Geometry Optimization in Orca 5.0: Complete Workflow Tutorial

Density Functional Theory (DFT) is a Nobel Prize-winning (1998) method for solving the Schrödinger equation in computational chemistry. Unlike wave function methods that calculate each electron's wave function, DFT derives electron distribution (electron density) directly. This approach includes many electron correlation effects at modest computational cost. The premise is that molecular energy can be determined from electron density rather than wave function. DFT can calculate chemical reactivity, structural properties, electronic properties, and thermodynamic properties. Basis sets are mathematical functions representing electronic wave functions, chosen based on atoms, electron count, and polarization requirements. Different theoretical methods correspond to different approximations of the Schrödinger equation: Hartree-Fock methods are least expensive but least accurate, semiempirical methods are low-cost qualitative methods useful for large molecules, and DFT provides a middle ground with better accuracy at reasonable computational cost.

In ab initio Hartree-Fock theory, molecular wave functions are constructed as linear combinations of atomic orbital basis functions (LCAO), where basis sets determine the mathematical flexibility for describing electron density; Gaussian-type orbitals (GTOs) are preferred over Slater-type orbitals because they enable efficient numerical integration of electron repulsion integrals, and basis set quality is systematically improved through contraction (combining multiple primitive Gaussians into contracted functions), split-valence expansions (providing multiple basis functions per atomic orbital), and polarization functions (adding higher angular momentum orbitals like d and f functions) to achieve balanced accuracy across different molecular properties.

DFT balances computational cost against accuracy across different functional types. Hartree-Fock is fast but lacks electron correlation. DFT provides good accuracy with reasonable cost but depends heavily on functional choice. Correlated wave function methods (coupled cluster) offer highest accuracy but scale exponentially with system size. Functional categories range from GGA (BP86) to hybrid (B3LYP) to double hybrids (B2PLYP). The Resolution of Identity approximation accelerates pure DFT calculations using auxiliary basis sets, enabling medium-sized molecules on standard hardware. Recent developments (RICOSX) extend this acceleration to hybrid functionals. Basis set notation follows systematic conventions: '6-31G' indicates core orbitals with fixed Gaussian contractions, valence region with contracted functions, polarization functions ('d', 'f') increase valence flexibility, and diffuse functions ('+') extend to outer regions.

In DFT calculations, selecting appropriate functionals and basis sets requires balancing accuracy and computational cost; functionals follow Jacob's Ladder (from Hartree-Fock through LDA, GGA, meta-GGA, hybrid, to double-hybrid), with dispersion corrections improving non-covalent calculations, while basis sets should generally use triple-zeta with polarization and diffuse functions for most calculations, and effective core potentials for heavy atoms to reduce computational expense.

Wave functions are expanded in basis sets (Gaussians, plane waves) that must be complete. Orthonormal basis functions allow efficient computation: integrating products picks out individual basis functions. Plane waves (e^(iG·r)) form complete bases for periodic systems but require cutoffs on G-vectors to limit computational cost. Higher G-values correspond to shorter wavelengths and higher kinetic energies, becoming unphysical beyond certain thresholds. The choice of basis set critically affects computational cost and accuracy, balancing completeness against feasibility.

The XYZ file format represents molecular structures by storing the number of atoms on the first line, a comment on the second line, and then atom records with atomic symbols followed by their x, y, and z coordinates (typically in angstroms) starting from the third line; this Cartesian coordinate system requires 3N total coordinates for N atoms, but only 3N-6 (for non-linear molecules) or 3N-5 (for linear molecules) are unique internal degrees of freedom since translations and rotations do not convey new structural information.

Two primary coordinate systems exist for specifying molecular geometries: Cartesian coordinates and Z-matrix (internal coordinates). Cartesian coordinates specify each atom's position in 3D space relative to an origin, which becomes computationally expensive for large molecules because moving one atom requires rearranging all others around it. Z-matrix uses chemically natural parameters: bond lengths (distance between two atoms), bond angles (angle between two bonds), and dihedral angles (twist between planes formed by four atoms). For n atoms, there are 3n-6 independent coordinates after accounting for three translations and three rotations.

Every atom in a molecule has Cartesian coordinates (x, y, z) that define its position in three-dimensional space. Computer-generated molecular structures are based on mathematical manipulation of these coordinate systems. Understanding Cartesian coordinates is fundamental to all molecular modeling software and calculations.

In molecular modeling, there are two coordinate systems: Cartesian coordinates use absolute x, y, z positions, while internal coordinates define molecular geometry through relative distances and angles between atoms. Internal coordinates are chemically more intuitive because molecular properties depend primarily on bond lengths rather than absolute orientation. However, Cartesian coordinates are required for molecular dynamics simulations since velocities must be defined along Cartesian axes.

Molecular modeling requires computer-readable representations. External coordinates use XYZ systems with 3n parameters for n atoms, sensitive to molecular position changes. Internal coordinates use bond lengths, angles, and torsional angles requiring only 3n-6 parameters, unaffected by absolute positioning. External coordinates are better for studying intermolecular interactions, while internal coordinates are preferable for studying conformational changes. Both representations are equivalent—external can be converted to internal using Euclidean distance and vector dot product formulas, and internal to external using algorithms like the fourth atom fixing method.

The command-line interface provides powerful tools for cluster management. 'squeue' displays all running and queued jobs with detailed information including job ID, user, partition, state (R=running, PD=pending), runtime, node count, and memory usage. 'sinfo' shows current node status (allocated, mixed, reserved, idle). 'scontrol' provides detailed job information including priority scores and failure reasons. Fair share scheduling reduces priority for heavy short-term users, ensuring equitable access. The SLURM wrapper suite generates equivalent original commands for transparency.

Essential Linux commands for HPC: pwd, ls, cd, cp -r, cat/more/less, man, module. SLURM combines resource management and scheduling with fair share (usage-based priority), backfill (fill gaps with small jobs), and expansion (quick priority for short jobs). Job submission workflow: salloc (interactive), sbatch (batch scripts), srun (parallel execution), squeue (monitoring), scancel (termination). C compilation demonstrates memory allocation, program counter loading, and CPU scheduling.

Linux provides a command-line interface for interacting with HPC clusters. Users type commands at the prompt, the system executes them, and displays output. The command prompt indicates the current working directory. Commands that run successfully produce output through standard output channels, while errors are directed to standard error channels. Anything following a pound sign (#) is treated as a comment and ignored by the system. Understanding this interaction model is fundamental to Linux-based HPC cluster usage and differs from graphical interfaces. The command prompt serves as the primary interface for system interaction, allowing users to navigate the file system, run programs, and manage resources through text-based commands.

SLURM is a job queue system that manages shared computing resources (such as GPUs and CPU cores) among multiple users by assigning jobs based on priorities and resource availability. Users can submit jobs using two main commands: 'srun' for interactive execution (which blocks the terminal until completion) and 'sbatch' for non-interactive execution (which redirects output to files). Key commands include 'sinfo' to view cluster status and available resources, 'squeue' to monitor job status and queue position, and 'scancel' to cancel jobs. Time limits can be set for each job to ensure fair resource allocation.

SLURM is a common batch scheduler used in HPC systems. Key commands: (1) sbatch submits a job script to the scheduler, (2) squeue (or sstat) checks job status and queue position, (3) scancel cancels a pending or running job, (4) srun runs single commands on compute nodes (often used within job scripts). These commands interface with the scheduler to manage job execution.