05 / REFERENCE

Parameter Reference & Calculation Review

Find out the physical effects, units and prerequisites of parameters, and establish calculation records that can be restarted, reproducible and explainable.

7 chapters It is recommended to read and practice in batches Updated on 2026.09.19

The input examples in this manual have been checked against the data and have not yet been verified by VASP actual calculations case by case. When using, it must be combined with software version, pseudopotential and specific system verification.

Understand the Parameters Before Using a Template

READ A TAG

INCAR specifies the calculation method and control items, POSCAR describes the structure, POTCAR provides PAW data corresponding to the order of elements, and KPOINTS or KSPACING determines sampling. Just because a label is written legally does not mean that the files are consistent with each other. When reading a parameter, confirm in order: what physical or numerical step it controls, the type and unit of the value, when it takes effect, which inputs it is related to, and whether your program version supports it.

The default value means "what the program will do if you don't write this", and does not mean it is the best for your material. Some default values also change with other tags, such as whether hybrid functional is turned on and whether wavefunction is read successfully, which will affect subsequent behavior. A number in the textbook is also just a starting point with assumptions; the calculation report should record the actual settings used and the convergence evidence, rather than just "using the default precision".

SYSTEM = Describe your model
ENCUT = 520       # eV; replace after convergence tests
EDIFF = 1E-6      ! electronic stopping threshold, eV
ISMEAR = 0; SIGMA = 0.05
LWAVE = .TRUE.

Here we only demonstrate ordinary assignments, comments and semicolons, which are not enough to independently define a reliable task. Logical quantities are usually written as .TRUE. or .FALSE.; arrays have both atomic items and species items, and they cannot be mixed. For example, MAGMOM is usually expanded in atomic sequence, while LDAUL, LDAUU, and LDAUJ correspond in POTCAR species order. 4*0.0 It means four identical array values, but it does not mean that only one number will be obtained after arithmetic multiplication.

Two "species numbers". The same chemical element may also be split into multiple species blocks in the input to set up different kinetics or other processing. The species-by-species array should correspond to the actual POSCAR/POTCAR block and cannot be filled in only by the number of different elements in the chemical formula.

The following cheat sheet organizes commonly used items and does not claim to contain all VASP parameters. Each item links to Official INCAR documentation Or the corresponding tab page; when special algorithms are involved, please continue to read their applicability and version notes.

I have read and understood this section Back to top ↑

Search Parameters by Purpose

PARAMETER INDEX

Search by tag, keyword or category. Look at the "function" and "premise" first, and then decide on the numerical value; the "accuracy" category deals with numerical representation, and the "electronic/magnetic" category may also change the desired physical state. They cannot be uniformly understood as bigger is better.

