It is hard to tell what is the problem here. Usually XRF has a larger information depth and will probably look much deeper into the sample than EDS. Provided that your sample is homogeneous enough I would not expect large differences. If homogeneity is not guaranteed you have to digg into the sample using FIB or similar and average over several measurements.

My material does not include some of elements in the result. Even my material is not homogeneus enough , i should have seen major elements in these 10 results from different points of my sample. ( Up to %60 Co)

Thanks christian,

These informatin was vital, i will check them again.

1. XRF and EDS are very close techniques (in most cases even the same detector). However XRF is always performed with certified standards while many people do EDS without any standards hoping to get right results, and sure they cannot get them. For proper EDS analysis you should calibrate you system with certified standards or at least some homemade standards which you believe are 100% (OK, 99%) right.

2. If you are looking for light elements, then both methods are not really good, but EDS is better. If XRF does not have vacuum chamber it will not detect light elements at all.

Generally the problem could be due to (in)homogeneity of your samples, due to own measurement setup or due to calculation towards quantification. Thus I think You should be more specific because many reasons could be behind this inconsistency of results. E.g. describe your sample (homogeneity, presence of grains, …), describe (qualitatively) results obtained by SEM-EDX and XRF, describe equipment (spot sizes, anticathode target material, accelerating voltage)… From the answers origin of problem may be deduced.

E.g. important question is, if there is difference only in quantification or as well between both measured spectra. From the answer to this question You can deduce, if the issue is caused only by quantification (calculation) or if the issue is directly connected with measurement.

As written above, penetration depth of X-ray within XRF is much deeper than penetration depth of e-beam (or better to say excitation volume) within SEM-EDX. However, also penetration depths of characteristic X-rays of particular elements must be considered. (And consequently also matrix effects, where heavy atomic mass elements may absorb characteristic X-rays of low atomic mass elements.)

Also, performances of particular equipment may play a role, especially for detection of low atomic mass elements.

Generally, for homogeneous samples, there is no reason to obtain roughly different results (with respect to principles of both techniques).

Use your EDS and XRF analysis in a sample that you know the elemental composition!

Typically, as mentioned in the above posts the XRF information depth is greater than SEM-EDS. XRF usually also detects elements at the ppm level, whereas EDS detection limit is typically at 0.1% level depending on the accelerating voltage you apply as well. So in XRF of same sample you may get extra elements.

In XRF usually a much larger surface area information on the sample is obtained at one go (irradiated area is large)..and if the inhomogeneities are micro-ones EDS will give differences at diff spots..hence its not always right to average the EDS information from several spots.

Above all XRF software can take care of matrix effects, whereas EDS corrections are ZAF based and gives only the relative concentrations even if you use certified standards.

In XRF, If there are light elements (not detectable) in your sample, you will need several certified standards of the same matrix as your sample to get correct calibration and hence quantification, ...But if all elements in the sample are detected in XRF it can give good quantification using standardless Fundamental Parameters (FP).

I would suggest running standards close to what you are after in order to determine if it is a question of calibration or something other than that. I agree with most about this and the argument relating to homogeneity. If your sample is not homogeneous and the areas being analysed are chosen at random and hence do not correspond to each other between the two instruments chosen, I'm afraid you will never get satisfactory results. It's not that the instrumentation is bad. It's just like that! I analyze old paint samples. They lack homogeneity and hence results are never the same, they vary even by a standard deviation of 30%. But its acceptable as it is within such a sample's nature. Hence, standards should answer a lot of questions for you.

One tip though: make sure your beryllium window is clean, sample is not contaminated and that the peaks you are seeing are truly from the sample and not analytical artifacts.

Hope this helps.