Skip to main content

Vibrational frequency calculation: concept simplified?


Vibrational frequency calculation holds a key that can decide whether the obtained optimized structure is a minimum or a transition state or some thing else. I always wanted a simplistic adaptation with minimal understandable and required information that in a matter of minutes can help a new person understand the concept as easily as possible.

which makes it really easy for anybody. So, what we least need to know when you submit a frequency calculation.

In frequency calculation also called force calculation, we are solving a hessian matrix. What is a Hessian matrix?
The Hessian matrix is the matrix of second derivatives of the energy with respect to geometry.

What is the trouble in frequency calculation?

Although first derivatives are relatively easy to calculate, second derivatives are not. To overcome this problem, the double derivative is evaluated in terms of first derivatives. Here, initially for a given geometry, corresponding first derivative is evaluated. Then, the geometry is subjected to a finite displacement. After the SCF calculation for this displaced geometry, once again first derivatives are evaluated. The second derivatives are then calculated as a difference of the first derivatives for these two geometries divided by the step size. The unidirectional displacement based second derivative calculation is called single-sided derivative.

Owing to the sensitivity of the hessian towards geometry, they are calculated at stationary points only and using double-sided derivatives i.e. double derivatives calculated considering displacement in either direction w.r.t. initial geometry

The elements of the Hessian are defined as:
\begin{displaymath}H_{i,j} = \frac{\delta^2E}{\delta x_i\delta x_j}\end{displaymath}

Thus employing double sided derivative for the evaluation of each element is then done in two steps.

Each atomic coordinate xi is initially first incremented by a small amount and the gradient is calculated. Then the coordinate is decremented w.r.t. to the original position and again the gradient corresponding to this displacement is calculated. The second derivative is then obtained from the difference of the two derivatives and the step size.
\begin{displaymath}H_{i,j} = \frac{(\frac{\delta E}{\delta x_i})_{_{+0.5\Delta x......rac{\delta E}{\delta x_i})_{_{-0.5\Delta x_j}}}{\Delta x_j}.\end{displaymath}
This is done for all 3N Cartesian coordinates.

In order to calculate the vibrational frequencies, the Hessian matrix is first mass-weighted:
\begin{displaymath}H^m_{i,j} = \frac{H_{i,j}}{\sqrt{M_i*M_j}}.\end{displaymath}

Diagonalization of this matrix yields the force constants of the system.
Diagonalization of this matrix yields eigenvalues, e, (force constants ) which represent the quantities (k/m)1/2 , from which the vibrational frequencies can be calculated:
\begin{displaymath}\bar{\nu}_i = \frac{1}{2\pi c}\sqrt{\epsilon_i}.\end{displaymath}

P.S: I have used some part of the text from the webpage as it is because there was no better and simplified way to explain it.

Comments

Popular posts from this blog

Material Studio - selective molecule movement

So, working with Material Studio is awesome. You get to handle bigger, complicated and even unimaginable system. But I am facing problems in very minute functioning. I recently faced two problems, but managed to resolve on my own. Somehow, tutorials doesn't help in resolving them and you end up doing some exercise on your own to get the solution. Among these, the problem of bringing two molecular structures in one window and then adjusting one molecule by moving it around, while other structure remains untouched, as it is. Actually, this should not be a problem to do. There is a simple way of doing this. Select the molecule to be moved and then there is a symbol (button) to select, which allows you to selectively move that molecule only. This is the same way it can be done in Material studio too.  But among all the dozens of symbols, the new person handling MS, kind of gets lost. So, what he does is check the tutorials, for any help. But the tutorials don't talk about suc...

xleap and ionic systems

An extension of earlier problem in generation of prmtop and inpcrd files follows here. After a long time I had to use xleap to generate parameter and coordinate files for a system comprising of an organic molecule, protonated methanol and Cl atom. Now, organic molecule is not an issue. antechamber tutorial clearly explains how to do that. So, any organic molecule like MEOH is also not a problem. But MEOH2 could be. I have written about how to deal with systems like MEOH2 or extra protonated oxygen systems. But if you want to bring an organic molecule, protonated methanol(MEOH2) and Cl atom together and create parameter and coordinate file for that, it may not be as easy as it appear. Since I had lost touch with xleap for sometime, I totally forgot the small tricks I found to work in xleap. I started working with: * the separate pdb files for each molecular unit. * then create a single pdb file using packmol for the whole system. * create prepin or mol2 file corresponding to combined p...

Generating a structure PDB file using Gabedit

If Gabedit is installed and functioning on your system, type gabedit  in terminal. Gabedit will open and you can find an option "Geometry" in the tool bar. Select "Geometry"--> "Draw" After some trials and playing you will finally learn how to generate a structure. Once your structure is ready, you need to set the atom types and charges. For this click on "M" button in the top left of the "Draw Geometry" window and select options as: M-->"Set"--> "Atom Type & charge using PDB template" Here if you use "Atom Types using connection types", it helps. You can check if the atom type and charge has been set according to the amber requirements by an indirect method. Try to run a quick MM calculation. Run MM calculations by clicking: "M"-->Molecular Mechanics-->optimization It may sometimes end up with a pop up window that suggests that either atom type or charge is not...