Quantum ESPRESSO learning pathways
MODEL · CONVERGE · VALIDATE · INTERPRET
Start with traceable pseudopotentials and a defined periodic model. Converge the target quantity before interpreting relaxed structures, electronic states, perturbative response or trajectories; preserve data between program stages.
Inputs are unexecuted teaching examples. Record executable versions, data provenance and units; establish convergence and scientific validity for your own system. External tools are identified separately.
Choose a calculation question
1 Build the calculation correctly
Record pseudopotentials and converge numerical controls.
1.1 A reproducible installation, not just an executable
A successful launch proves that a binary runs, not that the environment is appropriate for your science.
Study the workflow →1.2 Choose and audit pseudopotentials
A pseudopotential is part of the Hamiltonian, not a cosmetic file attachment.
Study the workflow →1.3 Geometry, units and reciprocal space
A valid parser can accept the wrong crystal.
Study the workflow →1.4 Silicon SCF: follow the self-consistency loop
SCF solves a fixed-point problem: the density determines the effective potential, its orbitals produce a new density, and mixing stabilizes the update.
Study the workflow →1.5 Engine, workspace and input grammar
Choose traceable UPF data and identify every stage of the workflow.
Study the workflow →1.6 Converge the quantity you intend to publish
Test both cutoffs, k sampling and electronic tolerance separately.
Study the workflow →1.7 Shared inputs, conventions and evidence
Complete parent inputs, unit conventions and reproducibility evidence for this course.
Study the workflow →2 Atoms, molecules and metallic sampling
Distinguish vacuum-isolated systems and periodic crystal conventions.
2.1 Hydrogen atom: spin is part of the reference
An isolated H atom has one valence electron.
Study the workflow →2.2 H₂ binding curve: cancellation with consistent references
A molecular binding curve is an energy difference between well-defined states, not the raw total energy of the molecule.
Study the workflow →2.3 Aluminum metal: occupations and the Fermi surface
Metallic occupations change rapidly near the Fermi surface.
Study the workflow →2.4 Converge the observable, not a familiar cutoff
Electronic iteration, plane-wave basis, density grid and Brillouin-zone sampling are different error sources.
Study the workflow →2.5 Complete silicon SCF calculation
Build an explicit primitive cell and evaluate the electronic success evidence.
Study the workflow →3 Geometry and mechanics
Validate atomic relaxation, cell stress and mechanical response.
3.1 Relax a water molecule with trustworthy forces
Geometry optimization follows forces, so noisy forces can create a false minimum even when the total energy changes little.
Study the workflow →3.2 Variable-cell silicon: pressure and Pulay error
Atomic relaxation and cell relaxation solve different problems.
Study the workflow →3.3 Equation of state: fit a physically resolved minimum
An equation-of-state fit turns an energy–volume curve into equilibrium volume and curvature.
Study the workflow →3.4 Finite-strain elasticity: clamped ions versus relaxed ions
Elastic constants are derivatives of stress with respect to strain.
Study the workflow →3.5 Relax atoms and cells with distinct criteria
Verify the optimizer, final structure, forces and stress independently.
Study the workflow →4 Electronic interpretation
Interpret states, reciprocal sampling, occupations and spin models.
4.1 Silicon bands and DOS: two different sampling problems
A band path displays selected reciprocal-space directions; a DOS integrates the entire Brillouin zone.
Study the workflow →4.2 PDOS and charge density: projections are not unique atoms
PDOS answers how much a state resembles chosen local orbitals; it does not uniquely partition electrons into atoms.
Study the workflow →4.3 Iron magnetism: search states, do not trust one seed
Spin-polarized DFT may converge to different local electronic minima.
Study the workflow →4.4 Spin–orbit coupling: a different spinor problem
SOC couples spatial and spin degrees of freedom.
Study the workflow →4.5 DFT+U on NiO: define the subspace before choosing U
DFT+U changes the treatment of a specified localized subspace.
