EDS

My New Lab Partner Part 2 (East Coast Edition)

Jens Rafaelsen, Applications Engineer, EDAX

During a recent trip to our Draper lab in Utah for a training class, I got a first-hand look at Matt’s new lab partner (https://edaxblog.com/2017/02/14/my-new-lab-partner/). I must admit that I am a little envious of his new microscope and how easily you get great looking images (even at low acceleration voltage or high beam current) compared to the systems we have in our Mahwah lab. However, I must also admit that he needed an upgrade a lot more than we did. While his old XL has been very reliable (and still seems to be, even after moving it to another room), it was always a bit of a worry conducting a training class with only one microscope available and one that was at end of service life at that.

Around the time when Matt got his new microscope we also had an addition to our Mahwah lab as seen in the picture below:

OK, it’s definitely not an ARM or a TITAN, it only goes to 120kV, it’s not quite as new and fancy as Matt’s microscope, and the firmware might read 1994 when you hit the ON button, but it’s still good to have a TEM in the building once again. One of the things that’s great about older scientific instruments is that they often include full vacuum and wiring diagrams, schematics, and troubleshooting directions. Not so great: pressure readings in arbitrary numbers… I did some creative plumbing and mounted extra gauges on the line of the microscope gauges so now I know that a pressure reading in the buffer tank of 26 corresponds to roughly 10-1 mbar and that the camera chamber goes down to the mid 10-5 mbar. As an added bonus, several people in the building have been around long enough to have experience with the CM12 both as users and service and have had their memories jogged for how to run and align it. This also spurred the comment: “That’s right, this is why I decided to get out of field service…”.

Having had very limited TEM experience it’s been a bit of a learning curve for me but I think it’s getting there. There’s still a lot to learn when it comes to fine tuning of the instrument, diffraction, and aligning for dark field imaging, but at least I am able to get bright field images at over 500k magnification without spending too much time. And some of the images actually have somewhat decent resolution and recognizable features at that:

Holey carbon at 660.000x magnification

Of course, a lot of what we do at EDAX doesn’t really require great resolution or the newest instruments. While it’s always nice to have pretty pictures to go along with things, the X-rays don’t really care much about your astigmatism or spot size (unless you are trying to map of course). But there’s a significant difference in what you see in your spectra whether your electrons are hitting the sample with 15 kV or 120 kV. There are also very different considerations and limitations between a SEM and a TEM when it comes to actually mounting the detector, designing collimators, and even what materials can be used. With that being said, I hope that with my “new” lab partner we will move things along so that we can show you new applications, software, and hardware specifically for the TEM in the near future.

Considerations for your New Year’s Resolutions from Dr. Pat

Dr. Patrick Camus, Director of Research and Innovation, EDAX

The beginning of the new calendar year is a time to reflect and evaluate important items in your life. At work, it might also be a time to evaluate the age and capabilities of the technical equipment in your lab. If you are a lucky employee, you may work in a newly refurbished lab where most of your equipment is less than 3 years old. If you are such a fortunate worker, the other colleagues in the field will be envious. They usually have equipment that is much more than 5 years old, some of it possibly dating from the last century!

Old Jalopy circa 1970 EDAX windowless Si(Li) detector circa early 70’s

In my case, at home my phone is 3 years old and my 3 vehicles are 18, 16, and 3 years old. We are definitely evaluating the household budget this year to upgrade the oldest automobile. We need to decide what are the highest priority items and which are not so important for our usage. It’s often important to sort through the different features offered and decide what’s most relevant … whether that’s at home or in the lab.

Octane Elite Silicon Drift Detector 2017 Dr. Pat’s Possible New Vehicle 2017

If your lab equipment is older than your vehicles, you need to determine whether the latest generation of equipment will improve either your throughput or the quality of your work. The latest generations of EDAX equipment can enormously speed up throughput and the improve quality of your analysis over that of previous generations – it’s just a matter of convincing your boss that this has value for the company. There is no time like the present for you to gather your arguments into a proposal to get the budget for the new generation of equipment that will benefit both you and the company.
Best of luck in the new year!

