What are your spectra calibrated to? Is it possible that your calibration is wrong or inconsistent? It sounds like you are calibrating to carbon and the topmost surface has enough carbon signal to get an accurate calibration then while sputtering the carbon signal goes away and with it the calibration. Then at the interface with the SiO2 there is more carbon from the substrate and calibration become accurate again.

Are you using a flood gun for charge neutralization?

If you want to verify the existence of a charging effect, analyzing one peak isn't enough.

A charging effect can be seen as verified, if several peaks of the compound (here Ti and O) exhibit the identical shift, usually to higher binding energies.

Christopher Goodwin, we did not reach SiO2 yet. In case of depth analysis, would the calibration matter? We did not take C scans. We use a flood gun .

Jürgen Weippert, we have a negative shift in case of less conductive sample(pure TiO2). I will check the O scan as well!

Regarding the O scans, both samples showed increasing negative shifts with time of sputtering;

However, Ti scans are different. Black TiO2 showed positive shift, while bare TiO2 increasing negative shift with time of sputtering;

here is the comparison

Just as Jürgen Weippert said a reference peak for binding energy calibration is needed to determine some change, whether it is oxygen, carbon, or even the valence band.

Also looking closer at your spectra it seems like there is more going on than just TiO2 it may be that you have some mixed oxide and the peak position is indicating a change in chemical specie.

Ar+ ions are used for etching. So, Ti3+ and Ti2+ species are expected to appear during the process. As I know, those species may cause a negative shift. However,there is a positive shift in case of black TiO2 for Ti peak.