4DS - Anything but Charting, page-28516

@Dramapig
what they are not saying is there is data (retention) loss across time - the state of resistance dissipates as ion migration occurs
it is an inherent trait of area based non filamentary rram - those who question this simple material characteristic in the process can jump on the web and find it - there is plenty there

Then in this type of non filamentary device you must factor in the environmental conditions of operation
why?
because the ion drift/migration ( and there for by association the retention value) is also reactive to heat ( stored or applied) as well as operating current / voltage strength and switching values related to time in operation.
So retention will be observed but at this stage dissipating over time outside the norms of recorded retention benchmark measurements.( 'persistent' indicates it will reoccur and 'tuneable' indicates it is relevant to applied current)

Testing for this continues at 60 and remanufacture at 60 to refine the process and achieve improvement.
Manufacture at 20 has (based on 60 process improvements and etch success) been deemed an appropriate step to test manufacture at that process node for success in all the same benchmark areas of the data.

so simply on retention only ....( understanding the great result on speed and endurance)
they will attempt to get it working at 20 first and foremost
then most likely on another Lot at 20 they will attempt to stabilise the resistance values relative to applied current there by removing the unintended changes described above and achieving that tolerance level of acceptable resistance change which will reveal the retention numbers everyone wants to see....they have not got that far yet imo

in my view It would be better if they said this themselves / or whatever they have data on in relation to resistance rather than this....


I find it insufficient ....but that is just my opinion