Thank you very much for your responses!
@Adrian: I did suspect a method-related instability causing the spikes
but I didn’t think of the reasoning you suggested.
It sounds quite plausible and I would be extremely grateful if you could
point me to the right literature that suggests this possibility. I read
some papers that suggested that adding up the kinetic term nullifies
the spikes but that doesn’t apply in my case since I checked that there
are no atom crossings.
@Steve: I am dynamically allocating memory for the vector (of a
static length) in the constructor of the compute style and it gets
deallocated in the destructor. There is no other vector/array that is dynamically allocated inside the compute.
@Axel: Thanks for your suggestions! I will get back to you after trying them
out.
Thank you all once again!