With over 1000 registrants and 2500 viewings, our webinar with Edax on the Past, present and future - the evolution of x-ray analysis was another highly popular event. Due to its popularity, we are publishing the Q&A from the event.

Once again, thank you to Jens Rafaelsen and Tara Nylese for such a fantastic webinar. It is available to watch on demand here.

Question 1:
In the past time around 20 years ago, I knew that the EDX technique can not quantify the C, H, N, O, B contents in steel. This is because of the X-ray background interference. However, I have seen in some reports they use EDX to quantify those elements in steel. Can the EDX technique today quantify those elements today? How? I have asked many users as well as a few authors they could not clarify these questions? Which company develops this scientific acceptable technique? How they calibrate it? Could you please explain?

Modern EDS systems are quite capable of identifying and quantifying light elements, typically down to Beryllium. Previously Beryllium windows were used to protect the detector and this prevented the transmission of low energy x-rays, but with polymer windows and now silicon nitride windows, response in the light element range is not much of an issue. The quantification of light elements is similar to quantification of other elements, but more subject to matrix correction factors, particularly absorption of the low energy x-rays in the sample. Also, due to the low energy of the x-rays involved, correct charge collection, electronic noise and pulse processing becomes very important. The models used for the quantification is typically based on corrections for elemental x-ray yield, backscatter loss, fluorescence and absorption in the material. All EDS manufacturers have their own quantification routines, but they are mostly based on published scientific papers and data.
For a standard less analysis, no calibration is needed besides the normal voltage-energy calibration of the detector system. While it is no problem to do standard less quantification on light elements, standards can be used to ensure maximum quality and reproducibility. For steel in particular the ANSI 316 standard is often used.

Question 2:
What is the importance of Be window of the detector of EDX instrument? How can I fix or replace the Be window if it is destroyed?

The window protects the EDS detector from contamination and maintains a vacuum within the detector module. While windowless detectors are available, one must be extremely careful to protect the detector from damage and contamination and be sure that it is at ambient temperature before venting to prevent condensation on the detector.
A Beryllium window is not user replaceable. Beryllium dust is toxic if inhaled, and even if a new window is available, a vacuum must still be established inside the detector module prior to sealing.

Question 3:
I would like to know if it is possible to calculate the porosity of materials.

While it is in principle possible (though not necessarily easy) to extract information about the density of a material through EDS, it would not be possible to distinguish whether changes in density was caused by open or closed porosity. If the pores of the material vary by gray level, image analysis can be done to calculate the area fraction of the pores in a field of view.

Question 4:
The use of x-rays on a small scale.

As we showed in the presentation, it is quite possible to extract x-ray information from areas/volumes on the nm scale. Simply decreasing the SEM acceleration voltage will limit the volume x-rays are being generated in, as can be confirmed by electron trajectory simulation. But since the x-ray absorption is matrix and element dependent, there is no simple answer to the spatial resolution of the x-ray image; it will be different for different elements in the same sample and if the sample matrix changes in one area, so will the resolution.

Question 5:
I am puzzled if can it use or how to use Gaussian FWHM and FWHM_RT , the full widths at half maximum, to character the dislocation density in martensite steel.

By far the largest contribution to the FWHM of EDS peaks comes from the detector itself. For dislocation and crystallographic measurements, EDS is often combined with Electron BackScatter Diffraction (EBSD) measurements. Since both techniques can be run simultaneously in the SEM, compositional and crystallographic information can be recorded and correlated in a single scan.

Question 6:
How to use X-ray Analysis for the purpose of detecting lighter elements (lower than sodium) convincingly?

With modern x-ray window technology light element sensitivity is not a significant problem and detection down to Be can routinely be achieved.

Question 7:
Why electron hit K shell electron of Sulfur?

With a very high number of electrons striking the sample in a confined area, the chance of an electron from the e-beam striking a bound electron becomes a simple matter of probability.

Question 8:
Why K consists of two lines?

The atomic configuration of Sulfur has 2 electrons in the K shell, 8 in the L shell, and 6 in the M shell. If an electron in the K shell is removed, electrons from either the L or the M shell can fill the void. This leads to two emission lines if a K shell electron is removed, the L to K transition (K?) and the M to K transition (K?). Since there are two different electron orbitals in the L shell for which the transition is allowed, the L to K transition will have two different energy lines (K?1 and K?2) though these are very close together.

Question 9:
How can you get the SEM image at 30 V if energy of secondary electrons is about 50 eV?