Adding a New Dimension to Analysis

Dr. Oleg Lourie, Regional Manager A/P, EDAX

With every dimension, we add to the volume of data, we believe that we add a new perspective in our understanding and interpretation of the data. In microanalysis adding space or time dimensionality has led to the development of 3D compositional tomography and dynamic or in situ compositional experiments. 3D compositional tomography or 3D EDS is developing rapidly and getting wider acceptance, although it still presents challenges such as the photon absorption, associated with sample thickness and time consuming acquisition process, which requires a high level of stability, especially for TEM microscopes. After setting up a multi hour experiment in a TEM to gain a 3D compositional EDS map, one may wonder Is there any shortcut to getting a ‘quick’ glimpse into 3-dimensional elemental distribution? The good news is that there is one and compared to tilt series tomography, it can be a ‘snapshot’ type of the 3D EDS map.

3D distribution of Nd in steel.

3D distribution of Nd in steel.

To enable such 3D EDS mapping on the conceptual level we would need at least two identical 2D TEM EDS maps acquired with photons having different energy – so you can slide along the energy axis (adding a new dimension?) and use photon absorption as a natural yardstick to probe the element distribution along the X-ray path. Since the characteristic X-rays have discrete energies (K, L, M lines), it might work if you subtract the K line map from the L line or M line map to see an element distribution based on different absorption between K and L or M line maps. Ideally, one of EDS maps should be acquired with high energy X-rays, such as K lines for high atomic number elements, and another with low energy X-rays where the absorption has a significant effect, such as for example M lines. Indeed, in the case of elements with a high atomic number, the energies for K lines area ranged in tens of keV having virtually 0 absorption even in a thick TEM sample.

So, it all looks quite promising except for one important detail – current SDDs have the absorption efficiency for high energy photons close to actual 0. Even if you made your SDD sensor as large 150 mm2 it would still be 0. Increasing it to 200 mm2 would keep it steady close to 0. So, having a large silicon sensor for EDS does not seem to matter, what matters is the absorption properties of the sensor material. Here we add a material selection dimension to generate a new perspective for 3D EDS. And indeed, when we selected a CdTe EDS sensor we would able to acquire X-rays with the energies up to 100 keV or more.

To summarize, using a CdTe sensor will open an opportunity for a ‘snapshot’ 3D EDS technique, which can add more insight about elemental volume distribution, sample topography and will not be limited by a sample thickness. It would clearly be more practical for elements with high atomic numbers. Although it might be utilized for a wide yet selected range of samples, this concept could be a complementary and fast (!) alternative to 3D EDS tomography.

Help!

Dr. René de Kloe, Applications Specialist EDS, EBSD, EDAX

The job of an applications engineer is to help people. Help sales people to explain to customers what a system can do. Help customers to get the most out of their system and help them to understand their materials better. Help the marketing group with nice examples. And help the development team to devise applications that have not been tried before.

One thing you need in order to be able to help is knowing the EDAX analysis systems inside-out. But the other thing you need is samples. Lots of samples. Every function or analysis tool in the software, regardless if it is for EDS, EBSD, WDS, or XRF is best shown with a specific material or combination of elements or phases. Some of these, like chemical standards with known composition, you have to make or perhaps buy. Others you have to collect yourselves, but from where? A great source for new materials are our customers. People often send me materials to evaluate our systems, or for help on how to best analyse their samples. When I then get permission to keep a bit of the material it goes directly into my collection, together with valuable information on the current analysis requirements in different scientific disciplines.

