of silicon, which is a very, very, very weak signal. And when we look at our confocal volume, the lower part resembles the fourth order silicon, but not all of our volume is within the material.
02:15:23
Thomas Meyer: Cost silicon is basically non transparent, at least to the five thirty two nanometer laser. Then we have half of the signal coming from the air, which is the nitrogen. And here one fast. One measure is to compare the fourth order silicon versus the nitrogen, and we say we are in the same order. Better to say we are better. We have a higher silicon fourth order signal this is one of our key features in terms of sensitivity and confocality. So to summarize the first part, we have this chemical sensitivity. So our materials itself are the markers. So for example, for the polymers we don't have to stain them. For biosamples we don't have to put fluorescence dyes on them. The crystals, as we mentioned before, we see the phonon buttons. So this is already the material itself contains the information. We have the high confocal resolution, so three D images are possible. We have the high sensitivity and last but not least, we do not destroy the sample. Except we use a little too high laser power.
02:16:44
Thomas Meyer: We do carbonisation or, as the user once said to me, trial by fire. This all this combination of properties make Raman A nice tool or i ideal for investigation of semiconductors, 2D materials, polymers and of course. Okay, now let's dive into the full wafer research. Of course, when we want to look at full wafers, we need a larger stage then for coupons. So in this case we have our semiconductor addition. Which has 300 times 350 millimeter stage. We have a set range of one centimeter. Of course, when the samples are larger and it is required, we can just unmount the wafer chug and use it without the wafer chuck. The height for our samples. In this system, what we have here, we have two lasers equipped, at the three hundred fifty five UV laser and five thirty two nanometer laser. Combination with the high resolution Hexalyte 600. And last but not least, what's also a very nice feature is the true surface. True surface is a kind of profilometer in a single path technology. So while we're measuring Raman, we still get the
02:18:13
Thomas Meyer: The height profile of our sample. And this is especially helpful if the sample is not perfectly plane or if the sample shrinks during the measurement. We can always keep our measurement in the same distance ensemble. So, when we look at semiconductor sample. Six inch, a one hundred fifty millimeter wafer of silicon carbide. A very typical for the semiconductor industry. So what we see basically we see these three phases. So first of all we see the orange part which is the 4H silicon carbide, the epilayer. Then we see off with a higher doping concentration, which is the facet. And then the green pulp reassembles some defects mainly in the middle is here. And what you get for free. Is the topography.
02:19:30
Thomas Meyer: So the wafers they are not perfectly flat. In this case, 15 centimeters, we have a height difference of. Forty three microns. And we always want to have the sample the perfect distance to the objective, that we get always the nice signal. And this information we get for free, so for the customer. Also interesting to calculate the bow and the wall pinch of the so before we dive deeper, let me explain the spectrum a little. This E two mode is sensitive to stress and strain. So when we have stress and strain, the position of this bond changes. And the A one L O mode or the L O P C mode, this is sensitive to doping. And let's dive a little deeper into these two characteristics so when we look at the stress and strain of the material of the silicon carbide, we see a very typical pattern all over the wafer. So basically in the center of the wafer we see compressive stress and at the edge we see more tensile stress. This is due to the cooling when the wafer was produced.
02:20:49
Thomas Meyer: And this induces this internal stress field. One can c calculate it also on mega pascal because most people are not interested in a Raman shift but in measurement and make out possible. When we look at the full bit of half max of the peak, which is which resembles the crystallinity, we see a very homogeneous usual all over the wafer, except at this very edge here. And when you remember, there was the facet located. The facet: we have a higher amount of nitrogen atoms, for example, and this. Impairs the crystal structure. So at this point I expect a little higher or a little less crystallinity. And this is also what we see here. Still, it's a bit noisy, but we can can we sample it? So a second application on full waveform maps is the charge carrier concentration. And here the samples the producer of the wafer sent me the samples that was anti-doping, so with nitrogen.
