# [Materials studio]geometry optimization

#### gungunlin

Dear friends:
anyone who used the materials studio software before?
these days, i used the software to do the geometry optimization of a compound's structure.
it has already taken about 30hours to do the optimization job till now and it has not finished yet. i am really anxious and don't know when it will finish the job.
there is a chart called "Dmol3 optimization convergence" while processing the optimization calculation. In the convergence chart, there are 3 lines, energy change, max force, and max displacement. do you know how to read the chart? and does it yield the information when the calculation will be finished?
thanks a lot! and hope you can solve the problem for me!

Related MATLAB, Maple, Mathematica, LaTeX, Etc News on Phys.org

#### alxm

I've never used Materials Studio. Haven't even heard of it before, I think.
Looking at their web page, "Dmol3" (which I'm unsure whether I've heard of either, despite being in the quantum-chemical field) is allegedly a DFT method. I'll wager a guess that it's B3LYP* or a variant of it.

You've left out some very critical information here: What's the size of your system and what kind of machine are you running on? Regardless, 30 hours is not a long amount of time. A system of a few hundred atoms could easily take weeks on a supercomputing cluster.

Geometry optimization is a simple thing. You calculate the energy, then you calculate the gradient w.r.t. nuclear coordinates (i.e. the forces on the atoms)**, then you use that to try to minimize the energy w.r.t. the nuclear coordinates, using a gradient descent method or similar. What you're seeing plotted is the total energy, the average forces, and the average displacement (i.e. how much it's moved the atoms).

Ideally, this should all converge nicely and smoothly. The energy should get lower (in general) and the forces and displacement get smaller. The optimization is done whenever the changes get below certain convergence limits which you can also set. Typically around 1E-5 Hartrees. What you want to look out for is indications that the thing isn't converging. That would appear as large spikes in the energy/force, or periodic behaviour (i.e. converges to a point, then returns to another point and converges to the same point again).

* Becke's exchange functional hybridised in three parameters with Hartree-Fock exchange and the Lee-Yang-Parr correlation functional
** The Hellman-Feynman theorem at work.

Last edited:

### Physics Forums Values

We Value Quality
• Topics based on mainstream science
• Proper English grammar and spelling
We Value Civility
• Positive and compassionate attitudes
• Patience while debating
We Value Productivity
• Disciplined to remain on-topic
• Recognition of own weaknesses
• Solo and co-op problem solving