Eight phase FeSi alloy Brass with NiMnSi particles 

This goes a long way in getting good example materials, but I always keep my eyes open for new interesting things. When I see a metal strip in an anti-theft label in clothing I keep it (after buying the item of course), when a droplet of lead-tin solder falls on the floor, I stick it in the microscope to see if it looks good. I also scrutinize things that get thrown away, ranging from the lid of a vegetable jar to a damaged bellows of an EBSD system. That has given me beautiful cast aluminium samples for EDS mapping, multiphase brass alloys for ChI-Scan EDS-EBSD analysis, and recently an unexpected copper-plated zinc-aluminium-silicon alloy for EBSD phase identification from a broken belt buckle.

Grain structure of a staple Grain structure of a key ring 

Luckily I don’t always have to go dumpster diving to get my example materials. One of my favorite sample mounts contains different types of heavily deformed ferrite, duplex stainless steel, and also martensitic structures. That sounds perhaps complicated, but on the outside the same sample just looks like staples, a paperclip, a key ring, and a screw.

The screw, for example, I polished after doing some DIY work at home and because a certain type of screw kept breaking off when I tightened it, I wanted to take a close look why that happened. It turned out that there were lots of small cracks along the thread, which then also lined up with trails of carbides further inside the screw. That turned out to be a really bad combination and when you tighten the screw, the cracks propagate, connect with the carbide trails and the screw head snaps off. The replacement screws that I used instead had a much finer structure without any cracks and that is what is still holding things together in the house. This shows how microstructures literally shape our daily life. And it also provides a beautiful example to help illustrate the importance of microstructural characterization to new EBSD users.

Weak screw Strong screw

The huge variation in materials and microstructures makes the collection of demonstration samples the most important tool for an application scientist and from this place I hereby want to thank all people who have given me a piece of some material during my years at EDAX to use to help others.

By the way, I would appreciate it very much if the person who briefly “borrowed” my marble sample last year gives it back soon …

Metals, Minerals and Gunshot Residue

Dr. Jens Rafaelsen, Applications Engineer EDAX

During a recent visit to a customer facility I was asked what kind of samples, and applications I typically see. It would seem that this would be a pretty easy question to answer but I struggled to narrow it down to anything “typical”. Over the past three weeks I have spent a couple of days each week at customer facilities and I think a brief description of each of them will explain why I had a hard time answering the question.

The first facility I went to was a university in the process of qualifying an integrated EDS/EBSD system on a combined focused ion beam (FIB) and scanning electron microscope (SEM). A system like this allows one to remove material layer by layer and reconstruct a full 3D model of the sample. The dataset in Figure 1 illustrates why this information can be crucial when calculating material properties based on the grain structure from an EBSD scan. If one looks at the image on the left in the figure, it seems obvious that there are a few large grains in the sample with the area between them filled by smaller grains. However, the reconstructed grain on the right shows that several of these smaller “grains” seen in the single slice are actually interconnected and form a very large grain that stretches outside the probed volume.

Figure 1: Single slice EBSD map (left) and single reconstructed grain from 3D slice set (right).

Figure 1: Single slice EBSD map (left) and single reconstructed grain from 3D slice set (right).

While we spent a good amount of time documenting exactly what kind of speed, signal-to-noise, resolution and sensitivity we could get out of the system, one of the customer’s goals was to measure strain to use as a basis for material modelling. We also discussed a potential collaboration since our EBSD applications engineer, Shawn Wallace, has access to meteorite samples through his previous position at the American Museum of Natural History in New York and a 3D measurement of the grains in a meteorite could make a very compelling study.

Next up was a government agency where the user’s primary interest was in mineral samples but also slag and biological materials retained in mineral matrices. Besides the SEM with EDS they had a microprobe in the next room and they would often investigate the samples in the SEM first before going to the microprobe for detailed analysis (when and if this is required is a different discussion, I would recommend Dale Newbury and Nicholas Ritchie’s article for more details: J Mater Sci (2015) 50:493–518 DOI 10.1007/s10853-014-8685-2).