Parameters/Official description	Category/Unit	Function and use	Premise and common misuse
ENCUT ↗	Accuracy eV; positive real number	Plane-wave kinetic-energy cutoff Test the target energy difference, forces, or stress	Not an upper bound on electron energies; use one validated cutoff consistently across comparative calculations
PREC ↗	Accuracy String, such as Accurate	Numerical precision and FFT-grid strategy Check together with ENCUT, the mesh, and the task	Accurate does not mean every source of error is converged
EDIFF ↗	Accuracy eV; real number	Electronic SCF stopping threshold Check energy changes between electronic steps for the whole calculation cell	Not an error per atom; setting 0 requests a fixed NELM iteration count
EDIFFG ↗	Accuracy Positive values eV; negative values correspond to eV/Å	Ionic-relaxation stopping criterion Negative: force norms; positive: energy change between ionic steps	Does not establish adequate sampling in molecular dynamics
LREAL ↗	Accuracy Auto or logical option	Real- or reciprocal-space treatment of nonlocal projectors For larger systems, test the cost and error of Auto	Compare Auto and .FALSE. for the property of interest
ADDGRID ↗	Accuracy Boolean	Additional support grid for augmentation charges Test whether it reduces the relevant force noise	Interpolation issues can occur; do not enable universally without testing
LASPH ↗	Accuracy Boolean	Nonspherical gradient contributions within PAW spheres Check especially for d/f elements, +U, hybrids, and meta-GGA	Keep the setting consistent across calculations used for energy comparisons
KSPACING ↗	Accuracy Å⁻¹; positive real number	Spacing for an automatic mesh when KPOINTS is absent A smaller value generally produces more k points	Ignored when KPOINTS is present; inspect the mesh actually generated
KGAMMA ↗	Accuracy Boolean	Center the automatic mesh on Γ Use with KSPACING and the crystal symmetry	Does not override the mesh mode in an explicit KPOINTS file
ALGO ↗	Electronic String	Electronic minimization algorithm Start with a robust algorithm compatible with the method, then benchmark	Fast is not properly supported for hybrids, and VeryFast is unsupported; check method-specific guidance
NELM ↗	Electronic Positive integer; iterations	Maximum electronic iteration count Check whether the run stopped at this limit	A larger limit does not fix a faulty structure or persistent oscillations
ISMEAR ↗	Electronic Integer option	Occupation, integration, and smearing method Choose for the material and for energies or forces	Positive values are unsuitable for straightforward use in gapped systems; tetrahedra require an appropriate mesh
SIGMA ↗	Electronic eV; real number	Occupation-smearing width Test energies and forces together with the k-point mesh	Only Fermi–Dirac smearing has a direct electronic-temperature interpretation
NBANDS ↗	Electronic positive integer	Number of bands to calculate Metals, hybrid functionals, and optics may need more unoccupied bands	A default or ground-state-sufficient count does not establish convergence of excited-state properties
NELECT ↗	Electronic electron number; real number	Total electron count Define the charge state relative to the total PAW valence-electron count	Charged periodic systems require attention to the compensating background, finite size, and reference energies
ICHARG ↗	Electronic Integer option	Initialization or fixing of the charge density 11 is commonly used for compatible CHGCAR-based fixed-density steps	A fixed density is not an SCF solution; check orbital-dependent methods separately
ISTART ↗	Electronic Integer option	Wavefunction initialization and restart mode Choose based on the available compatible WAVECAR	Check the output to confirm a successful read; a filename is insufficient
GGA ↗	Electronic String, such as PE	Select an LDA/GGA form Match the intended method and PAW data; PE specifies PBE	One setting does not turn arbitrary inputs into a valid PBE workflow
METAGGA ↗	Electronic String	Select a meta-GGA functional Check version, PAW, and kinetic-energy-density requirements	Mesh requirements and fixed-density band procedures may differ from GGA
IBRION ↗	Ionic Integer option	Ionic update, MD, or response task 2 is commonly used for conjugate-gradient relaxation; 0 selects MD	Not every positive value means geometry optimization; consult the task-specific documentation
NSW ↗	Ionic Integer; ionic or MD steps	Maximum ionic or MD step count Define the task length together with IBRION	Exhausting NSW does not prove relaxation convergence; MD step count does not establish equilibration
ISIF ↗	Ionic Integer 0–8	Stress calculation and allowed cell/atomic degrees of freedom Distinguish atomic relaxation, volume changes, and shape changes	Do not freely relax a cell containing vacuum; 8 requires VASP 6.4.1+
POTIM ↗	Ionic fs for MD; Å for finite displacements; otherwise algorithm-dependent	Time step or ionic displacement/step scale IBRION=0 requires an explicitly chosen suitable time step	Meaning and units depend on IBRION
PSTRESS ↗	Ionic kbar	External pressure or related pressure term State the stress sign convention and target boundary conditions	1 kbar = 0.1 GPa; do not tune it to fit lattice constants without a physical basis
ISYM ↗	Ionic Integer option	Use of symmetry Consider the structure, initial moments, and distortions being studied	ISYM=0 is not identical to −1; disabling symmetry is not a universal fix
SYMPREC ↗	Ionic Real number; symmetry tolerance	Tolerance for symmetry recognition Check coordinate rounding and genuine small distortions	A loose tolerance can mistake a real distortion for higher symmetry
ISPIN ↗	Magnetism 1 or 2	Non-spin-polarized or collinear spin-polarized treatment When magnetism is possible, test several initial MAGMOM patterns	Use LNONCOLLINEAR for noncollinear calculations and check version-specific combination rules
MAGMOM ↗	Magnetism μB; atom-by-atom array	Initial magnetic moments and magnetic symmetry One component per atom for collinear spin; three for noncollinear spin	Does not fix the final moments; removing it at a restart may change symmetry
LNONCOLLINEAR ↗	Magnetism Boolean	Allow a noncollinear spin density Use vasp_ncl with the correct three-component input	ISPIN=2 is not required to activate it; check version compatibility
LSORBIT ↗	Magnetism Boolean	Enable spin–orbit coupling Use vasp_ncl; check magnetization directions and symmetry	Small magnetic-anisotropy energies require strict numerical convergence
