Bad C-O single bond length for ethers

openforcefield / smirnoff99Frosst

A general small molecule force field descended from AMBER99 and parm@Frosst, available in the SMIRNOFF format

Creative Commons Attribution 4.0 International

28 stars 9 forks source link

Bad C-O single bond length for ethers #95

Open cbayly13 opened 5 years ago

cbayly13 commented 5 years ago

b15 with smarts [#6X4:1]-[#8X2H0:2] has an egregiously short length="1.370" ; could we please change this to an improved length="1.430" as a starting point?

cbayly13 commented 5 years ago

Here is an example molecule. DrugBank_2592: c1nc(c2c(n1)n(cn2)[C@H]3C@@HOP(=O)(O)O)O)N

trevorgokey commented 5 years ago

Just to support this, I scanned the "group1" torsiondrive for "O(C)C" and it shows that 1.37 is for esters rather than ethers.

davidlmobley commented 5 years ago

@cbayly13 in a draft refit by Yudong, he's ending up with this as output: <Bond smirks="[#6X4:1]-[#8X2H0:2]" length="1.430632479559e+00 * angstrom" k="6.441291139148e+02 * angstrom**-2 * mole**-1 * kilocalorie" id="b15" parameterize="k,length"/>

Your estimate of 1.43 looks like it was right on the money.

davidlmobley commented 5 years ago

(This paid no attention to this particular parameter/particular molecule -- it's a refit of all parameters based on the Roche dataset (torsion drives and geometry optimizations). Not our full dataset yet, but a good chunk of it.)