Skip to content

shuozhixu/LAMMPSatOU

BranchesTags
 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

89 Commits
gsfe
gsfe
 
 
peierls
peierls
 
 
README.md
README.md
 
 
lmp_cms.sh
lmp_cms.sh
 
 

Repository files navigation

Introduction

My name is Anshu Raj, and I am a graduate research assistant in the Computational Materials, Mechanics, and Manufacturing (CM3) lab at the University of Oklahoma (OU). In this repository, I will show you how to run atomistic simulations via LAMMPS on the OU Supercomputing Center for Education and Research (OSCER). This page was created based on LAMMPSatUCSB, with the help of Dr. Shuozhi Xu, the PI of the OU CM3 lab.

OSCER

First, request a user account at OSCER. While requesting for an OSCER account, one needs to specify his/her group name. If you are a member of the CM3 lab, specify cm3atou as your group name. For all other new users, use your own group name. If you use an incorrect group name in your request, your request may be delayed or even denied.

Wait for the user account to be approved --- you will receive an email. In what follows, I will assume that your account is username and your password is username-pw.

To use OSCER, these webpages may help:

Usually, your $HOME directory has a data quota of 20 GB. But each cm3atou group member has a larger quota of 30 GB. To check how much space you have taken, execute the following command in $HOME:

du -sh

If the quota is exceeded, you won't be able to write anything to $HOME, and most likely all your jobs will stop. In that case, you can take advantage of the directory /scratch/username/, where files are automatically deleted after 2 weeks. Note that /scratch is a common space where everyone on OSCER can use it, until it is full. One may check his/her usage of the scratch partition by

df /scratch/username/

There are three ways to use /scratch/username/:

  • Submit jobs in $HOME and output data there. Then move some or all data to /scratch/username/;
  • Submit jobs in /scratch/username/ directly;
  • Add #SBATCH --chdir=/scratch/username/ in the *.batch file, and then all output data will be written to that directory.

In addition, members in the cm3atou group can also move files to

/ourdisk/hpc/cm3atou/dont_archive/username/

where the maximum storage space for the cm3atou group is 55.8 TB. Unlike in /scratch, files in /ourdisk won't be deleted. For large files (> 1GB), one can archieve files on OURRstore. For more information, please visit this page.

The cm3atou group currently has ten nodes, each of which has 128 CPU cores (with hyperthreading) and 257101 megabytes of memory. Jobs have no time limit. In practice, the time limit would be until the next scheduled maintenance outage.

Module

Many softare programs are already installed on OSCER. You can check what modules are available by

module av

If you want to know about a specific software program, e.g., LAMMPS, type

module av lammps

or

module spider lammps

To load a specific program so that you can use it, type

module load XXX

where XXX is the program name as shown by module av or module spider. To check what modules you have installed, type

module li

Note: As a general rule, the module load XXX command needs to be typed each time you log into your OSCER account. To avoid that, write it in both .bashrc and .bash_profile files in your $HOME, provided that you are using bash. Then the next time you log in, you don't need to retype module load XXX.

OU VPN (optional)

Unlike some other universities, you do not need the OU VPN to connect to OSCER.

FTP client

You need a file transfer protocol (FTP) client to transfer data between OSCER and your local computer. Feel free to use any FTP client. Here is a selected list.

I personally recommend FileZilla. Below is an instruction:

  1. Download and install Filezilla Client on your local computer.
  2. Open it.
  3. The first time you use it, File --> Site Manager --> New site --> rename it 'OSCER', then in the window on the right hand side:
    • Protocol: SFTP - SSH File Transfer Protocol
    • Host: schooner.oscer.ou.edu
    • Logon Type: Normal
    • User: username
    • Password: username-pw
    • Connect
  4. The next time you use it, File --> Site Manager --> select 'OSCER', then 'Connect'.
  5. To transfer files between OSCER and your local computer, please refer to this page.

Terminal emulator

You also need a terminal emulator to 'talk with' OSCER, e.g., submit a job. Feel free to use any terminal emulator. Here is a selected list.

On Windows, OU recommends MobaXterm and PuTTY.

On Mac and Linux, without installing any new emulator, you may open the default terminal and type

ssh [email protected]

then hit Return. Then you will be asked to provide your password. Type your own password, e.g.,

username-pw

then hit Return.

Hint: Type the password anyway even though nothing is showing up.

To check the status of the cm3atou partition, type the following in your terminal,

sinfo -p cm3atou

To check the list of users that are currently running jobs on the cm3atou partition,

squeue -p cm3atou

To add the number of cores for each job to the list above,

squeue -p cm3atou -t all --Format=jobid:10,username:14,statecompact:6,numcpus:6,minmemory:8,timeused:12,timelimit:12,nodelist:8

To check the status of all jobs of yours that are either running or pending,

squeue -u username

Then the first column is the JOBID.

