Do you see if there is an actual plasma? Otherwise it might be that it only ignites after a while which would result in a high reflected power during that period. Eventually, a sweetspot ignition may be helpful, i.e. you ignite at "guaranteed ignition conditions" which may involve a higher pressure or power and then quickly switch to your growth condition. If your system involves a shutter, that is the most elegant solution.

I cannot look in the chamber, there is no avialable window now. My RF generator has autotuning and at the beginning it tunes itself pretty fast, so the reflected power is 0 W. Then I can see very slow change of load and tune parameter as the generator/match box slowly heats up. Based on this, I expect to have the plasma.

I have a shutter. I start the plasma with shutter closed. Then if I directly open the shutter and do the sputtering for half an hour, I dont see almost any deposited film. On the other hand if I sputter for half an hour into closed shutter and then open it and continue another half an hour, only then I get thick and visible layer of YSZ. It is really confusing me.

1) In addition to the reflection, do you have a selfbias voltage measurement in your setup? Is it constant?

2) Did you also observe this behaviour with other oxides in that chamber?

The measurement of bias shows constant value (does only very slowly change like load and tune values do) during whole time of presputtering and sputtering. My colleagues do also TiO2on this spttering machine and they dont have any problems with it. They also started with YSZ, so it is not me being unexperienced or something. I was just asked to continue on the experimets with YSZ, because they don´t have time to search for good deposition conditions, since it doesnt work well at the beginning of the sputtering.

I will try to find a way how to look inside if the plasma really ignites. But since the measured parameters (voltage, load, tune) don´t change in time and the RF sputtering works without problems for TiO2, I expect the problem not to be in the plasma ignition.

I also want to put quarzmonitor inside to see the deposition rate in real time, maybe that helps to find when exactly the sputtering starts.

My colleagues did already few tries. One was to put target and sample in the chamber and pump it for very long time to find out if the remaining air in the vacuum or something adsorbed on sample or target surface can have an effect, but the result was same, no spputtered film at the beginning. The base pressure was almost 10-8 mbar, normally we pump down only to go below 10-6 mbar.

Then they tried to do sputtering in the shutter, then sputtering on the sample, then they replaced samples as quick as possible and then the sputtering on freshly introduced sample didn´t work again, which somehow suggests that the cleaningness of the substrate also plays a role. Can the water wapor and oxygen that desorbs from newly introduced substrate somehow affect the sputtering?

Very confusing for me.

I mean, oxygen is sort of an electron scavenger, so it can hamper the ignition, but in that case I would expect a change of the selfbias over time.

The same would be the case for a dark space plasma, i.e. a plasma in a place where it's not meant to be instead of being located over the target. That can easily happen in a customized chamber, but when that flips to the right place, I would also expect you'd see that in the selfbias log.

Thanx for all your answers. Would you say that YSZ target can somehow oxidize or create some surface layer on air, that can prevent sputtering in a way resembling target poisoning like we know it from reactiv sputtering?

It can't be oxidized any further, but at least on the surface a hydroxylayer can form:

Article Structure and Chemical Reactivity of Y‐Stabilized ZrO2 Surfa...

I grow my YSZ be reactive sputtering from a ZrY target, so these parameters can't be transferred; when I want to overgrow it epitaxially, I have to etch it via ICP with rather harsh parameters in order to "break" the surface:

Article Formation of {111} oriented domains during the sputtering ep...

However, I would expect that this effect should be mostly gone after the first sputtering process, so I'm not sure either of these papers is going to help you.