Difficult to answer. In my opinion, as with most analytical techniques (besides weighing and counting...) the accuracy depends mainly on the use of suitable standard materials.  So if the peaks are well separated and if the analyzed layer is thin (no self-absorption), you can get very good accuracy with elemental standards. Taking into account self absorption is more complicated. Then it is difficult to give a measure of accuracy if you calculate self absorption based on theoretical values. If you have the possibility to measure absorption with the same setup, you can get rid of this problem.

If you mean accuracy by detection limit/sensitivity, then WDS is having a higher detection limit/sensitivity then EDS. But both are poor in detecting trace elements ( of lower atomic numbers).

I think it depends on counting statistics and many other factors.

EDS is most often used for qualitative elemental analysis, simply to determine which elements are present and their relative abundance. Common problems relating to EDS in the AEM occur either during data collection or analysis. Data collection problems include (a) poor sample preparation, (b) incorrect sample geometry and (c) contamination. Data analysis pitfalls includes (a) artifacts - escape peaks and sum peaks, peak overlaps - the spectral resolution of EDS is not a great as WDS (b) escape peaks - there is a statistical probability that some of the X-rays, generated in the sample and impacting the solid state detector and (c) sum peaks - this phenomenon occurs where the count rates are moderate to high.

WDS was the original technique developed to precisely and accurately determine chemical compositions of microvolumes (a few cubic microns) of thick specimens, and the instrument used is the electron microprobe. WDS is used for non-destructive quantitative analyses of spots as small as a few micrometers, at detection levels as low as a few 10s of ppmw. The high spatial resolution of WDS not only allows quantitative analyses to be performed on small phases but also to detect chemical zoning on a small scale within a material When the electron beam is rastered, the WD spectrometers can allow x-ray image maps of individual elements to be constructed.

The disadvantages of WDS include (a) WDS cannot determine elements below atomic number 5 (b) some peaks exhibit significant overlaps that result in analytical challenges (c) WDS analyses are not able to distinguish among the valence states of such that this information must be obtained by other techniques and (d) multiple masses of an element (i.e. isotopes) cannot be determined by WDS, but rather are most commonly obtained with a mass spectrometer.

When compared to EDS, WDS exhibits superior peak resolution of elements and sensitivity of trace elements.

PLEASE REFER FOR COMPARISION - http://mcff.mtu.edu/acmal/electronmicroscopy/MA_EDS_WDS.htm

THEREFOR I LEAVE IT TO YOU TO DECIDE SELECTIVELY FOR WHICH COMPOUND TO USE EDS OR WDS.

BEST OF LUCK. 

@ A. Ghosh: nice summary of both techniques, Thanks. Nevertheless, the major topic of the question is not yet answered, although I admit that it is not easy to give a general answer. I also need to disagree the obviously general disbelieve that EDS is only useful for qualitative analysis. There is no doubt, that the precision of a WDS measurement is much better than EDS which is mainly related to small gradients or trace element analysis. However, the accuracy is for many applications absolutely OK. Otherwise it wouldn't explain all these numerous SEM or especially TEM which uses EDS for chemical characterization. Again, if you need to identify tiny changes, other techniques are better.

Your statement regarding the spatial resolution also sounds too exclusive. I would say that EDS has at least the same but probably a better spatial resolution compared to WDS. This is already given by the commonly much lower beam current used in SEM. Therefore the typical spot size used in a microprobe is one magnitude or more bigger then the focused beam in SEM.

Arindam, do you have any number about the better accuracy of WDS and EDS? My current assumption was, that WDS is extremely precise (decimal places), but for a good accuracy one need to find a good standard which matches the investigated material (which is quite difficult for an unknown material). Here the extremely good angular resolution turns into a disadvantage. Therefore, the pre-decimal point position is perhaps wrong, but the variation at the decimal places (relative changes) are unbeatable, i.e. the accuracy is much worse than the precision. For EDS it is (more or less) vice versa. Minimal changes are hard to see because of the bad energy resolution and the superimposition of all elements (in WDS you still only measure one element, as far as I know). However, assuming no critical composition the pre-decimal point positions are quite accurate (but not precise). 

From my very limited point of view we have two techniques: one is high-precise but not simultaneously very accurate, and the other one is less precise, but compared to the reachable precision quite accurate. Both statements are relative which explains my question. Are the accuracies of both techniques comparable or is one technique in general more accurate than the other? WDS is definitely much more precise than EDS.  