Secondary electrons are defined as electrons generated by the primary electron beam and there is not a stringent energy cut-off as such. In a typical detection setup, electrons with energies below 50 eV are used for imaging, but this can be adjusted by changing the bias on the detector grid (depending on detector and setup). Though only a limited number of secondary electrons will be generated at low acceleration voltages advances in detection techniques has enabled imaging at these very low voltage conditions.

Question 10:
Is detector registrates any X-ray foton or several at time?

While multiple x-rays can enter the detector within one pulse processor time window, the counts in the spectrum originate from individual x-ray events. The detection time difference is on the order of nanoseconds. If multiple x-rays events occur, the electronics cannot differentiate which electrons were generated from which x-ray and will therefore be unable to correctly measure the x-ray energy. To avoid erroneous data, the pulse processor electronics will reject these measurements, which leads to increased detector dead time.

Question 11:
How many electrons do we need for one X-ray photon to generate?

The correlation between number of x-rays and number of electrons depends on several factors including the energy of the electron and the composition of the sample. The higher the energy of the incoming electron, the more x-ray generating collision it can be involved in before losing its energy. Consequently, if the binding energy of the atom (energy required to generate an x-ray) increases, the total number x-ray generating collisions the incoming electron can generate will reduce. One must also take into consideration that the ejected electrons from an x-ray event also can cause new x-ray events and that some of the incoming electrons will be lost to back-scatter events. So the number of x-ray events generated by one incoming electron must be calculated for a specific energy and sample.

Question 12:
When I go through to measure a particle present in a multi-component alloy, if I've known what the kind of this particle (elemental constituents) and I like to make sure if this composition is right, should I consider all elements that constitute the alloy or just consider the elements that are forming elements of this particle?

The EDS models for quantification typically assume a perfectly flat, homogenous sample that is infinitely thick from the point of view of the electron beam, but these conditions are often not met when analyzing particles or inclusions. The perfectly flat requirement can be negated to some extent by using a peak-to-background based quantification method, but the issue is often the homogenous and infinitely thick requirements. Depending on the composition of the particle and the electron energy, the interaction volume of the electrons in the sample can be several microns. There are several programs available to model the interaction volume, but only way of reducing it is to lower the electron energy/acceleration voltage. One approach to quantification of inclusions would be to lower the electron energy to the extent that the interaction volume is smaller than the particle, in which case only the elements in the particle is detected. If this is not possible, one can include the elements of the alloy and recalculate the atomic or weight percentage without these elements. However, under these conditions, the numbers will have some deviations from the true particle composition, as the matrix corrections factors are not well defined.

Question 13:
Do all EDX-Detector producers use products from the same SDD-chip-manufacturer or are there any significant differences in the production of these chips?

While the number of SDD-chip-manufacturers is limited, there are differences between both the chips and the pulse processing electronics that the chip interfaces to. EDAX recently started collaboration with Amptek in developing the next generation of SDDs and these new chips are currently being used in our Element product line.

Question 14:
What general condition (high/low voltage, large/small spot size etc.) should I consider to analyze light elements like carbon and boron?

For light element analysis it is usually beneficial to use an acceleration voltage in the 5-10 kV range, and sometimes even lower. Since the light element x-rays are low in energy, they are more readily absorbed than the heavier element x-rays. At higher acceleration voltages, a large interaction volume is excited, but since the low energy x-rays are absorbed they can only escape from a relatively shallow region close to the surface, while high energy x-rays can escape from a much larger region. Therefore we will see relatively more counts on the heavier elements with higher acceleration voltage. Tilting the sample towards the detector will shorten the x-ray escape path and increase light element escape and intensity in the spectrum. When analyzing light elements, it is also important to have good statistics, so a high beam current/large spot size is desirable as long as there is no beam damage to the sample and the detector dead-time is reasonable.

Question 15:
How to overcome peaks of different elements that merge together?

Some peak overlaps are inherent to EDS since typical detector resolution is larger than 120 eV at Mn K-alpha. The S-Mo or S-Pb overlaps are typical examples. The S K-alpha line is at 2.308 keV while Mo L-alpha is 15 eV lower at 2.293 keV and Pb M-alpha is 35 eV higher at 2.343 keV. While these lines are all found very close, they can often be separated through the deconvolution routines that are involved in the quantification algorithms. Most EDS software will report an error rate of deviation number that indicates to which extent the elements are present in statistically significant concentrations. The success of the deconvolution depends on the spectrum quality (more counts/better statistics is always a help) and the height of the peaks relative to the background in the region. However, in some cases it is impossible to separate the peaks using EDS and alternative techniques must be applied. An often used technique is WDS, which can be fully integrated with EDS in the EDAX systems.

