Large HSize

[PY 222]

I've been consistently getting larger and larger HSizes on my clients.

They were usually around 9 but now I am seeing it going all the way up to 2000+ and they are more common nowadays.

So, what does this translate to in terms of stats points and the science behind it?

[and_ped10]

Hi,

The HSize describes how large a subsystem that has to be analyzed when the code is running. In order to make the code as efficient as possible it is only the relevant/important information that is processed.

In the simulation certain atoms are locked (not allowed to move). The locked atoms should be a part of the simulation, but their position should not changed when the system evolves. To describe how the system evolves a matrix know as the Hessian is used. The variable HSize describe how large a Hessian (subsystem) that is being used to describe the current behavior. The size of the Hessian depent on different things and it is therefore hard to say that a certain HSize is correct.

Kindly Andreas

[graeme]

Very large HSize values do generally indicate that the search has gone bad. An HSize of 2000 means that many hundred of atoms are moving in the process. Low energy processes generally involve less than 20 moving atoms, which corresponds to an HSize of less than 100 ((moving atoms + neighbors) * 3).

There are some settings to control how aggressive the searches are. Cautious settings will reduce the number of wild searches, such as those with large HSize values, but this will also tend to reduce the varity of processes found. We will keep an eye on these stats and adjust the settings accordingly.