02:22:08
Thomas Meyer: And a variation 10 to the power of 15 to 10 to the power of 17 atoms per, so nitrogen atoms per cubic centimeter. In this case I did the measurements thirty minutes per wafer, what can even go faster if necessary. And here I evaluated the A1 node as a measure for the charge carrier concentration. Funny story when I measured these wafers I thought I naive as I am as I am, I thought oh the the customer would have sent me some very homogeneous wafers. I did the evaluation, so oh no. When I discussed with the customer, they said, Yeah, we wanted to test you So I think I passed the test and out of this? Distributions you see we have the I can calculate curve, and then of course I can calculate the charge carrier concentration in every single pixel.
02:23:24
Thomas Meyer: If we have just the logarithmic scale, which is nice but magnitude is that one. This was the control measurement or the measurement done by the customer. And as you can see, they measured like 10 to 15 points all over the wafer. And that just created this nice image. This looks maybe a bit more noisy but just this. Let's say twenties points. Done in ten minutes. Here, this is 10,000 points done in 30 minutes. So if you want to enhance the speed, you just reduce the resolution a bit and then can easily surpass this. So yeah, and as you see we get the real distribution and not some strong interpolation. Okay, let's have a look on the defect. In this case you maybe remember from the first site of the full waveflow we saw this freeze oh this free C silicon carbide at the center. Which is resembling some defect. So it caught my attention. I moved there. And then, even in the white light image, I saw this.
02:24:45
Thomas Meyer: Triangular structures. So I was thinking, okay, what what can we do to investigate what this might be? Sure. Let's say modes of evaluation are accessible to us. First of all, the structural properties revealed by Raman and the bound gap properties revealed by photoluminescence sense. So first of all I did a 3D image just because I can. We are fast, we have a high sensitivity, so doing this high resolution mapping is easily possible. And what we learned from this is how this defect propagates. I think it comes more to visibility when we look at the three D reconstruction of the image. You nicely see how it grew. I talked to the producer of the wafer, and this is basically. Or what we see here the defect starts in the substrate. And then when the API layer is applied, you have some some kind of flow. And this will create this free diagram. That is triangular structure and you see also how this
02:26:03
Thomas Meyer: Propagates from top to bottom. Point is this causes some structural impairments. So the material there is not as. Or it doesn't have a structural homogeneity there, so the device out of this part might be might break even faster. Then from the other parts. This is not everything. We have the structural impairment, but also changes in the band gap properties. So our device will behave in a different manner compared to the device built from here. So what we see we see this changes in the band gap properties and the energies. And more, we see also some changes which are not following the structural impairment, the structural. Pattern. But here we see this other structure where we have not only the barn-edge emission but some internal relaxations and then an emission from another energy line. So even though we still see the same crystal structure, the material here behaves different.
02:27:20
Thomas Meyer: Then here to here. So also we have some differences in the behaviour. And this is also something nice, so this complementary methods you with one method you see this, and with the other method you see this. And putting both things together gives you the real view, like looking at the world with one eye and the other eye, and when you open both eyes, you have a 3D image. Okay let's go quickly to this firmer imaging. When he has a set up all the layers, we have our integrated circuit and then we build our devices. In case it's a gun high electron mobility transistor. It this basically better than the conventional silicon transistors and is smaller. And the key is now to understand the behavior of the device under operation load. And what I did here, I wanted to create thermomechanical map saw. I basically put it on a heating stage.
02:28:36
Thomas Meyer: And then turned up the temperature. And what we see, first of all, we see a peak shift to when we heat up, we see tensile stress as the material expands. And also we see the change in the behaviour of the stokes versus anti-stokes because of the higher temperature, the anti-stokes the intensity rises compared to the stokes. This is the formula. It looks rather difficult, but you just have to type it in. And then you can nicely measure the or you have the temperature and the aperture stage and peak shift again, I like to make calibration curves. Because creating this calibration curves brings Raman from a semi quantitative method into a quantitative realm. So we can really measure the temperature. Here I also did it with the anti-Stokes temperature calculation. Yes. What we get by the end of the day is a thermomechanical map where we can measure the temperature in every pixel.