To cancel a running job,

scancel JOBID

To find out more about a running job, including the direcotry where you submitted it,

scontrol show job JOBID

To find out the specification of a node,

scontrol show nodes NODEID

If you are not familiar with Linux, please refer to these webpages:

You also need a software package to edit text files on OSCER. Again, feel free to use anything. Here is a selected list. I recommend vim, which is already installed in OSCER (and most, if not all, Mac and Linux systems). If you are not familiar with vim, please refer to these webpages:

LAMMPS

LAMMPS is an open-source software package for atomistic simulations. So you first need to understand how atomistic simulations work. There are three main types of atomistic simulation methods

To learn the basics of MD and MS, please read, respectively, Chapter 9 and Chapter 6 of this book. The Google Drive link is private, so you need to request access.

And here are more references on MD:

To learn LAMMPS, you may start with this page and this page.

Multiple versions of LAMMPS are installed on OSCER (try module spider lammps as mentioned earlier). However, we prefer to install the latest version ourselves. To do that, download the file lmp_cms.sh from this GitHub repository to your local computer, then upload it, via Filezilla, to your $HOME on OSCER. Then in your terminal emulator, cd to $HOME, and execute

sh lmp_cms.sh

Note that the fourth command in lmp_cms.sh loads a module. If you cannot load it, try module purge first.

Once the sh run is finished, you will find a file lmp_mpi in the software/lammps-cms/src/ directory on OSCER. And that is the LAMMPS executable with MANYBODY, EXTRA-COMPUTE, MC, and MISC packages.

If you need additional packages, you can add them yourself by editing the file lmp_cms.sh, i.e., adding more mpi yes-XXX before make mpi .

Example 1: Calculating the GSFE curve in a BCC metal

First, to understand the generalized stacking fault energy (GSFE) curve, read these materials:

The GSFE curve is just one curve taken from the GSFE surface, also known as the γ-surface, which is usually calculated in FCC metals. To know more about the GSFE surface, please read

Running this specific LAMMPS simulation requires five files. The first four files can be found in the gsfe directory in this GitHub repository, including

  • lmp_gsfe.batch, which is for job submission
  • lmp_gsfe.data, which is the LAMMPS data file
  • lmp_gsfe.in, which is the LAMMPS input file
  • gsfe_curve.sh, which is the post-processing bash script

The fifth file is

  • MoNbTi_A_atom.eam.alloy, which is the interatomic potential file and can be downloaded from this page

First, download all five files from GitHub to your local computer.

Then on OSCER, create a new directory in your $HOME. Say the directory is named oscer_gsfe. The command is

mkdir oscer_gsfe

Then upload, via Filezilla, the five files from your local computer to oscer_gsfe on OSCER.

Then, in your terminal emulator, type

cd oscer_gsfe

then hit Return. Then submit the job by typing

sbatch lmp_gsfe.batch

then hit Return. To check the status of the job, type

squeue -u username

then hit Return. You will see two lines. In the first line, there is a term ST, which stands for 'status'. If, at the same location of the second line, you see PD, the job is pending. Recheck the status later. If you see R, the job is running.

Once the job starts running, feel free to disconnect from OSCER or turn off your local computer. Your job will continue running on OSCER until it is finished by itself or you cancel it using scancel JOBID.

If you only see one line following squeue -u username, the job is finished. This, however, can mean one of the two things:

  • The job was finished because of an error. In this case, check these three files: lmp_gsfe.out, lmp_gsfe.err, and log.lammps. They provide you information on what caused the error(s). In particular, the last file is the log file of LAMMPS, which would present an error message in the last line. Please refer to this page for the explanation of each error message. Once you figure out what went wrong, fix the problem, and resubmit the job
  • The job was finished successfully. In this case, the last line of the log file should contain Total wall time. Then proceed to the next step.

You will find a lot of files in the directory. One file is called gsfe_ori. In the same directory on OSCER, type

sh gsfe_curve.sh

then hit Return. You will find a new file called gsfe. The first and second columns of this file, respectively, are the x and y axes of the first half of the 'MoNbTiA' curve in Figure 2(a) of this paper. Use Filezilla to download gsfe to your local computer, plot it, and see if you get the same curve.

As usual, feel free to use any software to plot the curve. Here is a selected list. I recommend Matplotlib.

Now, go back to the file lmp_gsfe.in and read it. Look up the meaning of each LAMMPS command on this page.

Note: Only the cm3atou group members can use the partition cm3atou in the batch file. If one were to use an OU-wide partition, change the partition name in lmp_gsfe.batch to something else.

OVITO

In the directory on OSCER, oscer_gsfe, you will find a lot of dump files, which contain information of atomic positions. To visualize these files, download them, via Filezilla, to your local computer. Then install OVITO on your computer. Read this page to learn how to use it.

Example 2: Calculating the Peierls stress of a screw dislocation in a BCC metal

First, to understand dislocations, I recommend these readings depending on how much you already know and how much more you want to know.

Some Google Drive links above are private. You may request access.

To learn all kinds of defects in crystals, read the website by Föll and/or the book by Cai and Nix.

Running this specific LAMMPS simulation requires four files. The first three files can be found in the peierls directory in this GitHub repository, including

  • lmp_peierls.batch, which is for job submission
  • lmp_peierls.data, which is the LAMMPS data file, containing a screw dislocation on the {112} plane
  • lmp_peierls.in, which is the LAMMPS input file

The fourth file is

  • MoNbTi_A_atom.eam.alloy, which is the interatomic potential file and can be downloaded from this page

Make sure that you have all four files in your local computer, which you should after you tried the first example.

