7.1 Is this optimized molecule really a minimum?
These are unexecuted teaching inputs and starting models. Original diagrams are schematics, not calculated results. Validate version-specific syntax, licensed or authorized data, numerical convergence and the scientific model before using this workflow.
7.1.1 Model, units and provenance
Keep basis/potential files and executable versions traceable. Grid controls use Ry; common energy/force outputs use hartree and hartree/bohr. Read the unit in each output heading.
Shared inputs, conventions and evidence
7.1.2 Unexecuted inputs and explicit deltas
Use the accompanying instructions to identify the parent calculation and placement of every delta; a snippet is not automatically a standalone input. Preserve all blank-line and file-provenance requirements.
7.1.2.1 Input block 1
! Existing GLOBAL: replace RUN_TYPE
RUN_TYPE VIBRATIONAL_ANALYSIS
! Root-level addition
&VIBRATIONAL_ANALYSIS
DX [bohr] 0.01
INTENSITIES T
FULLY_PERIODIC F
THERMOCHEMISTRY F
&PRINT
&MOLDEN_VIB ON
&END MOLDEN_VIB
&END PRINT
&END VIBRATIONAL_ANALYSIS
! Inside FORCE_EVAL/DFT/PRINT
&MOMENTS ON
PERIODIC F
MAX_MOMENT 1
&END MOMENTS
7.1.3 Worked investigation
Question and intuition. A geometry optimizer can stop with small forces while an unstable direction remains. Harmonic analysis asks how the forces change when each nucleus moves slightly. Assemble the Cartesian Hessian H, mass-weight it as D = M^−1/2 H M^−1/2, and diagonalize D. A positive eigenvalue describes a restoring motion; a negative eigenvalue gives an imaginary frequency. This is a local statement about one stationary structure, not a guarantee that it is the global minimum. The pedagogical target is an isolated water molecule: three atoms provide three internal vibrations, while translation and rotation account for the other six Cartesian degrees of freedom. A linear molecule instead has 3N−5 internal vibrations.
Prerequisites and input contract. Begin with a tightly optimized, neutral, closed-shell water molecule, using a converged GPW/PBE calculation, consistent GTH pseudopotentials and basis sets, and isolated electrostatics. Both SUBSYS/CELL and DFT/POISSON must describe the intended nonperiodic system; adequate vacuum alone does not change a periodic Poisson problem into an isolated one. Reuse the final coordinates, not the initial XYZ file. Keep the same Hamiltonian and numerical settings during optimization and Hessian evaluation. Before applying the following delta, set DFT/SCF/EPS_SCF to 1.0E−8 or tighter and verify any OUTER_SCF tolerance as well. This follows the finite-difference accuracy warning in the 2025.2 vibrational-analysis reference.
Original input delta: valid section syntax. Replace RUN_TYPE inside the existing GLOBAL section; add VIBRATIONAL_ANALYSIS at the root. Insert MOMENTS inside the existing FORCE_EVAL/DFT/PRINT. Do not create duplicate nonrepeatable parent sections.
See input block 1 above.
DX is a displacement in bohr, not angstrom. Frequencies are normally reported as wavenumbers; read the actual output heading before exporting them. INTENSITIES requests intensity evaluation but does not itself request the dipoles. Here MOMENTS supplies them for molecular IR; no Raman polarizability calculation is configured. The MOMENTS reference distinguishes the molecular dipole operator from the periodic Berry-phase expression.
Numbered procedure. 1. Archive the final geometry, residual forces, basis and potential file versions, cell, SCF tolerance, and grid settings. Confirm that optimization ended by convergence, not its iteration limit. 2. Evaluate a single-point force at this geometry with the intended Hessian settings. A newly tightened grid may reveal that reoptimization is needed. 3. Build the finite-difference Hessian. Inspect every displaced calculation for SCF convergence; a completed parent job is insufficient evidence. 4. Repeat with DX = 0.005 and 0.02 bohr. Keep the central geometry fixed while assessing displacement sensitivity, then tighten numerical settings if necessary. 5. Animate all internal modes with a compatible Molden-file viewer. Identify bending and symmetric/asymmetric stretching by displacement patterns, not frequency order alone. 6. Tabulate mode character, wavenumber, IR intensity, and sensitivity to DX and grid. Include soft or imaginary modes rather than silently deleting them.
Interpretation and checks. Tiny residual external-mode frequencies can arise from numerical noise and imperfect stationarity. A substantial imaginary internal mode requires investigation: displace in both directions along its eigenvector, optimize both structures, and compare consistent energies. A transition state should have the expected unstable direction; simply counting one negative value is not a complete transition-state validation. Intensity measures the change of dipole along a normal coordinate, so a strongly moving mode need not be IR bright. A Hessian is harmonic and local: anharmonic shifts, finite-temperature broadening, and solvent rearrangements require other models. The ETH Zürich 2018 infrared exercise provides a useful complementary static-versus-dynamic teaching workflow.
Pitfalls and exercise. Too small DX amplifies force noise; too large DX samples anharmonic curvature. Do not treat gas-phase rigid-rotor/harmonic-oscillator thermochemistry as automatic solution or surface free energies. As an exercise, replace hydrogen masses by deuterium masses through an explicitly documented isotope setup, keeping the electronic potential surface fixed. Predict which modes shift most, then explain why not every frequency scales by exactly 1/√2. Deliver the convergence table and annotated mode images; leave numerical cells blank until actual calculations exist.
7.1.4 Related calculations
- 6.4 RDF, mean-square displacement and honest uncertainty
- 7.2 From a dipole trajectory to an IR spectrum