SAXIS ↗	Magnetism Three-component direction; dimensionless	Spin quantization axis and spin-coordinate basis Interpret MAGMOM and output components in the appropriate frame	Does not apply an external magnetic field; changing the axis changes how vectors are interpreted
LDAU ↗	Magnetism Boolean	Enable a local interaction correction Specify a consistent set of LDAUTYPE, LDAUL, U, and J settings	Define the corrected orbitals and justify the interaction parameters
LDAUTYPE ↗	Magnetism Integer option	Select a +U or related linear-response form 2 selects the Dudarev form; define the effective U−J	U and J have different meanings in different forms; literature values are not interchangeable
LDAUL ↗	Magnetism Species-wise integer array	Angular-momentum channel to correct 2 for d, 3 for f, and −1 for no correction on that species	Follow POTCAR species order; this is not a per-atom list
LDAUU ↗	Magnetism eV; species-by-species array	U or related interaction parameter in the chosen form Document its source, calculation, or fitting procedure	Consider whether comparing absolute total energies at different U values is meaningful
LDAUJ ↗	Magnetism eV; species-by-species array	J or related interaction parameter in the chosen form In the Dudarev form, U and J determine U−J	Its meaning differs for LDAUTYPE=3; consult that form's documentation
LMAXMIX ↗	Magnetism nonnegative integer	Maximum one-center angular momentum in mixing and charge files Fixed-density +U commonly needs 4 for d and 6 for f throughout both stages	Raising it later cannot recover information omitted from an earlier file
LWAVE ↗	Output Boolean	Write wavefunction-related files Save them when needed for restarts or post-processing	If writing is disabled, an old WAVECAR does not belong to the current run automatically
LCHARG ↗	Output Boolean	Write charge-density-related files Plan ahead for fixed-density bands or charge analysis	An old CHGCAR may be incompatible with the current structure
LORBIT ↗	Output Integer option	Local orbital projections and related outputs 11 is commonly used for angular-momentum-resolved projections	Projections are method-dependent, not unique atomic charges or oxidation states
NEDOS ↗	Output positive integer	Number of energy-grid points for the DOS Set the intended grid with EMIN and EMAX	Increasing it cannot replace more k points or unoccupied bands
EMIN ↗	Output eV; real number	Lower edge of the DOS and related energy windows Use a consistent energy reference	Not automatically a lower bound measured relative to the Fermi energy
EMAX ↗	Output eV; real number	Upper edge of the DOS and related energy windows Use with EMIN, NEDOS, and the actual calculated band range	A window beyond the computed unoccupied bands does not create the missing states
LVHAR ↗	Output Boolean	Write the ionic-plus-Hartree local potential State which potential is used, for example in a work-function calculation	Check the LOCPOT reference, vacuum plateau, and dipole correction
LAECHG ↗	Output Boolean	Reconstruct and write densities with all-electron features Choose the reference density required by the chosen Bader workflow	Files represent different components, such as core, atomic-superposition, or converged valence density; do not confuse them
LHFCALC ↗	Advanced Boolean	Enable HF-type or hybrid exchange Use a compatible functional, AEXX, screening, and solver	A single switch does not fully define HSE06
AEXX ↗	Advanced dimensionless real numbers	Fraction of HF-type exact exchange Record it with the chosen hybrid-functional form	Changing the fraction changes the method; it is not an ordinary convergence setting
HFSCREEN ↗	Advanced Å⁻¹; real number	Range-separation parameter Standard HSE06 uses 0.2 with the other required definitions	Arbitrary combinations of GGA/meta-GGA and screening are not supported
IVDW ↗	Advanced Integer option	Pairwise or many-body dispersion correction Record the correction and its pairing with the functional	11 and 12 use different damping; avoid double-counting dispersion
LDIPOL ↗	Advanced Boolean	Include dipole corrections in the potential and forces Use with IDIPOL, DIPOL, and sufficient vacuum	Not a universal correction for errors in arbitrary charged periodic systems
IDIPOL ↗	Advanced Integer direction option	Direction for dipole and related electrostatic corrections Match the surface normal and cell conventions	A mismatch in direction or dimensionality changes the physical interpretation
DIPOL ↗	Advanced three-component fractional coordinates	Reference center for the dipole moment For surfaces or molecules, often start near an appropriate center of mass	Fractional coordinates, not Å; check the correction region and vacuum placement
IMAGES ↗	Advanced nonnegative integer	Number of calculation images or replicas In NEB, specify intermediate images with separate endpoint directories	Also used by other multi-replica methods; check the task and directory conventions
SPRING ↗	Advanced NEB spring option, eV/Å²	Spring and NEB treatment between path images A negative value selects a nudged elastic band in native VASP	A larger value does not guarantee a better path; keep optimizer settings specific to the implementation
MDALGO ↗	Advanced Integer option	Thermostat or lattice-dynamics algorithm Specify with IBRION=0, ISIF, and the thermostat settings	A temperature setting alone does not establish the ensemble or equilibration
TEBEG ↗	Advanced K; real number	Initial or target temperature setting Interpret with the velocity initialization, thermostat, and TEEND	Different from electronic SIGMA smearing; instantaneous temperature fluctuates
LANGEVIN_GAMMA ↗	Advanced ps⁻¹; species-by-species array	Langevin friction coefficients for atomic degrees of freedom For MDALGO=3, follow the actual species blocks	Units are ps⁻¹, not fs⁻¹; NpT also needs lattice friction and mass settings
NCORE ↗	Parallel positive integer	MPI ranks cooperating on each band For pure MPI, benchmark choices for the basis, hardware, and problem	Do not also set NPAR; GPU and relevant OpenMP paths reset it to 1
NPAR ↗	Parallel positive integer	Number of band-parallel groups Understand its relationship to NCORE when reading older inputs	Takes precedence over NCORE if both are set; modern tuning generally favors using NCORE
KPAR ↗	Parallel positive integer	Number of k-point parallel groups Must divide the available MPI ranks; consider the irreducible-k-point workload	Few k points limit the benefit of larger values; specialized algorithms may give it another meaning

