AMS

Introduction

AMS (Amsterdam Modeling Suite), originally named ADF (Amsterdam Density Functional), is the SCM Software for Chemistry and Materials. AMS offers powerful computational chemistry tools for many research areas such as homogeneous and heterogeneous catalysis, inorganic chemistry, heavy element chemistry, various types of spectroscopy, and biochemistry.

The full SCM module products are available:

• ADF
• ADF-GUI
• BAND
• BAND-GUI
• DFTB
• ReaxFF
• COSMO-RS
• QE-GUI
• NBO6

Running AMS on Graham

The ams module is installed on Graham only due to license restrictions. The license is an Academic Computing Center license owned by SHARCNET. You may not use the Software for consulting services and for purposes that have a commercial nature. To check what versions are available, use the module spider command as follows:

[name@server $] module spider ams

For module commands, please see Using modules.

Job submission

Graham uses the Slurm scheduler; for details about submitting jobs, see Running jobs.

Because Graham's scratch file system is aged now, large AMS jobs may run slow on the cluster. For now, we recommend user to run AMS jobs using cpus within one node (one full or partial node) and use the local disk for SCM_TMPDIR. Each computer node has about 950GB local disk space. See example .sh file below

Example scripts for an adf job

This H2O_adf.sh script is for a whole-node job.

File : H2O_adf.sh

#!/bin/bash
#SBATCH --account=def-pi
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=32          # 1 node with all 32 cpus, MPI job
#SBATCH --mem=0                       # request all memory on node
#SBATCH --time=00-01:00               # time (DD-HH:MM)
#SBATCH --output=H2O_adf-%j.log       # output .log file

module unload openmpi
module load ams/2024.102
export SCM_TMPDIR=$SLURM_TMPDIR       # for cpus within 1 node and file size <900GB. comment out this line when you run large across multi-node job
bash H2O_adf.run                    # run the input script

This is the input file used in the script:

File : H2O_adf.run

#!/bin/sh
# This is a shell script for AMS2020
# You should use '$AMSBIN/ams' instead of '$ADFBIN/adf'

AMS_JOBNAME=H2O_adf $AMSBIN/ams <<eor
   # Input options for the AMS driver:
   System
      Atoms
         O             0.000000     0.000000     0.000000
         H             0.000000    -0.689440    -0.578509
         H             0.000000     0.689440    -0.578509
      End
   End
   Task GeometryOptimization
   GeometryOptimization
      Convergence gradients=1e-4
   End

   # The input options for ADF, which are described in this manual,
   # should be specified in the 'Engine ADF' block:

   Engine ADF
      Basis
         Type TZP
      End
      XC
         GGA PBE
      End
   EndEngine
eor

Example scripts for a band job

File : SnO_EFG_band.run

#!/bin/sh
# The calculation of the electric field gradient is invoked by the EFG key block
# Since Sn is quite an heavy atom we use the scalar relativistic option.

$AMSBIN/ams <<eor

Task SinglePoint
System
   FractionalCoords True

   Lattice
      3.8029  0.0  0.0
      0.0  3.8029  0.0
      0.0  0.0  4.8382
   End

   Atoms
      O   0.0  0.0  0.0
      O   0.5  0.5  0.0
      Sn  0.0  0.5  0.2369
      Sn  0.5  0.0 -0.2369
   End
End

Engine Band
   Title SnO EFG
   NumericalQuality Basic      ! Only for speed
   Tails bas=1e-8              ! Only for reproducibility with nr. of cores
   ! useful for Moessbauer spectroscopy: density and coulomb pot. at nuclei
   PropertiesAtNuclei
   End

   EFG
      Enabled True
   End

   Basis
      Type DZ
      Core none
   End
EndEngine
eor

The following slurm script is similar to the one used for a single adf run (H2O_adf.sh), except it's not a whole-node job.

File : SnO_EFG_band.sh

#!/bin/bash
#SBATCH --account=def-pi
#SBATCH --nodes=1 
#SBATCH --ntasks-per-node=16     # a 16 cpus MPI job
#SBATCH --mem-per-cpu=3G         # memory; 3G per cpu in this example
#SBATCH --time=00-10:00          # time (DD-HH:MM)
#SBATCH --output=SnO_EFG_band-%j.log

module unload openmpi
module load ams/2024.102
export SCM_TMPDIR=$SLURM_TMPDIR       # for cpus within 1 node and file size <900GB. comment out this line when you run large cross node jobs
bash SnO_EFG_band.run                 # run the input file

Notes

1. The input for AMS is different from ADF, the previous ADF input file will not run for the new AMS. Some examples can be found in /opt/software/ams/2020.102/examples/
2. Except the output .log file, other files are all saved in a subdirectory AMS_JOBNAME.results. If AMS_JOBNAME is not defined in the input .run file, the default name is ams.results
3. The restart file name is ams.rkf instead of the TAPE13 in previous ADF versions.
4. You can watch a recorded webinar/tutorial: An Update on ADF/AMS software on Graham

For more usage information, please check the manuals in SCM Support

Running AMS-GUI

Rendering over an SSH connection with X11 forwarding is very slow for GUI applications such as AMS-GUI. We recommend you use VNC to connect if you will be running AMS-GUI.

Graham

AMS can be run interactively in graphical mode on a Graham compute node (3hr time limit) over TigerVNC with these steps:

1. Install a TigerVNC client on your desktop
2. Connect to a compute node with vncviewer
3. module load ams
4. amsinput

Gra-vdi

AMS can be run interactively in graphical mode on gra-vdi (no connection time limit) over TigerVNC with these steps:

1. Install a TigerVNC client on your desktop
2. Connect to gra-vdi.computecanada.ca with vncviewer
3. module load SnEnv
4. module load clumod
5. module load ams
6. amsinput

Message Passing Interface