Study the workflow →4.6 Hybrid silicon: converge exchange sampling separately
Hybrid functionals add nonlocal exchange, which brings a second reciprocal-space convergence problem and substantial cost.
Study the workflow →4.7 Bands and DOS: different sampling tasks
Preserve parent data and reciprocal conventions in two electronic workflows.
Study the workflow →5 Surfaces and defects
Track reference energies, boundary conditions and finite-size effects.
5.1 Al(001) slab: separate thickness from vacuum
A slab has two independent finite-size errors: interaction between its two surfaces through the material, and interaction between repeated slabs through vacuum.
Study the workflow →5.2 Hydrogen adsorption: define stoichiometry and coverage
An adsorption energy depends on the chosen reservoir and surface coverage.
Study the workflow →5.3 Charge rearrangement and work function
Adsorption-induced density change and work-function change probe related but distinct phenomena.
Study the workflow →5.4 Silicon vacancy: finite-size and charge-state bookkeeping
A defect supercell is a periodic array of defects.
Study the workflow →6 Lattice response
Validate perturbations, force constants and reciprocal-space interpolation.
6.1 Γ-point phonons: curvature, not a decorative frequency list
A phonon calculation measures the second-order response around a stationary structure.
Study the workflow →6.2 From a q mesh to a phonon dispersion
DFPT evaluates dynamical matrices on a coarse reciprocal-space mesh.
Study the workflow →6.3 Phonon DOS and harmonic thermodynamics
Harmonic thermodynamics integrates all vibrational modes, so a path plot is insufficient.
Study the workflow →6.4 MgO dielectric response, Born charges and LO–TO splitting
In a polar insulator, a long-wavelength longitudinal vibration creates a macroscopic electric field.
Study the workflow →7 Spectroscopy and method boundaries
Separate response models and the executables or external tools they require.
7.1 Independent-particle optical spectra with epsilon.x
An optical spectrum needs transition matrix elements, not just a difference between occupied and empty energies.
Study the workflow →7.2 Molecular absorption with turboTDDFPT
Linear-response TDDFT describes the density's response to a weak time-dependent perturbation.
Study the workflow →7.3 GW and BSE: a verified handoff to an external code
GW quasiparticle corrections and electron–hole BSE spectra are different calculations from semilocal DFT, hybrid DFT and independent-particle optics.
Study the workflow →8 Dynamics, pathways and production
Preserve parent data, sampling evidence and reproducible execution.
8.1 Born–Oppenheimer MD: integrate accurate forces
In Born–Oppenheimer MD, electrons are converged for each ionic configuration before forces advance the nuclei.
Study the workflow →8.2 Car–Parrinello dynamics: keep fictitious electrons adiabatic
cp.
Study the workflow →8.3 NEB: find a path, then resolve its saddle
NEB optimizes a chain of images between known endpoints.
Study the workflow →8.4 Restarts and HPC: preserve correctness while scaling
Production work fails most expensively when a job appears restartable but its required files were never saved or were overwritten.
Study the workflow →8.5 Debugging, restarts and reproducibility
Separate new stages from interrupted-job continuation and archive evidence.
Study the workflow →Reference guides
- 1.5 Engine, workspace and input grammar — Choose traceable UPF data and identify every stage of the workflow.
- 2.5 Complete silicon SCF calculation — Build an explicit primitive cell and evaluate the electronic success evidence.
- 1.6 Converge the quantity you intend to publish — Test both cutoffs, k sampling and electronic tolerance separately.
- 3.5 Relax atoms and cells with distinct criteria — Verify the optimizer, final structure, forces and stress independently.
- 4.7 Bands and DOS: different sampling tasks — Preserve parent data and reciprocal conventions in two electronic workflows.
- 8.5 Debugging, restarts and reproducibility — Separate new stages from interrupted-job continuation and archive evidence.
What counts as a completed calculation?
Explain the model, input choices, units, parent/restart data, convergence of the target quantity and acceptance criteria. Preserve raw inputs and outputs with version and dataset provenance. A successful process exit, numerical convergence and physical interpretation require separate evidence.