Question 16:
What is the accuracy of chemical composition analysis of metallic alloys? Is 5%, 1% or 0.1 % accuracy achievable? Let us say we have like NiAlFe alloy or CuZnBe alloy, How precisely the chemical composition can be established?

The accuracy is highly dependent on sample and microscope conditions. Typically an element is considered as detectable once the peak is above the background with a roughly 95% confidence level. That is, at the detection limit the peak height is comparable to the statistical noise in the background and the error rate of the measurement is close to 100%. As the peak height/counts increases, the accuracy will increase as well and when taking all statistical and instrumental variations into account, the error rate will tend towards roughly 3% (at one standard deviation) unless standards are applied. However, it might not be practical to acquire the spectrum long enough to reach this limit.
For a sample with 30 wt% Ni, 20 wt% Al, and 50 wt% Fe and a simulated spectrum using a detector resolution of 123 eV, acceleration voltage of 25 kV, and acquisition time of 25 minutes at 10,000 counts per second, the minimum detection limits are around 0.04 wt% for all elements if using the K-lines. For a C82800 Beryllium Copper sample with 96.6 wt% Cu, 2.6 wt% Be, 0.5 wt% Co, and 0.3 wt% Si using 10 kV acceleration voltage, the detection limits are around 0.05 wt% using the Cu L-line, 0.01 wt% using Be K-line, 0.16 wt% using Co L-line, and 0.03 wt% using Si K-line.
When analyzing materials with compositions close to the detection limit, special attention must be paid to sample preparation. The samples should be polished flat and be free of all surface contamination. Especially when light elements are analyzed, any carbon contamination on the surface can lead to additional absorption and reduction in analysis accuracy.

Question 17:
With older technology, how can you make sure the carbon readings you get are from the sample and not the polymer window?

The contributions from polymer window will usually be taken into account when the quantification calculations are performed. However, carbon contamination can build up on the window over time, and this will affect the performance of the system by absorbing incoming x-rays, particularly low energy x-rays. Measuring the spectrum from a carbon containing standard will help determine if there are any adverse effects of contamination and for some systems, it is possible to include window contamination in the modelling.

Question 18:
Did you find any effect of Si3N4 grain size on interaction with acids?

We have not tested the solubility of the silicon nitride windows in acid and would hope that this is not a situation anybody would encounter in their microscope.

Question 19:
Good after noon, we have a SEM: Hitachi TM 1000, how to know if its EDS is correct?

A collection of spectra from traceable standards is a very good way of characterizing the performance of and EDS system. If the measurements are repeated at set time intervals, they can also help track any contamination or other problems as they might arise.

Question 20:
Is silicon nitride window used for all SDD nowadays?

No, silicon nitride windows are a very new addition to the EDS window technology. Currently only EDAX offers this option.

Question 21:
Will silicon nitride window generates Si and N peaks in spectrum?

Typically the absorption at the energy lines of the window materials are more of a concern. In general the silicon nitride has better transmission than the polymer window, though of course there are absorption lines from the window, support grid, and light reflector. However, the window characteristics and absorption lines are taken into account when quantifying the spectra.

Question 22:
Could you elaborate more on plasma cleaners for detector windows?

The use of plasma cleaners is not recommended for polymer windows as there have been some indications that it can create pinholes in the window and/or light reflector. We are currently testing the resilience of the silicon nitride windows using long term plasma exposure.

Question 23:
Can you separate on fast map Sulfur Lead and Molybdenum?

The S-Pb-Mo overlap can be very tricky and typically it is difficult to separate the peak overlaps directly using fast maps. Successful separation relies upon good statistics and long collection times. However, often phase mapping can be a solution for the problem. In phase mapping the difference in the raw spectra are used to separate the phases, and if a third component is found together with S-Pb but not with S-Mo, the phases can be separated successfully.

Question 24:
How do you measure an 8 nm carbon coating?

The example in the presentation was a calculated spectrum where the 8 nm was an input parameter, but quartz crystal microbalances is a technique that is routinely used to measure deposition layer in the nm range. The TEAM software carbon coating correction analysis the intensity of the carbon peak under the analysis conditions and will calculate the thickness of the coating. This is particularly useful in quantification when light elements are present, which can be absorbed by the carbon.

Question 25:
Can you see boron?

Yes, energy lines down to Be can be detected on most modern EDS systems.

Question 26:
What are the typical beam currents used at the extreme low energy analysis (e.g. <1keV)?

As the electron energy decreases so does the x-ray yield, and more current/more electrons are needed to get acceptable statistics in a reasonable amount of time. Often beam currents in the tens of nA are used, as long as the sample does not show beam damage.