02:29:57
Thomas Meyer: And remember this sixty times sixty three thousand six hundred spectra with a pretty decent signal to noise ratio were acquired in three minutes. Three minutes, thirty three seconds, to be honest. And then we get this thermal image. So this is now let's say the the real experiment of course is then putting the device under load. And then you see how the device heats up on this let's say on a spatial distribution. And you might see where the device might crack earlier or later. Even with advanced experiments you can trigger the measurement with with switching on and off the device, so you can get the heating costs. So you can really understand your device, how it acts when it's put under load. Loss ah here the two D materials I almost forgot. I will quickly about this. It's not a two D material, it's one D material, but that's but
02:31:14
Thomas Meyer: So basically what I'm looking at now is the radial breathing mode. So and this this is like the breathing of the tube. And as you can imagine, this briefing the frequency of the breathing is related to the diameter of the carbon nanotube. So with some Mathematics: we can follow the literature, we can calculate the diameter of the carbon nanotube. Which is the structural information. If we look at the G-band at the position and the height of the G-band, we can distinguish between. Semiconductive and metallic properties. So we can say this particular nanoplota. Carbon nanotube has a diameter of one nanometre and is metallic. Which is very useful when you want to put something when when you create something. Stuff out of it. 2D materials where we have the graphene, for example, exfoliated graphene, we want to know is it one layer, two layers or three layers?
02:32:35
Thomas Meyer: Can be easily distinguished by the ratio of the G to two D band. And yeah easy to distinguish between one, two or three or more layers. And of course we can have a look at the chirality, which is also important for the material properties. So there's another nice example of the graphene. This is an electron image, and of course, here we don't know is this already graphene or is this just the silicon substrate? What do we have where? So basically we see oh this is already one layer of graphene. But then we see even more faces. We see the two layers of graphene with a twisted angle. The twisted angle can be calculated out of the full width of half maximum and the precision of the 2D band. And again we see another angle. Here we see another peak arising and another twisted angle. So twist angle, not twisted angle. So again, here we see another feature.
02:33:54
Thomas Meyer: All this information combined, we get the real information. In this case, it's important to know how the mismatch is of the layers of the graphene. How they behave. 'Cause this mismatch of course changes the properties on how it acts. Okay, if we look at the which are promising in terms of getting the integrated circuits below the five nanometer scale. It's important to know are they one layer, two layers, three layers, because then they change behavior between direct and indirect semiconductors. Again, with Raman, this is possible because the shear mode is only visible in the and a multi-layer and also for P L. That the PL is much stronger where we have a single layer compared to two layers again. Direct was indirect enough arm cap. Yeah. If we look at the second harmonics Non-linear effect of higher harmonics, we can learn about the grain boundaries. So, how these crystals grew together.
02:35:15
Thomas Meyer: So with one method we get the complete image Raman. The photoluminescence, the band gap properties and the grain boundaries. And now, last but not least, let's watch a little video about polymerisation of resin varnish. Basically it's a transition of a C double bond into a C single bond. And here the ta the the the the customer thought the varnish would harden homogeneously. And when I did the experiment
02:35:56
Thomas Meyer: I see. How the hardening took place, which was pretty surprising for the customer because it's not homogeneous, but from top to bottom. By the end of the day they bought the system exactly because of this finding. Because it really contradicts their their expectations. Basically we had two varnishes here. One was fast, the other was slow. And we learned also that the the fast hardening varnish just stopped up after twenty four hours. The slow one continued, so it took longer, but the process went on and on and on. Okay, so that's quickly summarize. First of all, when we produce our wafers. We learned that we learned about the stress field and the doping concentration, which is important when we want to build further devices. Of course we also see the defect, so we already learn where our material might be impaired, where we don't where we cannot expect the devices to be fully functional afterwards.
02:37:22
Thomas Meyer: We can investigate the fully processed devices so we can learn how it operates under load. You can learn about the novel materials, the 2D material. This is more let's say basic research. The universities but hopefully it will make the way its way into industry. And then, last but not least, of course when the product, the semiconductor chip is finished, is produced, we want to implement it in something which we can use. So, for example, in a car, we want. But we need to learn also about the polymerization. So I think that's it. If you have any questions feel free to ask or find me in the labs in the afternoon.
