Gaussian 16 Linux [NEW]

| Problem | Solution | |---------|----------| | g16: command not found | Check PATH and g16root | | Cannot open license file | Verify .secret file in bsd/ | | Out of memory | Reduce %mem, increase system swap | | Disk full in scratch | Clean /scratch; use larger partition | | linda error: cannot connect | Set GAUSS_LFLAGS='-vv -n2' for MPI |

For large calculations (e.g., CCSD(T)), use:

%nosave

to avoid writing large checkpoint files.

Slurm script (run_gaussian.slurm):

#!/bin/bash
#SBATCH --job-name=G16
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=8
#SBATCH --mem=16G
#SBATCH --time=12:00:00
export g16root=/opt/gaussian16
export GAUSS_SCRDIR=$SLURM_TMPDIR/gaussian
source $g16root/g16/bsd/g16.profile
g16 < input.com > output.log

Submit with:

sbatch run_gaussian.slurm

---

Gaussian 16 uses a license file named g16/bsd/cred. Copy your license credential file (provided by Gaussian, Inc.) into:

/opt/g16/bsd/cred

Set the file permissions:

chmod 600 /opt/g16/bsd/cred

For reproducible research, you can containerize Gaussian 16 using Docker or Singularity. Example Dockerfile (simplified):

FROM ubuntu:20.04
RUN apt update && apt install -y csh libgfortran4
COPY g16 /opt/g16
ENV g16root=/opt
ENV GAUSS_EXEDIR=/opt/g16
ENV PATH=/opt/g16:$PATH

Build and run:

docker build -t g16 .
docker run --rm -v $(pwd):/data -w /data g16 g16 input.com output.log

| Error | Likely Cause | Solution | |-------|--------------|----------| | g16: command not found | Environment not sourced | Run source ~/.bashrc | | Error: Cannot open scratch file | No write permission to GAUSS_SCRDIR | chmod 1777 /scratch or use local path | | Illegal instruction | CPU too old (missing AVX) | Request newer hardware or use %cpu=-AVX | | Segmentation fault | Insufficient memory | Increase %mem or reduce job size | gaussian 16 linux

---