Question 27:
Could also organic Domains such as 10-20 nm in size be resolved with EDAX in materials typically used in bulk heterojunction solar cells?

As organics are primarily Carbon, the analysis depth remains quite large even at low acceleration voltages. At a voltage of 5 kV the analysis depth in a pure Carbon sample is several hundred nanometers, so spatial resolution for Carbon is limited by the physics of x-ray generation and attenuation/escape depth. Depending on sample composition and geometry this means that the signal will typically be averaged over a fairly large volume. However, if we are considering Carbon inclusions in a heavier element matrix, the Carbon escape depth in the matrix will be much smaller, and good spatial resolution can be achieved, though 10-20 nm would definitely be a challenge.

Question 28:
Can you refer us to a paper discussing the technique used to get the FWHM resolution?

The detector resolution is calibrated by measurements on an Al-Cu sample and reported for the Mn K? line according to the ISO 15632:2012 standard. The FWHM of the spatial resolution in the presentation was calculated by taking the derivative of a line profile across the interface and fitting the curve with a Gaussian peak.

Question 29:
How to get better signal of Nitrogen in spite of its scarcity in my sample?

Lighter elements and/or low energy peaks are often challenging and care must be taken to obtain the best quality data. The first step is to ensure that the sample is at the intersection distance where the x-ray count numbers are maximized. Secondly an appropriate acceleration voltage must be selected. Lower acceleration voltages tend to favor low energy detection due to a reduced interaction volume and low energy x-rays being able to escape from a larger fraction of the excited volume. Detector resolution also comes into play and often the slower amp times will produce better results for low energy peaks, especially if several peaks are close together. Finally count statistics are important, the beam current should be adjusted to keep a dead time should below roughly 30% at the chosen amp time and the acquisition time should be long enough to ensure good spectral quality. If this still does not provide acceptable results, tilting the sample towards the EDS detector can also provide a slight increase in the light element response.

Question 30:
Did you perform of quantification of the maps? How did you subtract the bremsstrahlung?

Yes, most of the maps in the presentation were quantified to remove the region-of-interest overlap between Si and W. The quantification of the maps can be done automatically in the EDAX TEAM software using a similar quantification method as is applied to normal spectra. This includes geometry factors, ZAF correction, detector and window response, and background fitting/subtraction. Bremsstrahlung subtraction is thus part of the quantification routine and relies on software routines.

Question 31:
If we acquire spectra with extremely low voltage and very large probe/aperture, do we loss spatial resolution?

Yes and no, it depends on which resolution we are referring to. While very impressive resolutions in electron images have been shown at low voltage, the resolution will get worse with large spot size/probe size and with large apertures. However, the x-ray resolution and the electron resolution originate from two physically very different mechanisms. The interaction volume for x-rays is primarily governed by the electron energy, and will decrease with decreasing energy (resulting in better spatial resolution). The x-ray spatial resolution is typically orders of magnitude worse than the resolution of the electron beam, but gets significantly better as the beam energy decreases.

Question 32:
Can you say something about minimum detection limits?

Minimum detection limits are dependent on the individual sample, microscope parameters, x-ray lines in question, acquisition time, and detector characteristics. There is no single detection limit that can be obtained across the board independent of these factors. Typically we define the minimum detection limit as the required concentration to have a peak height that is above the background with 95.45% confidence limit. Often detection of below 0.1 atomic percent is feasible, but it is highly sample and parameter dependent.

Question 33:
How reliable is the Williamson hall analysis for nanocrtystalline material (size < 5nm)?

The Williamson Hall analysis is based on Bragg diffraction in the sample where the x-rays are elastically scattered. In EDS an electron beam is incident on the sample which leads to characteristic x-ray and continuum (Bremsstrahlung) generation. On top of that, the EDS detector is mounted in a fixed position and is incapable of measuring the signal as a function of angle. For these reasons, Williamson Hall analysis is not applicable to EDS signals.

Question 34:
How can we determine the phases in very thin film formed on glass substrate because it shows amorphous nature which may come due to glass? What will be the appropriate method according to you?

When we discuss phases in EDS, we are referring to areas with different chemical composition, but EDS does not give us information about the structure and whether the region is amorphous or crystalline. For crystalline materials EBSD is a very powerful tool that offers great material insight when combined with EDS, which enables both chemical and crystallographic analysis. When analyzing sub-micron thin films, one must also take the escape depth of the x-ray radiation into consideration, as this can easily exceed the film thickness. Depending on the exact problem in question, alternative techniques might include ellipsometry, x-ray diffraction (XRD) and x-ray fluorescence (XRF).