ADF

Introduction

ADF (The Amsterdam Density Functional package) is a software for first-principles electronic structure calculations. It is particularly popular in the research areas of homogeneous and heterogeneous catalysis, inorganic chemistry, heavy element chemistry, various types of spectroscopy, and biochemistry.

BAND, also called ADF-BAND, is the name of a software package closely related to ADF. BAND can be used to study periodic systems: solids, surfaces, slabs, polymers, and reactions on surfaces. BAND is a separate software product: you can use it without using ADF.

ADF on Graham

ADF 2017.207 and 2016.106 are installed on the new cluster Graham with modules. Free access for users on Graham.

Version selection

User can chose which version of ADF to use, 2016.106 or 2017.207

#for ADF 2017.207
module load adf/2017.207

#for ADF 2016.106
module load adf/2016.106

Job Submission

Graham uses your ComputeCanada(CC) user account, not your Sharcnet account anymore. Each research group has a default account with your supervisor's userid (PI'sid) named def-PI'sid (type 'groups' on Graham will show the groups you are in).

If this is your first time to use Graham, you can add the following line to your .bash_profile file:

export SBATCH_ACCOUNT=def-PI'sid

Otherwise, you will have to include #SBATCH_ACCOUNT=def-PI'sid in submit script for each job.

Graham uses Slurm, which is different from the sqsub command used on other Sharcnet clusters. A sbatch script is needed to setup the compute resources for the job. ADF is a MPI-based parallel software, user has the option to choice how many cpus and compute nodes to use for a job.

Here is an example for a full-node ADF job named adf_test.sh:

#!/bin/bash
#SBATCH --nodes=1 --ntasks-per-node=32  # 1 node with 32 cpus, you can modify it
#SBATCH --mem-per-cpu=2G                # memory, should be <4G for full node job
#SBATCH --time=00-03:00                 # time (DD-HH:MM)
#SBATCH --output=adf_test.log          # output .log file
module load adf/2017.207
ADF adf_test.inp                # ADF command

If the job needs more mem-per-cpu, then the job has to use multi-nodes, for example:

#SBATCH --nodes=4 --ntasks-per-node=8  # 4 nodes with 32 cpus, 8 cpus per node
#SBATCH --mem-per-cpu=6G              # 6G memory per cpu

adf_test.inp is an example input file for one-step ADF calculation

 Title WATER Geometry Optimization with Delocalized Coordinates

 Atoms
    O             0.000000     0.000000     0.000000
    H             0.000000    -0.689440    -0.578509
    H             0.000000     0.689440    -0.578509
 End

 Basis
 Type TZP
 Core Small
 End

 Geometry
  Optim Deloc
  Converge 0.0000001
 End

 End Input

Examples

Example for adf_test can be found on Graham under

/home/jemmyhu/tests/test_ADF/2017.207/test_adf/

Some user may do multi-step ADF runs within one job, see GO_H2O.run and GO_H2O.sh example under

/home/jemmyhu/tests/test_ADF/2017.207/test_adf/

The same procedure applies to BAND job, see band_test.inp and band_test.sh examples under

/home/jemmyhu/tests/test_ADF/2017.207/test_band

Message Passing Interface