A typical workflow would be to map out an area of the sample and identify the different phases present to calculate the area fraction and total composition. Since the users of the facility work with minerals all the time, they could easily identify the different parts of the sample by looking at the spectra and quantification numbers, but I have a physics background and will readily admit that I would be hard pressed to tell the difference between bustamite and olivine without the help of Google or a reference book. However, this specific system had the spectrum matching option, which eliminates a lot of the digging in books and finding the right composition. The workflow is illustrated in Figure 2, where one starts by collecting a SEM image of the area of interest and, when the EDS map is subsequently collected, the software will automatically identify areas with similar composition and assign colors accordingly. The next step would then be to extract the spectrum from one area and match it up against a database of spectra. As we can see in the spectrum of Figure 2, the red phase of the EDS map corresponds to a obsidian matrix with slightly elevated Na, Al, and Ca contributions relative to the standard.
Figure 2a

Figure 2: Backscatter electron image (top left) and corresponding phase map (top right) showing different compositions in the sample. The bottom spectrum corresponds to the red phase and has an obsidian spectrum overlaid.

Figure 2: Backscatter electron image (top left) and corresponding phase map (top right) showing different compositions in the sample. The bottom spectrum corresponds to the red phase and has an obsidian spectrum overlaid.

The last facility I visited was a forensic lab, where they had purchased an EDS system primarily for gunshot residue (GSR) detection. The samples are usually standard 12.7 mm round aluminum stubs with carbon tabs. The sticky carbon tabs are used to collect a sample from the hands of a suspect, carbon coated and then loaded into the SEM. The challenge is now to locate particles that are consistent with gunshot residue amongst all the other stuff there might be on the sample. The criteria are that the particle has to contain antimony, barium and lead, at least for traditional gunpowder. Lead free gunpowder is available but it is significantly more expensive and when asking how often it is seen in the labs, I was told that apparently the criminal element is price conscious and not particularly environmentally friendly!

The big challenge with GSR is that the software has to search through the entire stub, separate carbon tape from particles down to less than 1 micron, and then investigate whether a particle is consistent with GSR based on the composition. The workflow is illustrated in Figure 3 and is done by collecting high resolution images, looking for particles based on greyscale value in the image, collecting a spectrum from each particle and then classifying the particle based on composition. Once the data is collected, the user can go in and review the total number of particles and specifically look for GSR particles, relocate them on the sample, and collect high resolution images and spectra for documentation in a potential trial.

Figure 3: Overview showing the fields collected from the full sample stub (top left), zoomed image corresponding to the red square in the overview image (top right) and gunshot residue particle from the red square in the zoomed image (bottom).

Figure 3: Overview showing the fields collected from the full sample stub (top left), zoomed image corresponding to the red square in the overview image (top right) and gunshot residue particle from the red square in the zoomed image (bottom).

Three weeks, three very different applications and a very long answer to the question of what kind of samples and applications I typically see. Each of these three applications is typical in its own way although they have little in common. This brings us to the bottom line: most of the samples and applications we come by might be based in the same technique but often the specifics are unique and I guess the uniqueness is really what is typical.

From Intern to Analyst – Studying the Impact of ‘Non-Ideal’ Samples on Quant Results

Kylie Simpson and Robert Rosenthal, 2016 Summer Interns at EDAX

Being surrounded by equipment worth more than your average college student can even fathom is incredibly daunting. Your heart still skips a beat at every hiss or beep that the microscope produces. Not to mention the fear of ramming into the pole piece while inserting the EDS detector (we later learned there was a hard stop to prevent this but it never quite seemed to alleviate the fear). It’s hard to summarize all of the experiences from our internship at EDAX this summer. While it was only about two and a half months, the sheer amount knowledge we gained through hands on experience is unquantifiable. The five day EDS training course in itself contained enough information to be taught over an entire college semester.

