CXO888 / any other attendees, page-13

yeah - that goes to retention which is described as not critical ( paraphrasing)
they talk about tunable retention....this is true but it is a material (pcmo) mitigation statement
PCMO exhibits resistance drift between hrs and lrs switching - there is measurable drift to the lower resistance states
this occurs because there is a local electric field in the bulk area of the device...this field is lower than that of the interface exchange area resulting in 'some' ions/oxygen vacancies migrating or drifting to the bulk area and getting stuck there ....and not returning to the interface area - affecting retention - that is why to date retention has not been characterised - it remains a randomised data set because the drift alters as the write voltage is altered / optimised dynamically

the fact the 'stuck' ions do not return from the bulk area to the interface lowers the amount of ions at the interface reducing the lrs.
So tunable because you have to adjust the voltage to achieve optimal switching....

this is where and why the company has used the term tradeoff ....it is essentially drift mitigation known to occur inherently in pcmo.

they categorically cannot announce a killer use case application until this is characterised and set in accurate repeatable data - I suspect that's why 60nm gets another run

20nm is by no means a given as these material characteristics / issues need resolution and verification at 60 and then same same in a successfully manufactured 20nm process node.

if this is the path there is plenty still to go until any application category is firmed up successfully on a winning
cost / performance ratio - that's what sells it to a major - not another salesman - anyway there is nothing to sell until these things are worked out

plenty on the web about this stuff - its the material characteristics not my opinion
something to think about