H-P bonds (not in parm99/parm@Frosst)

[bannanc]

Looking at GAFF2 for inspiration. It has a dedicated hydrogen type for hydrogen bound to phosphorous (hp).

Here are the H-P bonds:

hp-p2  266.54   1.336                     SOURCE3      87    0.1706
hp-p3  200.57   1.412             SOURCE3_SOURCE5     123    0.0510
hp-p4  253.60   1.349                     SOURCE3      17    0.1577
hp-p5  196.39   1.418             SOURCE3_SOURCE5      31    0.0077

Relevant phosphorous atom types

p2 30.97         1.538               Phosphate with two connected atoms
p3 30.97         1.538               Phosphate with three connected atoms, such as PH3
p4 30.97         1.538               Phosphate with three connected atoms, such as O=P(CH3)2
p5 30.97         1.538               Phosphate with four connected atoms, such as O=P(OH)3

I'm having a hard time making a decision about the hydrogen-phosphorous bond, but it seems like it would be better to keep just 1 H-P bond for now.

[davidlmobley]

In the absence of any deep chemical insight on my part, I'm tempted to go with a length of about 1.38 and a force constant which is the average of the ones listed. :) But if you

This is one it would be interesting to revisit when we explore partial bond order -- I wonder if the force constant and length differences GAFF2 is attempting to capture here are something which is captured by the semiempirical calculation so we could just interpolate smoothly...

[bannanc]

@davidlmobley I pretended like I remembered o-chem yesterday and tried to draw out some structures and think about what would make the H-P bond stronger. Naively I assume something in the p2 and p4 types makes it more likely to polarize the bond making it stronger, otherwise H and P have similar electronegativities. I also tried to make guesses about which of these phosphorous is more common. It seems like there are possibly two distinct groups here, I think it makes more sense to round one of them knowing that might be bad for some case and good for others, instead of being bad for everything.

[#1:1]-[#15:2]    200.    1.4

[davidlmobley]

Sounds good to me, @bannanc . The idea of two distinct groups means this would also be a good case for follow up with QM in the not-too-distant future, which would be nice.

[bannanc]

In pull request #43