Working with the Applications team gave us a real feel for what EDAX is all about. Not only did we get to work on a summer-long project, we also got to work with the marketing, engineering, and software teams on a regular basis. We also helped with support for the new APEX software. This work setting provided us with a plethora of new knowledge, not only of the physics and programming behind EDAX software but also of the inner workings of the company and the crucial role that teamwork plays in accomplishing tasks. Having access to an electron microscope as well as the specialized knowledge of the members of the Applications team enabled us to get the most out of our summer here at EDAX. After sitting in on a meeting with other members of the Applications team, we were exposed to some of the real-world problems faced by customers on a regular basis and decided to investigate this further with our summer project.

When collecting quantification results for EDS, the ZAF matrix corrections are based on the assumption that the sample is flat, homogeneous, and infinitely thick to the electron beam. Although these are the ideal collection requirements, many customers run into problems when their samples do not meet these assumptions. We spent our time here testing the impact of ‘non-ideal’ samples on quant results while also determining ways for customers to improve the accuracy of quant results with these samples. We tested samples with rough topography by scratching up and polishing a stainless steel and a pyrite sample (Figure 1). By collecting a counts per second map for the steel (Figure 2), we were able to visualize the impact of rough samples and confirm the need for sample prep.

Figure 1. Pyrite particles and polished pyrite Figure 2. CPS maps of stainless steel surfaces

We also tested inhomogeneous samples, including a Lead-Tin solder sample and a stainless steel sample (pictured below). By collecting spectra of these samples at different magnifications, we observed the correlation between lower magnification and a higher accuracy of quant results.

Figure 3: Lead-Tin solder and stainless steel samples

Figure 3: Lead-Tin solder and stainless steel samples

Finally, we tested the impact of thin samples on quant results using an aluminum coated piece of silicon. This sample was very hard to obtain, being that we had to coat the silicon five separate times, but it yielded very interesting results (see graph (left) in Figure 4 below). Our results illustrated the influence and importance of collecting spectra while also allowing us to back-calculate the thickness of each aluminum layer (pictured in Figure 4 (right) below).

Figure 4.

Figure 4.

Overall, we thoroughly enjoyed our summer at EDAX and will take away not only knowledge of EDS, EBSD, SEMs, computer programming, and teamwork, but also valuable problem solving skills applicable to classes, professions, and other real-world scenarios that we will encounter in the future.

Meet the Interns

Kylie Simpson: Kylie is currently a student at the Thayer School of Engineering at Dartmouth. She is participating in a duel-degree program with Colby College and Dartmouth College and is studying mechanical engineering and physics.

Robert Rosenthal: Robbie is currently a student at the University of Colorado at Boulder. He in going into his junior year studying Mechanical Engineering.

“It’s not the size of the dog in the fight, it’s the size of the fight in the dog.” (Mark Twain)

Dr. Oleg Lourie, Senior Product Manager, EDAX

San Javier, Spain, October 18, 2015: Airbus A400M airlifter escorted by Sains Patulla Aguila squad on their 30th anniversary celebration event.

Many of us like to travel and some people are fascinated by the view of gigantic A380’ planes slowly navigating on tarmac with projected gracious and powerful determination. I too could not overcome a feel of fascination every time I observed these magnificent planes, they are really – literally big..  The airline industry however seems to have a more practical perspective on this matter – the volume of the A380s purchase is on decline and according to the recent reports Airbus is considering reducing their production based on growing preference towards smaller and faster airplanes. Although the connection may seem slightly tenuous,  in my mind I see a fairly close analogy to this situation in EDS market, when the discussion comes to the size of EDS sensors.

