This page very briefly summarizes the "negative process ID" issue
Once in a while, processes that should have two daughters
in the final state, have only one left in the GEANT KINE bank.
What's the reason for that?
When looking at Monte Carlo events, one notices that sometimes
the process ID is negative. Why?
The short answer
You can't possibly keep all the particles created and propagated
in a GEANT simulated heavy ion collision, in the output.
There is no way. Something has to give.
The slightly longer answers
In general, there is a cutoff parameter in GEANT, which
determines whether an electron is kept in the KINE bank
or not. Soft electrons (which are obviously copious) at
some point are dropped, while the energy they deposited
is still accounted for. That explains the case when a e+e-
pair was apparently created but only e+ (or e-) is actually
found in the GEANT output.
Now, the negative process ID issue.
The rationale for that (and quite a bit of technicality) can be found in the
"Notes on new shower track philosophy" page which is dated but relevant nevertheless.
In summary, we are being selective in picking GEANT tracks to be saved. We ignore
the history of particles generated in the showers (while of course the energy
of the hits is fully accounted for). On the other hand, if the track in question
does satisfy our criteria (which is typically to leave hits in a tracking detector
such as TPC) we obviously need to save it. This leads to the situation where the
information about the parent of the saved track, is effectively being lost, and
we have a "disconnected vertex". This is flagged by the negative process ID.