I am optimizing some structures at the doublet potential energy surface. I know that the eigenvalue of operator S2 should be near to 0.75 on the doublet potential energy surface (S(S+1)=(0.5(0.5+1)=0.75). But, my structures show 1.45 for the eigenvalue of operator S2. How can I eliminate these spin contaminations? Does anyone have experience regarding the elimination of the spin contamination?
Generally, spin contamination problem can occur high spin state system invovling transition metal (ex Mn, Ni, etc)
There is option to eliminate spin-contamination
You can insert qc keyowd with scf but you should know that time cost is much higher than any option.
My suggestion is that you can run a job within 2 or 3 step employing scf=qc then, restart normal calculation from previous geometry
Hello Moein,
what computational approach do you use? Quite often, DFT can reduce spin contamination.
Best regards,
Aleksey.
Hello Aleksey
I am optimizing my structures at the B3LYP/6-31G(d) level.
I have newly found two important points about spin contamination. First, if we optimize our structures at the Restricted Open-shell (RO) level instead of unrestricted Open-shell (UO), we will have less spin contamination (ROB3LYP instead of UB3LYP). But, using Restricted Open-shell (RO) level increases the cost of calculations.
Second, big spin contamination is warning regarding unfavorable distribution of spin density. In this step, my suggestion is to start the optimization from another input.
I will be very grateful to know other opinion about spin contamination.
The cost of calculation at RO level is higher than restrict calculations. Unrestricted wave function will double the size of your basis set, however, after an UO calculus it is possible to project out the spin contamination using a spin projection methods. At MP2 level Gaussian does the projection automatically. Now, I am working with charged species at doublet state at M06-2X level and the spin contamination is negligible. Maybe you should try to change your DFT method to verify if the spin contamination decreases.
Generally, spin contamination problem can occur high spin state system invovling transition metal (ex Mn, Ni, etc)
There is option to eliminate spin-contamination
You can insert qc keyowd with scf but you should know that time cost is much higher than any option.
My suggestion is that you can run a job within 2 or 3 step employing scf=qc then, restart normal calculation from previous geometry
Hello,
How to estimate spin contamination particularly triplet spin contamination from the TD files.
- Badges
- Science topic