A last comment: X-rays are formed by the electrons of an element. Therefore, the above mentioned isotopes cannot be distinguished as long as they don't affect the electron configuration remarkably. The valence state, however, is from my point of view an image of the electron configuration and therefore detectable by WDS. Certainly not always that clear since for heavy elements the change of the valence electrons does not affect remarkably the inner electrons which are used for WDS, but for elements with small element number like O, N, P , S, C etc generating with respect of their valence or even bonding character a so-called chemical shift which is the reason for the required (ideal) standards in WDS. Please correct me if I am wrong!  

@ R. Krishna: accuracy is (as far as I know) a number which defines how close the result describes the true value. Precision defines how good is the reproducability, i.e. a technique can determine a value quite accurate but with a comparatively strong scattering (accurate but not precise which is mainly given by EDS), whereas another can be very precised but perhaps gives a less accurate value (which might be the case for WDS if you do not have any reliable standard). The given Wikipedia link introduces the term trueness which describes in combination with the precision the accuracy of a value.  

https://en.wikipedia.org/wiki/Accuracy_and_precision

Sir Gert, from my side I think you are correct. I had preferred EDS over WDS for my samples probably as because I find it easier to conclude. To be frank, I am also not so clear about the numbers about the better accuracy but WDS gives more sharp peaks when compared. I am attaching two figures. I think they will help you. Hope you will like and upvote my answers. Also, thanks for the detailed analysis of my answer. 

@Gert Nolze 

I have worked a lot with microprobe (3 WDS, but no EDS), but it was 20+ years ago. Since then I am working exclusively with different systems of EDS, so no direct comparison on the same instrument from me.

You have mentioned many times that WDS requires standards, but EDS also requires the same standards, no difference here (standardless analysis is pretty unreliable). 

From physic's point of view there is no substantial difference between EDS and WDS, just different detectors with different abilities. For quantification purposes WDS is always better: much better energy/wavelength resolution, much better peak/background ratio. Energy shifts (chemical shifts) are determined by interaction of electrons with specimens, WDS is just better suited for their detection. So, both accuracy and precision are better for WDS.

So, whenever good quantification required WDS should be detector of choice.

But: WDS is a bit slower (may take 5 minutes to acquire the full spectrum when about the same quality spectrum could be acquired in 1-3 min with EDS. 

WDS is highly sensitive to topography - small changes in specimen height  can lead to significant changes of signal intensity

WDS is sensitive to.beam deflection from the center of field of view, so that maps of homogeneous material at lower magnifications (usually less than x1000) will show lower intensity at their edges (it could be corrected with software, but any software corrections a prone to errors). EDS is much better suited for everyday use, when only qualitative/semi-quantitative results are needed, and more forgiving to insufficient level of operator's training. However, with proper standards (yes, standards), EDS can provide very good quantitative results.

Vladimir, I know still from former times that standards are also required for EDS. Newer systems, however, mainly used standardless approaches, and I have to admit, it seems to me like for XRD where 20 years ago a phase fraction was also only possible if you generate standard mixtures. This time has gone since full pattern refinement. I guess it might be similar for EDS as well. All commercial systems doing standardless analysis, although I guess because of the complexity a standard-related analysis might be still more accurate. Also the development of Silicon Drift Detector with slightly better energy resolution and much higher count-rate ability increases the statistic further. This all has nothing to do with accuracy and resolution but it makes deconvolution more reliable. Did you check the difference between standard and standardless analysis?

Thanks! 

No, I did not check difference. I just know that standarless analysis is developed for EDS not because of some special features of EDS, but because EDS is used mostly not for high-precision measurements, where standarldess analysis can thrive. Standarless software can be developed for WDS also, but what for? WDS from the beginning was used in areas where more precise (means - with standards) measurements were needed. 

I wouldn't say that standardless EDS has been developed because it is mainly used for quantitative analysis. Nowadays, practically all measurements are done by standardless techniques since they are obviously reliable enough. (you are certainly right that they are not sufficient to classify unknown phases). The major advantage of standardless EDS is the opportunity to measure at any acceleration voltage which was before usually only possible for 20kV or partially for 10kV since only there element profiles were available. The applied models are obviously that good that measurements in variable pressure, with coatings, or even the thickness of a specific layer can be determined. I don't want say that the models are perfect (this is also not the case for XRD). But for standard applications it works surprisingly good if one knows about the weak point (described above by  A. Gosh). Unfortunately, I don't know any paper where it has been critically compared (which is my fault since I simply don't know it. I cannot believe that nobody did this already). 

Have you seen this paper (authors are "high authority" in SEM/EDS/WDS):

http://download.springer.com/static/pdf/828/art%253A10.1007%252Fs10853-014-8685-2.pdf?originUrl=http%3A%2F%2Flink.springer.com%2Farticle%2F10.1007%2Fs10853-014-8685-2&token2=exp=1444591651~acl=%2Fstatic%2Fpdf%2F828%2Fart%25253A10.1007%25252Fs10853-014-8685-2.pdf%3ForiginUrl%3Dhttp%253A%252F%252Flink.springer.com%252Farticle%252F10.1007%252Fs10853-014-8685-2*~hmac=417781d275279775e0d425f1ba3c481a695a8dc2e350eb9e7c84b5b29f597ccd

They state that EDS is very good now if used with standards.

