You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What we seem to be observing is that when you compute an excess property, and you do the calculation (for enthalpy for example), then with H_mixing = <H_mixture>- (MF_0<H_0>-MF_1<H_1>), the MF_0 and MF_1 are taken from ThermoML mole fractions, but it is more accurate to take them from the actually simulated mole fractions.
In our case, the ThermoML mole fractions are 0.5098 and 0.4902, and the simulations ended up being done with 510 and 490 molecules, as that is what 509.8 and 490.2 get rounded to.
In this case, one sees that if the molar enthalpy of component 0 is -44.879 and of component 1 is 251.418, and you get a mixture enthalpy 98.855, then if you use the simulation MF, you get (0.510)*(-44.879) + (0.490)251.418 - 98.855 = -1.45153, whereas if you use the ThermoML MF, you get (0.5098)(-44.879) + (0.4902)251.418 - 98.855 = -1.511, which a difference almost about 2x than the uncertainty; this -1.511 number is what Evaluator actually returned. If the ThemoML MF had been, say 0.4904/0.5096, then you would have had a simulation of 1000 molecules still rounded to 490/510, and using the ThermoML MF would have been (0.5096)(-44.879) + (0.4904)*251.418 - 98.855 = -1.57.
I would argue that the most important thing is to use the mole fraction consistent with the mole fraction of the mixture solution actually simulated, as you want the ideal solution composition consistent with the mixture composition. The H_mix(x) is not going to change very much when changing by x by 0.002, but mismatches result in larger errors. More specifically, because the ideal components of the enthalpy are ~1-2 order of magnitude larger than the nonideal components, mismatching the ideal and nonideal solutions will give errors 1-2 orders of magnitude large than the changes in the nonideal contribution itself.
To Reproduce
Any calculation should should work to show this error, but ones with smaller numbers of molecules so the rounding will give a larger difference.
Output
No error in output log per se. We figured this out by reproducing the calculation of enthalpy and figuring out mole fraction is being used.
Computing environment (please complete the following information):
I don't specifically recall the version of Evaluator - within the last 6 months.
Additional context
The text was updated successfully, but these errors were encountered:
Describe the bug
What we seem to be observing is that when you compute an excess property, and you do the calculation (for enthalpy for example), then with H_mixing = <H_mixture>- (MF_0<H_0>-MF_1<H_1>), the MF_0 and MF_1 are taken from ThermoML mole fractions, but it is more accurate to take them from the actually simulated mole fractions.
In our case, the ThermoML mole fractions are 0.5098 and 0.4902, and the simulations ended up being done with 510 and 490 molecules, as that is what 509.8 and 490.2 get rounded to.
In this case, one sees that if the molar enthalpy of component 0 is -44.879 and of component 1 is 251.418, and you get a mixture enthalpy 98.855, then if you use the simulation MF, you get (0.510)*(-44.879) + (0.490)251.418 - 98.855 = -1.45153, whereas if you use the ThermoML MF, you get (0.5098)(-44.879) + (0.4902)251.418 - 98.855 = -1.511, which a difference almost about 2x than the uncertainty; this -1.511 number is what Evaluator actually returned. If the ThemoML MF had been, say 0.4904/0.5096, then you would have had a simulation of 1000 molecules still rounded to 490/510, and using the ThermoML MF would have been (0.5096)(-44.879) + (0.4904)*251.418 - 98.855 = -1.57.
I would argue that the most important thing is to use the mole fraction consistent with the mole fraction of the mixture solution actually simulated, as you want the ideal solution composition consistent with the mixture composition. The H_mix(x) is not going to change very much when changing by x by 0.002, but mismatches result in larger errors. More specifically, because the ideal components of the enthalpy are ~1-2 order of magnitude larger than the nonideal components, mismatching the ideal and nonideal solutions will give errors 1-2 orders of magnitude large than the changes in the nonideal contribution itself.
To Reproduce
Any calculation should should work to show this error, but ones with smaller numbers of molecules so the rounding will give a larger difference.
Output
No error in output log per se. We figured this out by reproducing the calculation of enthalpy and figuring out mole fraction is being used.
Computing environment (please complete the following information):
I don't specifically recall the version of Evaluator - within the last 6 months.
Additional context
The text was updated successfully, but these errors were encountered: