how to plot minimum energy path, after getting saddle point
Moderator: moderators
how to plot minimum energy path, after getting saddle point
Dear all,
Now I have a question about how can I draw the minimum energy path after I obtain the saddle point with the DIMER method, i.e, I know the initial point and the saddle point, but now how to konw the end point?
I have read carefully the paper
"A dimer method for finding saddle points on high dimensional potential surface using only first derivatives" J. Chem. P 111, 7010, 1999,
and find the following sentence:
Once the dimer has converged to a saddle point, it is easy to trace out the minimum energy path. ...
how to understand this sentence, and how to reach my goal?
Thanks a lot!
Now I have a question about how can I draw the minimum energy path after I obtain the saddle point with the DIMER method, i.e, I know the initial point and the saddle point, but now how to konw the end point?
I have read carefully the paper
"A dimer method for finding saddle points on high dimensional potential surface using only first derivatives" J. Chem. P 111, 7010, 1999,
and find the following sentence:
Once the dimer has converged to a saddle point, it is easy to trace out the minimum energy path. ...
how to understand this sentence, and how to reach my goal?
Thanks a lot!
I would use IBRION = 3 (damped dynamics) with POTIM <= 0.1. That should map out something relatively close to the MEP. After that, if you want to, you can extract a few approximately equally spaced configurations from the XDATCAR with xdat2pos.pl and use them as an initial guess for a further NEB calculation.
1) The two images should lie along the unstable mode. So use the two CONTCARs along with the interpolation script like:
interpolate.pl 01/CONTCAR 02/CONTCAR 1.1
for one direction and
interpolate.pl 02/CONTCAR 01/CONTCAR 1.1
for the other. Note that 1.1 may not be enough, you need to experiment with
that. Refer to the interpolate.pl documentation on http://theory.cm.utexas.edu/vtsttools/scripts/ for more info.
2) Just run a regular VASP minimization (ICHAIN = 0) for each image. The images
are now decoupled and have no idea of each other existence.
Andri
interpolate.pl 01/CONTCAR 02/CONTCAR 1.1
for one direction and
interpolate.pl 02/CONTCAR 01/CONTCAR 1.1
for the other. Note that 1.1 may not be enough, you need to experiment with
that. Refer to the interpolate.pl documentation on http://theory.cm.utexas.edu/vtsttools/scripts/ for more info.
2) Just run a regular VASP minimization (ICHAIN = 0) for each image. The images
are now decoupled and have no idea of each other existence.
Andri
Thanks your replies!
I browse the page of http://theory.cm.utexas.edu/vtsttools/scripts/, and find the usage of interpolate
interpolate.pl (POSCAR 1) (POSCAR 2) (fraction)
output: POSCAR.out file, to STDOUT
here, "fraction" is just the distance between the image and the saddle point ?
I browse the page of http://theory.cm.utexas.edu/vtsttools/scripts/, and find the usage of interpolate
interpolate.pl (POSCAR 1) (POSCAR 2) (fraction)
output: POSCAR.out file, to STDOUT
here, "fraction" is just the distance between the image and the saddle point ?
hi, all. I do some tests but still have something unclear.
1) in my optinion, the distance between the two images in 01 and 02, after finding the saddle point, is very short, so if fraction=1.1, the command :
interpolate.pl 01/CONTCAR 02/CONTCAR 1.1
will make a image quite near the CONTCAR-02.
For example, in my work, the distance between CONTCAR-01 and CONTCAR-02 is only 0.01 Angstrom. So I use fraction = 100, 200 ,300, and so on, to make a serial of images, then repeat the process for the opposite direction. What I do is correct?
2)In my work, I use the perl scripte "dist.pl" to get that the distance between the initianl state A and the final state B is about 9 Angstrom, and the distance between A and the saddle point is 3 Angstrom, but the distance between B and the saddle point is 15 Angstrom! Whether it implies that the saddle point is not between A and B?
Thank you any words!
1) in my optinion, the distance between the two images in 01 and 02, after finding the saddle point, is very short, so if fraction=1.1, the command :
interpolate.pl 01/CONTCAR 02/CONTCAR 1.1
will make a image quite near the CONTCAR-02.
For example, in my work, the distance between CONTCAR-01 and CONTCAR-02 is only 0.01 Angstrom. So I use fraction = 100, 200 ,300, and so on, to make a serial of images, then repeat the process for the opposite direction. What I do is correct?
2)In my work, I use the perl scripte "dist.pl" to get that the distance between the initianl state A and the final state B is about 9 Angstrom, and the distance between A and the saddle point is 3 Angstrom, but the distance between B and the saddle point is 15 Angstrom! Whether it implies that the saddle point is not between A and B?
Thank you any words!
I agree that a fraction of 1.1 is too small, because the distance between the points p1 and p2 from a dimer calculation is on the order of 0.01 Ang. To get initial points on either side of the saddle, I use a fraction of 10:
interpolate p1 p2 10
interpolate p1 p2 -10
If you generate these two images on either side of the saddle, you can then minimize the structures to get the initial and final states.
I can't see why you would need to generate a series of images, using fractions of 100, 200 and so forth. These will simply generate a set of images along the unstable mode at the saddle. These will not get you to the initial and final state minima.
You can't conclude too much from the distances. You can have situations in which the initial state is closer to the final state then it is to the saddle point. Using distance as a simple measure of the reaction coordinate really only applies if atoms move in a straight line between the initial and final states.
interpolate p1 p2 10
interpolate p1 p2 -10
If you generate these two images on either side of the saddle, you can then minimize the structures to get the initial and final states.
I can't see why you would need to generate a series of images, using fractions of 100, 200 and so forth. These will simply generate a set of images along the unstable mode at the saddle. These will not get you to the initial and final state minima.
You can't conclude too much from the distances. You can have situations in which the initial state is closer to the final state then it is to the saddle point. Using distance as a simple measure of the reaction coordinate really only applies if atoms move in a straight line between the initial and final states.
Thanks your replies! I am clear about the choice of fraction.
In fact, I use the DIMER method in my work to explore the diffusion path for a rotation between the metastable state A and the final state B. After finding the saddle point for a relatively lowest-energy diffusion path, I want to plot the fig for the relation between "energy and reaction coordinate".
So I want to know
1) how to determine the reaction coordinate for the saddle point relative to A or B. Because the state of A is not equivalent to B, I cann't simply scale the reaction coordinate of A (B) to -1(+1), and that of the saddle point to 0.
2) how to understand the method described in the paper of J. Chem. P 111, 7010, 1999,
The dimer is first allowed to rotate into the lowest curvature mode, the unstable mode, so that it is aligned along the reaction coordinate.
An image is placed on one side of the dimer along the direction Nˆ 1 . The distance of the image from the midpoint of the dimer (the saddle point) is chosen according to the desired resolution of the path. In a manner very similar to the algorithm used to rotate the dimer, the energy of this image
is minimized while keeping its distance from the previous image (in this case the saddle point) fixed. This procedure is repeated, each time placing a new image initially along the local path (the line between the two previous images) to minimize the number of function calls required to zero the tangential force on the new image. The process is stopped
when the minimum energy of an image is greater than that of the previous image. After the path is traced out in one direction from the saddle point to a minimum, the opposite direction must be followed to complete the minimum energy path.
I thank you for your patient to answer my question!
In fact, I use the DIMER method in my work to explore the diffusion path for a rotation between the metastable state A and the final state B. After finding the saddle point for a relatively lowest-energy diffusion path, I want to plot the fig for the relation between "energy and reaction coordinate".
So I want to know
1) how to determine the reaction coordinate for the saddle point relative to A or B. Because the state of A is not equivalent to B, I cann't simply scale the reaction coordinate of A (B) to -1(+1), and that of the saddle point to 0.
2) how to understand the method described in the paper of J. Chem. P 111, 7010, 1999,
The dimer is first allowed to rotate into the lowest curvature mode, the unstable mode, so that it is aligned along the reaction coordinate.
An image is placed on one side of the dimer along the direction Nˆ 1 . The distance of the image from the midpoint of the dimer (the saddle point) is chosen according to the desired resolution of the path. In a manner very similar to the algorithm used to rotate the dimer, the energy of this image
is minimized while keeping its distance from the previous image (in this case the saddle point) fixed. This procedure is repeated, each time placing a new image initially along the local path (the line between the two previous images) to minimize the number of function calls required to zero the tangential force on the new image. The process is stopped
when the minimum energy of an image is greater than that of the previous image. After the path is traced out in one direction from the saddle point to a minimum, the opposite direction must be followed to complete the minimum energy path.
I thank you for your patient to answer my question!
The method of tracing out the minimum energy path has not been implemented in vasp. It is not really critical for finding rates.
If you just have your initial, saddle and final states, you can make a plot of these with a reaction coordinate (distance) by using the information from the three points as if they were a nudged elastic band calculation. You can either make an neb.dat file (energies, force along the path (=0), and distance) by hand and use the spline script, or put the POSCAR and OUTCAR files in 00,01,02 directories, and use our nebresults script to make a figure of the minimum energy path with the 3 known points on it.
If you just have your initial, saddle and final states, you can make a plot of these with a reaction coordinate (distance) by using the information from the three points as if they were a nudged elastic band calculation. You can either make an neb.dat file (energies, force along the path (=0), and distance) by hand and use the spline script, or put the POSCAR and OUTCAR files in 00,01,02 directories, and use our nebresults script to make a figure of the minimum energy path with the 3 known points on it.