In modern microanalysis where the studies of a compositional structure rapidly become dependent on a time scale, the use of the large sensors can no longer be a single solution to optimize the signal. The energy resolution of an EDS spectrometer can be related to its signal detection capability, which determines the signal to noise ratio and as a result the energy resolution of the detector. Fundamentally, to increase signal to noise ratio one may choose to increase signal, or number of counts, or as alternative to reduce the noise of the detector electronics and improve its sensitivity. The first methodology, based on larger number of counts, is directly related to the amount of input X-rays determined by a solid angle of the detector, and/or the acquisition time. A good example for this approach would be a large SDD sensor operating at long shaping times. A conceptually alternative methodology, would be to employ a sensor with a) reduced electronics noise; and b) having higher efficiency in X-ray transmission, which implies less X-ray losses in transit from sample to the recorded signal in the spectra.

Using this methodology signal to noise ratio can be increased with a smaller sensor having higher transmissivity and operating at higher count rates vs larger sensor operating at lower count rates.

To understand the advantage of using a small sensor at higher count rates we can review a simple operation model for SDD.  A time for a drift of the charge generated by X-ray in Si body of the sensor can be modeled either based on a simple linear trajectory or a random walk model. In both cases, we would arrive to approximate l~√t dependence, where l is the distance traveled by charge from cathode to anode and t is the drift time. In regard to the sensor size this means that a time to collect charge from a single X-ray event is proportional to the sensor area. As an example, a simple calculation with assumed electron mobility of 1500 cm2/V-1s and bias 200 V results in 1 µs drift time estimate for 100 mm2 and 100 ns drift time for 10 mm2 sensors. This implies that in order to collect a full charge in a large sensor the rise time for preamplifier needs to be in the range of 1 µs vs 100 ns rise time that can be used with 10 mm2 sensor.  With 10 times higher readout frequency for 10 mm2 sensor it will collect equivalent signal to a 100 mm2 sensor.

What will happen if we run a large sensor at the high count rates? Let’s assume that a 100mm2 sensor in this example can utilize the 100 ns rise time. In this case, since the rise time is much shorter than the charge drift time (~1 µs), not all electrons, produced by an X-ray event, will be collected. This shortage will result in an incomplete charge collection effect (ICC), which will be introducing artifacts and deteriorating the energy resolution. A single characteristic X-ray for Cu (L) and Cu Kα will generate around 245 and 2115 electrons respectively in Si, which will drift to anode, forced by applied bias, in quite large electron packets.  Such large electron packets are rapidly expanding during the drift with ultimately linear expansion rate vs drift time. If the rise time used to collect the electron packet is too short, some of the electrons in the packet will be ‘left out’ which will result in less accurate charge counting and consequently less accurate readout of the X-ray energy. This artifact, called a ‘ballistic deficit’ (BD), will be negatively affecting the energy resolution at high count rates. It is important to note that both ICC and BD effects for the large sensors are getting more enhanced with increasing energy of the characteristic X-rays, which means the resolution stability will deteriorate even more rapidly for higher Z elements compare to the low energy/light elements range.

Figure 1: Comparative Resolution at MnKa (eV).

Figure 1: Comparative Resolution at MnKα (eV) *

As the factual illustration to this topic, the actual SDD performance for sensors with different areas is shown in the Fig. 1. It displays the effect of the acquisition rates on the energy resolution for the EDS detectors having different sensors size and electronics design. Two clear trends can be observed – a rapid energy resolution deterioration with increase of the sensor size for the traditional electronics design; and much more stable resolution performance at high count rates for the sensor with new CMOS based electronics. In particular, the data for Elite Plus with 30 mm2 sensor shows stable resolution below 0.96 µs shaping time, which corresponds to >200 kcps OCR.

In conclusion, conceptually, employing a smaller sensor with optimized signal collection efficiency at higher count rates does offer an attractive alternative to acquiring the X-ray signal matching the one from large area sensors, yet combined with high throughput and improved energy resolution. Ultimately, the ideal solution for low flux applications will be a combination of several smaller sensors arranged in an array, which will combine all the benefits of smaller geometry, higher count rates, higher transmissivity and maximized solid angle.

* SDD performance data courtesy of the EDAX Applications Team.