Your question is not clear.

X-rays are collected from a spot, so "direction" here is meaningless. If you mean line scan, please elaborate. 

Dear Prof. Vladimir

Thank you for your answer. Actually I took different EDX spots on a certain direction of an aluminum alloy sample I found the major element is Al. But after I took EDX for different spots on the transverse direction I found the dominant element is oxygen.

Dear  Sadeem,

It is necessary to remember that EDX is not the chemical analysis of a massive specimen. It is just method for estimation of elements distribution on the surface of your specimen (usually 1-3 µm depth) and is very sensitive to contaminants. Most likely, your specimen is just contaminated.

Good luck!      

P.S. What techniques of the sample preparation did you use?

Dear Denis

Alumina is expected in the prepared samples, the method is reactive spray deposition.

EDX can be direction sensitive if your sample is not flat. If there is something protuberant between your spot and the detector window some photons can be absorbed by the protuberance. The less is the energy of the photon the more is absorbed that the reason why light element (like oxygen) is easily stopped. May be you encountered this kind of problem. Different results like that can be obtained also with a tilted sample. Check if your sample is flat and horizontal.

Dear Stephant

Thank you for your participation. 

The same has been grinded and polished before test, so I think the effect of flatness suppose to be excluded.

Dear Sadeem,

Remember that EDX is a technique for superficial estimation of elements of your specimen (see comment of Dennis Korneev) an it is sensitive of direction an place than you develop the characterization. For this reason you can obtain different information  in the same especimen and in the sample the elements can be distributed non-uniformly.

You need to repeat your measurements. 

EDS is not "directional". So, whatever you have found can be a)real, b) contamination c) operator error (for example change in accelerating voltage can dramatically change relative heights of Al and O peaks).

What's other measurements that you suggest me to do to confirm the results or deny it?

Thank you

In order to prevent any speculation, why you do not post two of the spectra and an image from the surface?

Is the material embedded....and did you measured close to the embedding material?

@ Vladimir: It is not really related to this problem here, but what do you expect for your EDS signal if you have strong channeling? :-) 

@Gert Nolze

Strong channeling is rather something exotic for me. You need special specimens/specimen preparation techniques to see it. And I do not expect it will influence semi-quantitative EDS results. Anyway, as Sadeem Fadhil infromed us, specimens were just polished, so there should be no channeling contrast.

Dear Gert and Vladimir

Thank you for your valuable advice, I wish I can put it publicly but the image is now copyrighted and I need to get the agreement from the copyright owner first, which is now not possible till the review process is finished.

Actually I need to confirm the result but I think using the same test again will not change anything, what other tests you suggest me to do to check out the composition at these two directions?

Thank you

Sadeem

Sadeem, despite my short comment about the direction dependency I totally agree with Victor. It is more a curiosity I was pointing out. Channeling will definitely change the penetration depth of electrons, but the final effect on the information depth differences are in the scale of some percents only (in maximum). It would explain small deviations but not such big one you referred to. The Monte Carlo simulations usually applied to explain the origin of X-rays assumes no crystalline materials but amorphous. In case of typically prepared samples it is OK. However, immediately if you see orientation contrast of backscattered electrons an obviously remarkable channeling occurs. How much this affects the EDS signal hasn't been investigated yet (as far as I know). A bit surprising since the majority of material is crystalline and undergoes channeling, but I guess, the preparation is often so bad that the upper layer of the sample is highly damaged and prevents any channeling. The other thing is: who really wants it? :-) 

Sadeem, as has already been stated the technique is directionless. Please can you tell us what your "directions" refer to? Is it a bulk aluminium metal sample that has been cut into two, one of which is rotated through 90 degrees to show the transverse (the cut surface) - The two surfaces are both polished and then mounted together horizontally in the SEM, and a spot from each measured under identical conditions?

If you are not able to post the image can you post the data generated from each spot?

Dear Ian

Its one sample, can be considered as bulk. First, within a certain direction different spots have been taken there and the compositions of these points have been measured. Then the sample is rotated by 90 degree and the same process is repeated. For the first direction the dominant element was Al by 83% and oxygen was around 5% and for the second direction the dominant was oxygen with 64% and Al was only 28%. Both surfaces were grinded and polished before the measurement process.

Sadeem, that does seem quite bizarre.  If this is aluminium metal you should not be getting oxygen, so it seems not to be pure Al. the stable oxide is Al2O3, but is not metallic, and should give about 53%Al, 47%O. So neither of your analyses are close to this. 

The totals are 88% and 92%; what other elements do you have? What do you get if you analyse a pure Al metal and a pure Al2O3 sample under the same conditions? Is your system calibrated and checked against certified reference materials?

Sometimes aluminium can be a cause for problems because of incorrect collimator alignment on the EDX detector, since that is made of aluminium. There is also a lot of aluminium inside the SEM and can cause so-called system peaks, so it is important to check for these on a calibration standard, such as cobalt.

As for other techniques to check the results, can you run XRD on the two orientations to check for phase differences? Or even to identify what phase you have!!

Dear Ian

I agree with you, I need to accomplish XRD test in both directions.

Thank you