That has been always a hot, unsolved discussion at our department.

According to my experience, a "bad" standard sample and a bit of deviation from the optimized WD for WDS measurements can lead to a very un-accurate, un-precise result. EDS precision, however, depends less on the sample positioning inside the SEM.

Again according to my experience, EDS has better precision, but normally for higher accuracy we select WDS that could have less precision! (given the fact that wrong standard or WD can lead to a worse accuracy than EDS measurements)

Hi Ehsan, I don't believe (!) than EDS has a higher precision than WDS. Are you talking about a WDS attached on an  SEM or a microprobe? For the WDS/SEM combination you are perhaps right!

Thanks, Vladimir!

Advantages of WDS are

1) Higher  probe current (20 nA) is used so the amount of X-rays generated is higher . hence quantification/detection will be better.

2) WDS we collect  X- rays at  background  and peak positions,which ( P-B) is used to relate to composition. 

3) better for mineral  analysis provided proper reference standard is used.

Disadvantages:

1) depends on Quality of standard reference  and  closeness to the sample matrix being analysed.

2) time consuming.

3) optically flat sample required. any slight variation in focussing height can lead to erroneous result especially for light elements.

Dear Dr. Nolz,

Can you please let me know the energy resolution (eV) of your EDS detector , energy resolution (%) of WDS detector, Goniometer radius of your WDS system. I can try to give some idea about the accuracy of your systems.

Our EDS detector has a resolution of 123eV and we are using a microprobe Jeol JXY8900RL. 

Nevertheless, are you sure that you can tell me something about the accuracy?

I believe, from these data you can only make some rough estimation about the precision (comparing peak widths and anglular resolution of the goniometer).

Thanks in advance

G. Nolze

WDS because the spectral resolution of WDS is better the peak to background ratio is better, therefore, even if doing qualitative analysis of trace elements (0.1 or %) it is mandatory.

If one is interested in the coordination number of low atomic number elements (i.e. Si) EDS is mandatory, because only with EDS resolution the displacement of the spectral lines will be visible.

When low energies lines are to be measured, (low atomic number elements) or L  lines of medium atomic number elements, or M lines of heavy elements, WDS is necessary.

On the other side, one can say that its best virtues are its worse defects: because of its sensitivity, in order to get sensible results one needs: specimens that are well polished, standard specimens with similar composition to make on-the-spot calibration, and a well trained operator.

Let me repeat again:

For good quantitative results specimen preparation and quality of standards should be the same for both EDS and WDS. No difference. 

In my opinion Vladimir Dusevich has provided quite a wide spectrum of sound evidence for a comparison between EDS and WDS.

One aspect which is never enough been emphasized is the operator’s capability to obtain good data from one system or the other. In other terms, how well the operator is trained in electron probe microanalysis. Naturally, an EDS system is “good for all seasons” (standardless or not) for its intrinsic push button nature, whereas for a WDS system the operator normally has a wider choice of parameters (to be input by the operator), hence it’s more subject to the ability, training and/or sensitivity of the analyst sitting at the console.

EDS system is that the user can quickly collect a full spectrum simultaneously.

WDS, on the Bragg Equation, collect a full spectrum by scanning, using different analytical crystals. (Time consuming!).

EDS has a low resolution (be between 135-150 eV)

WDS system, for same X-ray line will only be about 10 eV wide.

WDS has a higher Peak/Backgroung Ratio, very important for detection of low concentrations in ppm range.

Can anyone explain why we can not use WDS for TEM?

X-ray signal from ultrathin specimens (TEM) is much weaker than from bulk specimens (SEM); too low for WDS.

thank you Vladimir Dusevich

"Can anyone explain why we can not use WDS for TEM?"

First it is also a geometry challange, a WDS is a big device which needs empty space to the X-rays emission point focussed. Even if engineering challange would be solved, it would be more hard to find the focus point (cross section between electron-beam and detector optical axis) of the X-ray emission (nm dimension) reproducable... and this makes the measurment reproducability more challanging. The EDS is with much higher acceptance of X-ray emission point and therefore more reliable in reproducability of the results... the only choice for TEM therefore.

And as already mentioned by Vladimir, the much bigger 'solid angle' of EDS collects more X-rays in same time, the small X-ray emission in TEM would need very high acquisition time with WDS.

I'd like to give additional statements to the initial question and some of the answers.

