In addition: Fe is easily oxidized, during of after deposition. EDS will overestimate amount of O if it is concentrated in a surface layer, especially if a signal is generated by electrons (SEM) rather than X-rays (XRF).

Do you preclean your wafers (ICP etching or acid)? Otherwise a natural oxide layer might be the reason.

Jürgen Weippert, thank you for your answer. The wafers were cleaned ultrasonically in acetone and ethanol successively, and then dried in compressed air blow.

To my extent of knowledge, ultrasonic cleaning is not capable of removing SiO2 from a surface, so the natural oxide layer will stay, for sure. In our institute, we use a variety of methods depending on the application, such as the aforementioned ICP etching, H2 plasmas, RCA cleaning or HF treatment. Those are, of course, harsh methods, but they have proven to be necessary for good results.

In addition: Fe is easily oxidized, during of after deposition. EDS will overestimate amount of O if it is concentrated in a surface layer, especially if a signal is generated by electrons (SEM) rather than X-rays (XRF).

I did not intended to remove SiO2 by ultrasonic cleaning, because I considered I can coat the oxide layer with an oxygen free FeC layer, which was not the case. The deposited layer is 100 nm thick. I will consider the methods recommended by you for cleaning the wafer. Thank you again.

Vladimir Dusevich, thanks a lot for these important information.

100 nm layer is thin for X-rays. So, you can see O from SiO2 thru the layer. Whether O comes from Si or Fe oxides, it is still coming from surface layers.

XPS measurements showed same O content after Ar ion cleaning-sputtering the film surface.

The same result? Nearly impossible. Did you used standards for you EDS quantification? For your specimen EDS of O is mostly qualitative, not quantitative. But may be your layer consists of Fe carbides and oxides? Some leaking in vacuum chamber, not pure enough Ar...

Vladimir Dusevich , the Ar purity is 99.999.