3.320  Atomistic Comp Model:Materials

Dear Class,

Over the past couple of days I've received several questions about the lab. For those of you still working, I've placed a FAQ under the "LAB 5"  assignment link in the course materials section of the website.

Good luck !

Chris

Hi Class,

As you work on Lab 5, feel free to email me with any questions or concerns. I will post questions and answers of general interest in the form of a FAQ on the class website.

Note that while a single MC trajectory runs quite fast, many will be needed to determine the phase diagram.

Good luck !

Chris

Dear Class,

LAB 4 due date is extended to 5/9/2008 (Next Friday). If you have any questions, please send us email (Oliviero Andreussi: oliviero@mit.edu and Xiaofeng Qian: qianxf@mit.edu). Please also check the FAQ on the cluster.
 
Oliviero and Xiaofeng

Hi Class,

Problem set 1 has been graded and the grades have been posted on Stellar.  If you turned in a physical copy, comments can be found on the physical copy.  Otherwise you should see comments in the comments section on Stellar.

A few general notes about the problem set:

• The potential models got the lattice parameter right because they were parameterized to reproduce the lattice parameter.  In general potential models do very well on the tasks for which they are designed, and getting the lattice parameter right is usually one of those.
• The lattice parameter is not the same as the bond length.  The lattice parameter is the length of a periodically repeated unit, and the bond length is the distance between nearby nuclei.
• If you are asked to do something accurately, relax the atoms.  This is especially true when using potential models and relaxation is very cheap.  Generally the positions of the nuclei should be frozen only if a particular unstable state interests you or if relaxation is prohibitively expensive.
• Relative error can be misleadingly small, given that it is dependent on the reference state.  For example, the difference of an energy of 9.5 eV and and energy of 9.6 eV might seem small, but what if you are comparing it to state with energy of 9.55 eV?  Absolute error is usually a better metric, and errors as small as a few meV / atom can be important.

Let me know if there are any questions.

Tim

Dear Class,

I have created a text file "FAQ" at /home/users/mit3320/LAB4/FAQ. It contains some tips for Lab 4. If you have any questions, you can send email to both me (qianxf@mit.edu) and Oliviero (oliviero@mit.edu). We will collect the information, fix the problem and update the above FAQ file. So please check the FAQ file frequently before you run the code.

Here are some brief summary of FAQs. For the detailed information please check the above file.

FAQ 0:
Please check your data frequently, especially the size of your directory by the following command: "du -sh /home/users/mit3320xxxx/LAB4", where "mit3320xxxx" should be your account on the cluster. If you find it is too large (more than 2G), please first stop the running program and then check your input file. You may increase "print-interval" and "dump-interval". Or let us know and check it for you.

FAQ 1:
When you do the problem 1a to test different timesteps, please check the timestep from 0.0001ps to 0.01ps. Moldy program above 0.01ps may generate huge output file and jam the /home/users/ directory on the cluster. And indeed timestep larger than 0.01ps is not appropriate for the MD simulation of Xenon system.

FAQ 2:
The "rdf-limit" should be smaller than the supercell size, otherwise, moldy will  stop with errors. The "cutoff" for Lennard-Jones potential should be smaller than half of supercell size to avoid image effect for the same particle.

FAQ 3:
"dump-interval" in the old "run_moldy.j" is changed to be "500" to avoid the large dump file.

Best,

Xiaofeng