Dr. Shangshang Mu, Applications Engineer, EDAX
I got into the EDS world about 10 years ago, when I started my PhD study at Boston University. In my first project, I needed to quantify the elemental composition of my experimental samples. My advisor told me that ideally, we should use an electron microprobe and that the nearest one was at MIT, which is literally on the other side of the Charles River. But after we heard of the hourly fee and estimated the number of samples I would need to analyze, we started to plan an alternative. We found out that there was a field emission SEM equipped with an EDS detector in our own Photonics Center, just across the street and most importantly it was free of charge. At that time, I knew neither microprobe nor EDS and decided to give EDS a try first. Even if it did not work, I had nothing to lose except wasting some time. Graduate students have plenty of time to waste. So finally, I didn’t cross the beautiful Charles River but Commonwealth Avenue for the EDS.
As a result, I was introduced to the EDS world by the EDAX Apollo 40 SDD detector and EDAX was the only EDS manufacturer I knew as a beginner. It is such a good brand name standing for Energy Dispersive Analysis of X-rays so in the first couple of years I was under the impression that it was the term of this technique. For quantitative analysis, I used known standards that were close to my experimental samples in composition to standardize and got pretty good results fast and easy. In the subsequent projects, I collected a lot of EDS maps and the Apollo 40 never disappointed me. Since the EDS detector worked well for my PhD projects, I no longer considered other techniques. By the way, although I missed the opportunity to learn how to use the microprobe across the river due to the budget issue, my first full-time job in the states was to manage a much fancier microprobe acquired by a former user of the MIT’s microprobe. He received his PhD in geochemistry from MIT and I got most of my microprobe skills from him.
As an entry level EDAX user in graduate school, I had no way to imagine that I turned myself into an EDAX applications engineer and right now I am celebrating my one-year anniversary. After I joined EDAX, I got to know that the Apollo was the first generation SDD detector series of EDAX and our current EDS detectors have much better all-around performance. At the beginning of my second year with EDAX, I looked back into the EDS data I collected in graduate school and noticed that they were collected at a much slower amp time (the time the detector processes one X-ray count) compared to current ones and the EDS maps looked kind of noisy in comparison to my current perspective. Prompted by these findings, I wanted to initiate the discussion with how advancements in detector technology shaped the way we use EDS.
I believe all EDS users have heard of this rule of thumb: keep the dead time between 20% and 40%, or something like this. At least I was taught to keep the percentage within this range and as far as I know a lot of EDS users are following this rule. This is a traditional perspective that came out in the past when detector amp times were much slower. The dead time is all about throughput and affected by the amp time. Historically, if the dead time is below 20%, it means the detector either doesn’t receive enough X-ray counts per second to ensure high data quality or the amp time is too fast to maintain an optimal detector resolution. On the other hand, if the dead time is over 40%, the Input Counts Per Second (ICPS) is too high to be handled optimally by the current amp time and may lead to excessive summed peaks. We can get the dead time back in the range by either decreasing the beam current to lower the count rate or choosing a faster amp time.
In the past, we usually did not consider the second option since it would sacrifice the detector resolution a lot for throughput. However, the current generation of EDAX EDS detectors is equipped with CMOS based pre-amplifiers that allow much faster amp times ranging from 0.12 μs to 7.68 μs, while keeping very good resolution and high throughput. For example, if I have the EDAX Octane Elite detector in my lab, I can run eight times faster by moving the amp time from the slowest 7.68 μs to the intermediate 0.96 μs, the resolution is only decreased by roughly 4 eV (from 124 eV to 128 eV). This intermediate amp time can handle at least 80% of the job, however under this condition it is hard to make the dead time go over 20% unless the detector is receiving over 100K ICPS. Even if I use the fastest amp time at 0.12 μs, the resolution is still below 150 eV, which is not a significant decrease in resolution if you know that the detector resolution for those equipped with traditional JFET pre-amplifiers is about 250 eV at this fastest amp time. Since resolution is no longer a limiting factor, feel free to open your aperture to increase the count rate and choose a faster amp time to lower the dead time. In return, you will get a higher throughput, which means more statistics and higher quality of data. Although we can go below 20% dead time to have a throughput improvement, it still makes sense to apply the 40% upper bound since the detector will not convert X-rays efficiently once the dead time is beyond this maximum percentage.
When do we want to hit hundreds of thousands of ICPS and take advantage of the fast amp time and high throughput brought by current EDS detectors? While I was using the Apollo 40 at Boston University to collect EDS maps, I stuck to 12.8 μs amp time believing the higher the detector resolution, the better the map quality. Now, I have realized that the major limitation to the quality of EDS maps is not the detector resolution but the limited statistics at the pixel level. Not to mention for our current detectors, the degradation in resolution when running fast is very little. The quality of peak deconvolution is primarily determined by the level of statistics, even when dealing with tricky peak overlaps.
I did a quick study on a piece of floor tile in my lab that contains both calcium phosphate and zirconium silicate, so P and Zr are in distinct phases. P K and Zr L peaks are heavily overlapped with only 29 eV of energy difference. I used the Octane Elite detector to do a quick five-minute net intensity map on this floor tile at 26 nA beam current and 0.96 μs amp time. This combination gave me 160K ICPS and 28% dead time. For the purpose of comparison, I used the slowest amp time at 7.68 μs to yield the highest detector resolution and recollected the maps for 6.5 minutes. To keep the dead time at 28% I had to lower the beam current to 3.2 nA to constrain the ICPS at 20K. Obviously, in the first run Zr and P were separated out nicely in the sharp images (Figure 2 left), as it built up eight times more statistics than the second run in roughly the same amount of time. In the second run at the highest detector resolution, the separation was not quite as good (Figure 2 right). As we can see, the images are kind of pixelated and the coral pixels are mixed within the green in the overlay, so the slightly better detector resolution did not help at all. If I were able to know this trick as a rookie, I would be able to get higher quality maps in the same amount of time or get the same quality of maps faster.
Figure 2. Net intensity maps of P/Zr overlay and P collected at 0.96 μs amp time (left) and 7.68 μs amp time (right) for roughly 5 minutes.
From my point of view, the advancements in detector technology, the experience we gain, or a combination of the two change the way we use EDS.