How to deal with dsciATAC data, it's about 50,000 cells per sample.

[Goood2001]

I'm working with an object which is about 50,000 cells per sample. When I use ArchR addDoubletScores() function to inferring dsciATAC-seq Doublets, there are only a few cells can be detection as doublets and most of cells have low doubleScore. About this issue my solution is using subsetArchRProject() to split my ArrowFiles into 3 or 5 parts and calculate each DoubleScore. Then I can get normal doublets rate.
Thus, is there any max level of cell numbers per sample that ArchR can deal with? I guess it's around 10k cells?