I didn't read the mentioned paper, but from past experience oxygen in the film is due to a leakage (feedthrough).

I also would guess a leak or some residual air in the line after plumbing the sources. Often oxygen remains as -OH, so check the binding energy of O1s by XPS

Since Al has a high affinity to oxygen, even the smallest amount of residual gas in your vacuum (base pressure) would find its way to your film. Also, perhaps some of the oxygen that you're observing is coming from a native oxide on the surface. If you are seeking to reduce the oxygen further (even though it's pretty low; what's the error in the quantification method you're using?) you could try standard maintenance on your system, check for leaks in the gas manifold (bake the lines if possible), and use a higher purity gas sources. Also, process parameters such as flow-rates, pulse duration and temperature can have an indirect effect on oxidation. 

This oxygen content can have different ressources within your ALD system. Examples include base pressure (moderate vacuum or high vacuum), the presence of moisture (usage of load-lock?) or the purity of the process gases (the purity as parchased or the purity reduced by leaks).

Some groups reported significantly lower oxygen content when gases of better purity were used.

Thank you for the good answers.

My question was more general one than related to the ALD tool in Jyväskylä, Finland. We have indeed a load-lock and can deposit for instance nitrides having well below 0.1 at.% of oxygen in them.

The analysis method we use is time-of-flight elastic recoil detection analysis (TOF-ERDA), which can quantitatively depth profile in principle all and in practice light thin film elements starting from hydrogen. The detection limit is typically below 0.1 at.%.

The first concern behind my question was that is it normal and generally acceptable to have even several at.% of impurities, like oxygen in nitrides after oxygen-free process, without knowledge of their origin. I would feel very worried.

The second concern was related to the presence of hydrogen. If the standard analysis tools like XPS, AES and SEM-EDX are available for ALD film studies, in most papers hydrogen is entirely forgotten from the film composition. After measuring about thousand ALD samples, I can conclude that films with hydrogen content less than 0.5 at.% are rare. Some processes like low-temperature thermal- and PE-ALD films can contain even above 30 at.% of hydrogen. To my experience, many times it is the high hydrogen content which is the key to understand other properties of the films.

Dear Timo,

you are absolutely right: the presence of hydrogen can change completely the interpretation of XPS or AES data. However, there is one trick to check for the presence of hydrogen. It is possible to identify hydrogen bonds from the REELS spectra. You can read more about this in the application notes on the Escalab 250Xi of the Thermo Fisher Scientific (these notes are available on line).