Strict energy comparison requires consistent functional, PAW data, truncation strategies, and energy definitions. Changing U, exchanging the mixing ratio or the number of electrons often means changing the research model. The lower the total energy, the better the method cannot be claimed. All key comparisons fall back Error and convergence of target properties .

I have read and understood this section Back to top ↑

Find the Prerequisites for Your Task

WORKFLOW MATRIX

Dependencies for common workflows are listed below. All rows still require compatible structures, methods, PAW data, and samples; "extra files" in the table does not mean that basic input can be omitted. Prioritize establishing independent directories for different tasks, leaving a clear source of the previous stage.

Task	What you should have before entering	Additional documentation or processing	Check after completion
First fixed structure SCF	Rational structure, electron number, magnetic state and sampling	You can start from the atomic density, it is not necessary to have CHGCAR.	Whether the electron stopping condition is reached; whether the magnetic state and occupation are reasonable
Geometry optimization	Tested energy and force accuracy	IBRION, NSW, ISIF and constraints must match each other	Residual forces, stresses, stopping causes and local stagnation point properties
final static energy	Optimized CONTCAR converted to POSCAR after inspection	Fixed structure; converge to finer sampling according to target quantity	The comparison object method is consistent, electronic convergence, and no midway structure is taken by mistake.
LDA/GGA fixed density band structure	Convergent charge of regular grid SCF	Compatible with CHGCAR, ICHARG=11, and the path corresponding to the unit cell	Energy zeros are from reliable SCF; paths are not used for total energy integration
DFT+U band structure	SCF with consistent U, J, magnetic states and sufficient LMAXMIX	Use compatible CHGCAR before and after; d/f channels check 4/6 respectively	Whether recalculating the original regular grid with fixed density can reproduce the original results?
Hybrid/meta-GGA band structure	Orbital/density states available for corresponding methods	Use the correct path based on algorithm and version; you cannot always just copy CHGCAR	unoccupied bands, exchange sampling, algorithm and version conditions
finite displacement phonon	Fully optimized structure with sufficiently low force noise	Displacement supercell with consistent force calculation settings	Displacement amplitude, supercell, acoustic mode and imaginary frequency mode shape
NEB	Respectively optimized initial and final states, identical atomic mapping and unit cell	Intermediate image directory; explicit implementation of native VASP or VTST	Whether the path passes through atoms, the projection force of each image, and the path resolution
AIMD	Stable structure, appropriate initial velocity/temperature, clear ensemble	Time step, thermostat/pressure bath and restart state compatible	Energy drift, equilibrium section, sampling length and statistical error

Version example: The meta-GGA fixed density band structure process that relies on kinetic energy density is different before and after VASP 6.6.0; the new version needs to retain the corresponding kinetic energy density information according to official instructions. Don't think of the old PBE tutorial's directory copying method as a universal operation for all functional operations. See details Official band structure steps and this lesson electronic structure tasks .

I have read and understood this section Back to top ↑

Local INCAR Input Review

LOCAL INPUT REVIEW

Paste the INCAR text below and the tool will recognize the normal TAG = VALUE, # with ! Comments, as well as semicolon separation on the same line, indicate duplicate tags and some common combination problems. The text is processed within the browser on this page; the tool does not submit cluster tasks or run VASP.