Then on OSCER, create a new directory, oscer_peierls, in your $HOME, by typing

mkdir oscer_peierls

then hit Return. Then upload, via Filezilla, the four files from your local computer to oscer_peierls on OSCER.

Then, in your terminal emulator, type

cd oscer_peierls

then hit Return. Then submit the job by typing

sbatch lmp_peierls.batch

then hit Return.

After the job is finished, you will find a new file called strain-stress. The first and second columns of this file, respectively, are the yz components of the strain tensor and stress tensor of the simulation cell. The strain is unitless and the stress is in units of MPa. Download strain-stress to your local computer, plot it, and you will see a point at which the stress-strain relation starts to deviate from linearity. Let's call it P1, which corresponds to

0.000401244438509872 -1172.52533757423

To visualize the dislocation core, download dump files to the same directory on your local computer. You do not need to download all of them at once, just selected ones, e.g., dump.0.load, dump.50.load, dump.100.load, ..., dump.500.load. Open any of them in OVITO by File --> Load File --> select the file --> Open. Then, Add modification --> Dislocation analysis (DXA), and change the "input crystal type" to "Body-centered cubic (BCC)". The blue and white atoms, respectively, are in BCC and disordered local structures. The green line is the dislocation line. White atoms exist in three locations: top layer, bottom layer, and center of the simulation cell. Those in the center are atoms in the dislocation core. Select one white atom using the crosshair button.

Say that you chose to open the file dump.0.load, now please click on dump.0.load [LAMMPS Dump] in the Data Source section on the right panel. Then in the External file section below, check the box before auto-generate below Search pattern:, you should see Found X matching files, where X is the number of dump files you just downloaded.

Next, go through all dump files frame by frame in OVITO and pay attention to between which two frames the dislocation core starts to move along the positive x direction. Why is this important? Because when the applied stress surpasses the Peierls stress, the dislocation line should move from one Peierls valley to another. Therefore, if the dislocation moves between one dump file and the next one, the Peierls stress is between the two stresses associated with these two dump files.

How do we determine whether the dislocation moves? Usually one of the two criteria is used: (i) does any white/blue atom become blue/white? (ii) does the green line move by a non-negligible distance? The keyword here is "non-negligible". Regardless of whether the Peierls stress has been reached, the green line may move a little bit between any two frames, especially when the dislocation is an edge dislocation. However, this may be because the entire simulation cell is sheared and so are all atoms within. If, from dump file A to dump file B, the dislocation moves a little bit, and from dump file B to dump file C, the dislocation moves by a longer distance, then likely the Peierls stress is reached somewhere between B and C.

Note: For a screw dislocation, it is important to check whether the dislocation moves within the xz plane. In many cases, the screw dislocation immediately crosses slip to a plane that is not parallel to xz. For more on this topic, read this paper. When this happens, the Peierls stress is not calculable. Write this down and move on to the next calculation. Sometimes the screw dislocation moves within the xz plane by a certain distance, and then crosses slip. In this case, the Peierls stress is considered calculable. Note that an edge dislocation does not cross slip, so its Peierls stress should always be calculable.

In the example provided in this GitHub repository, the dislocation moves between dump.350.load and dump.400.load. Then download dump.360.load, dump.370.load, dump.380.load, and dump.390.load from OSCER to the same directory (to which all previous dump files were downloaded) on your local computer. Open any dump file again in OVITO, by File --> Load File --> select the file --> Replace selected. Again, go through the newly downloaded dump files frame by frame and identify the two frames between which the dislocation core starts to move. The two frames are dump.390.load and dump.400.load.

Then download dump.391.load, dump.392.load, ..., dump.399.load to the same local directory. Open any dump file, go through these new dump files and identify the two frames between which the dislocation core starts to move. The two frames are dump.398.load and dump.399.load.

Then the Peierls stress is the stress of the simulation cell corresponding to dump.399.load.

In the file strain-stress, the first line corresponds to dump.0.load, and so line 400 corresponds to dump.399.load, i.e.,

0.000400240702603493 -1169.50115394567

As a result, the Peierls stress for the anti-twinning direction on the {112} plane is 1169.5 MPa, which is very close to the stress at P1 which was identified earlier.

In a general case, however,

  • The dislocation core may move along the negative x direction, depending on the Burgers vector of the dislocation and the shear direction.
  • Sometimes there is more than one point on the stress-strain curve at which the stress-strain relation starts to deviate from linearity. Let's say there are three such points and we call them P1, P2, and P3. Then the point at which the dislocation core starts to move may be one of them, or none of them. In other words, do not assume that any of these points is the critical point associated with the Peierls stress. If the point at which the dislocation core starts to move does not correspond to any turning point identified on the stress-strain curve, go with the former point instead of the latter.

References

If you use any files from this GitHub repository, please cite

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

2 stars

Watchers

0 watching

Forks

1 fork
Report repository

Releases

No releases published

Packages

No packages published

Languages

  • Shell 100.0%