"From my very limited point of view we have two techniques: one is high-precise but not simultaneously very accurate, and the other one is less precise, but compared to the reachable precision quite accurate. Both statements are relative which explains my question. Are the accuracies of both techniques comparable or is one technique in general more accurate than the other? WDS is definitely much more precise than EDS"

...

"You have mentioned many times that WDS requires standards, but EDS also requires the same standards, no difference here (standardless analysis is pretty unreliable).

So, both accuracy and precision are better for WDS. So, whenever good quantification required WDS should be detector of choice."

There are excellent works and publications by NIST group, search for "EDS measurements of X-ray intensity at WDS precision and accuracy using a silicon drift detector".

and

"Performing elemental microanalysis with high accuracy and high precision by scanning electron microscopy/silicon drift detector energy-dispersive X-ray spectrometry (SEM/SDD-EDS)" 2012/2015

The headline says already all. The major reason for this development was the SDD ability to collect high count rates, it overcompensates the weakness with worse signal to noise.

Then also, indeed the EDS is able to perform standardless, in meantime. One needs to say even the vast majority of all users do use standardless, regardless it is sure that the use of standards can improve accuracy further (required/recommended even in special element combinations....).

The remained superiority of WDS keeps in detection/analysis of small specimen constituants. It is due to the about one order of magnitude better P/B signal. This is especially beneficially to use if low constituants peaks do overlap with other major components X-ray line sereis (even if usually not severe overlaps are; or EDS pile-ups/sum-peaks do disturb etc). It was mentioned already at beginning, the better resolution helps in several element combinations. But with precise energy calibration (and perhaps standards comparison additionally) also the EDS is nowadays with no significant disadvantage (but as mentioned, clear disadvantage if low constituants are in the overlap).

===============

In my view the WDS can develop it's superiority even in combination with EDS then, it complements the EDS abilities which are in most of all cases sufficient for accurate and precise analytical results. Please visit the M&M 2022 contribution "The Promises when WDS Supports the EDS X-ray Analysis in SEM and the Evaluation Algorithms do Merge". There is even shown a way the WDS is even possible to utilize completely standardless (in combination only), means a SEM operator can use EDS/WDS beneficially without using or even having standards.

Poster M&M 2022 poster "The Promises when WDS Supports the EDS X-ra...

Article The Promises when WDS Supports the EDS X-ray Analysis in SEM...

In the case of WDS, a mechanical spectrometer is used to detect X-ray lines one at a time. Instead, an electronic spectrometer is used to detect X-ray lines simultaneously in EDS. Hence WDS has a significantly higher spectral resolution than EDS.

WDS: Better Resolution, can be variable, most suitable for monoenergetic incident beam on target(X-ray target of sample)- Better for Raman, Compton, plasmon X-ray scattering

EDS: Relatively lower spectral resolution, fixed resolution, sutable for polychromatic incident or relatively high energy beam compared to binding energy of the target element(X-ray target)- Multi-elemental analysis

In terms of the accuracy of determining component contents above 1%, the WDS and EDS methods are comparable (Newbury and Richie, 2013), while EDS provides the best accuracy in determining the ratio of components, and WDS is certainly more accurate in determining impurities. But for EDS it is important to use adequate software for spectrum processing, unfortunately, not all software for spectrum processing meets the requirements of quantitative analysis.

Dear @Nikolay Karmanov

How come "EDS provides the best accuracy in determining the ratio of components" when, according to your earlier statement, "accuracy of determining component contents above 1%, the WDS and EDS methods are comparable"? And, by the way, can you please to name a manufacturer, whose EDS software is inadequate?

I worked with both techniques for Compton scattering analysis (please find the attached file for better clarification). what I found

1. Resolution of WDS is better, we can get about 1 eV.

2. Resolution of EDS is 135 eV-200eV

3. Both the alpha and beta lines of a number of elements or multi-elements make the spectrum very crowed or sometimes superimposed with each other because of the resolution of EDS detector, that type of thing can be separated out using WDSthe

Subject to the requirements of quantitative electron probe microanalysis (correctly prepared and properly installed sample, standards, matrix corrections, etc.), the accuracy and correctness of determining the main (10-100%) and minor (1-10%) components by both methods are comparable. At the same time, at contents near the lower boundary (1-3%), WDS provides some advantage. There are good spectral line deconvolution algorithms for EDS, so even when examining rare earth minerals, for example, EDS provides better results than WDS (my EDS-experience is Oxford Instruments INCA and AZtec microanalysis systems). The thing is that there are no sufficiently reliable methods for taking into account spectral overlaps for WDS, when there is no "empty" space in the spectrum for measuring the background at least on one side of the peak.