Alchemical ion improvements

[lohedges]

1. Automatically keep the charge constant unless the user has specified a value, warning if value doesn't match.
2. If present, use existing ion parameters when creating the alchemical ion.

(I'm just finding a system to test this on locally.)

[mb2055]

I don't think this automatically keeps the charge neutral. If i pass a system with a charge difference of 1 without specifying --charge-difference I simply get the warning WARNING | somd2.runner._runner:__init__:187 - The charge difference of 1 between the end states does not match the specified value of 0..
Should it not automatically detect that there is a difference of 1 and create a corresponding co-alchemical ion?

[lohedges]

That should just be a warning that it doesn't match the specified value. Since the user hasn't set anything in this case (the default is zero) it should still create the ion. Do you not see anything else in the log, e.g. Water at molecule index.... I'll have a check too.

[lohedges]

No, you're right, it's not automatically updating. Will debug.

[lohedges]

I was using the wrong variable for the charge difference in one of the conditionals. Should hopefully work now. For example, if the actual charge difference is 1 but I specify 3, then I see:

2024-08-09 15:41:59.202 | WARNING  | somd2.runner._runner:__init__:181 - Charge difference between end states is not an integer.
2024-08-09 15:41:59.203 | WARNING  | somd2.runner._runner:__init__:187 - The charge difference of 1 between the end states does not match the specified value of 3
2024-08-09 15:41:59.211 | INFO     | somd2.runner._runner:_create_alchemical_ions:450 - Water at molecule index 489 will be perturbed to Cl- to keep charge constant.
2024-08-09 15:41:59.215 | INFO     | somd2.runner._runner:_create_alchemical_ions:450 - Water at molecule index 101 will be perturbed to Cl- to keep charge constant.
2024-08-09 15:41:59.219 | INFO     | somd2.runner._runner:_create_alchemical_ions:450 - Water at molecule index 861 will be perturbed to Cl- to keep charge constant.

But if I don't specify anything I get:

2024-08-09 15:43:39.006 | WARNING  | somd2.runner._runner:__init__:181 - Charge difference between end states is not an integer.
2024-08-09 15:43:39.006 | WARNING  | somd2.runner._runner:__init__:187 - The charge difference of 1 between the end states does not match the specified value of 0
2024-08-09 15:43:39.016 | INFO     | somd2.runner._runner:_create_alchemical_ions:450 - Water at molecule index 489 will be perturbed to Cl- to keep charge constant.

[mb2055]

Yes, it works now.
The only issue I foresee is inconsistency when charge-difference is specified with a value of 0.
Currently if I have a system with a non-zero charge difference and I specify somd2 Input.bss --pert-file Input.pert --charge-difference 0 the code will automatically introduce an alchemical ion:

2024-08-09 15:44:09.323 | WARNING  | somd2.runner._runner:__init__:181 - Charge difference between end states is not an integer.
2024-08-09 15:44:09.323 | WARNING  | somd2.runner._runner:__init__:187 - The charge difference of 1 between the end states does not match the specified value of 0
2024-08-09 15:44:09.332 | INFO     | somd2.runner._runner:_create_alchemical_ions:450 - Water at molecule index 489 will be perturbed to Cl- to keep charge constant.

This, as far as I can tell, makes it impossible to run without an alchemical ion in systems with a charge difference, which could be an issue.

[lohedges]

Ah yes, I need to think of the best way of checking that the value isn't default. And the user has explicitly set it to zero. This might be tricky, since I can imagine input coming from the command-line, YAML file, etc. We'd need to make sure not to write a default value to the YAML, since if it was re-read then it would appear as having been set, etc.

[lohedges]

I'll also add some checks for sensible numbers for the charge difference, e.g. so that it doesn't exceed the number of water molecules. (Not that they should be doing that anyway.)

[lohedges]

Just defaulting to None for the charge_difference should fix things. I've also added a check for the number of water molecules.

[mb2055]

All good 👍