Paste INCAR for review (only demo snippet below)

All analysis is completed in the local browser and no text is uploaded. It can only prompt the listed syntax and parameter combination problems; no prompt does not mean that the parameters are correct. VASP will not be run, and your compute directory will not be read.

"No prompt" does not mean that the settings are correct. There is no reading of POSCAR, POTCAR, KPOINTS, CHGCAR, or output files, and no knowledge of species ordering, magnetic ordering, program construction, and target accuracy. It does not verify convergence, does not predict structural rationality, and does not fully implement VASP's line continuation, array, block, and all syntax. The actual input taken still needs to be checked in OUTCAR.

For example, when you see "Check LMAXMIX", you should go back to the d/f orbital under study, the previous step of charge saving settings and the next step of fixed density process, rather than mechanically copying a number. When you see "NEB implementation needs to be confirmed", you should check whether the corresponding patches and optimizers are installed. You cannot write VTST parameters directly to unmodified native programs.

Exercise: The reviewer identified 20 parameters, no strong warnings, what else to check?

Check the elements, atomic numbers and coordinate order between the four basic input types; identify the sources of all restart files; confirm the compilation function and version; check the actual read parameters and stop reasons after running; and finally test the convergence of the target quantity to truncation, sampling and model size. None of this can be accomplished by looking at the INCAR text alone.

I have read and understood this section Back to top ↑

Parallelization: MPI, Threads, and GPUs

PARALLEL EXECUTION

Increasing the number of cores does not always shorten the time. Electronic states must be orthogonalized, FFT requires communication, and file reading and writing also have costs. MPI rank is a process, OpenMP thread is a parallel unit within a process, and GPU is another computing device; the number of tasks in the submission script cannot be copied indiscriminately into NCORE.

For common CPU MPI tasks without additional images or special algorithm groupings, KPAR divides the processes into k-point groups, NCORE specifies how many processes collaborate in a band structure, and NPAR is the number of corresponding band structure groupings. Available relationships at this time NPAR = Total MPI ranks / (KPAR × NCORE) Helps check divisibility relationships; when IMAGES or other algorithmic groupings exist, the corresponding levels must be deducted first. This is a division of labor relationship, and it is not recommended to write all three parameters at the same time.

NCORE and NPAR should not be set at the same time. If they coexist, the official explanation is that NPAR takes priority due to compatibility with old input; modern CPU MPI tuning usually uses NCORE first. Short tasks should be used to compare total time consumption, time consumption per electron step and peak memory under the same representative structure and the same physical parameters, and then select the appropriate combination. You cannot have a test change NBANDS, convergence steps, or hardware occupancy and attribute the entire difference to the parallelism parameters.

GPU and OpenMP paths are different. The current official description is that when the GPU path and the corresponding OpenMP thread parallelism are enabled, NCORE will be reset to 1; do not move the pure MPI NCORE tuning table directly. The GPU build is configured according to one MPI rank for each available GPU device, combined with CPU threads; some multi-chip devices have independently visible sub-devices, which should be understood based on the number of devices actually exposed by the computing center.

GPU builds Also not all algorithms are executed on the GPU. After verifying the specific VASP version, GPU implementation, compiled libraries, and feature support, use the output to confirm actual device allocation. KPAR can reduce some communication, but too many groupings may cause uneven load or increase memory pressure; Γ-only tasks do not have many k-points to allocate. For the latest relevant conditions, see NCORE, KPAR, GPU build and run instructions .

I have read and understood this section Back to top ↑

Restarts, Outputs, and Reproducible Records

RESTART & RECORDS

The restart file is a save of the calculation status of a certain step, and is not an acceleration package suitable for any new task. WAVECAR saves wavefunction related information, CHGCAR saves charge and PAW single center information, and CONTCAR saves the structure after the task is completed. It needs to be checked whether they are complete, come from a converged state, and are compatible with new unit cells, sampling, electron numbers, methods, and program construction.

When starting a new round of optimization from CONTCAR, the structure, atomic order and constraints are first checked and then written as POSCAR for the new directory. For dynamics, velocities and other continuation states may also be important; simply keeping the coordinates and re-randomizing the velocities cannot be called a strictly continuous extension of the original trajectory. The last round of logs should be retained before copying files to prevent OUTCAR with the same name from overwriting the evidence that caused the problem.

Output	can help answer	It cannot be concluded based on it alone
OUTCAR	Actual setup, electron/ion steps, force, stress, and method information	The existence of the file does not mean that the task is completed; you must look at the end and the reason for stopping
OSZICAR / standard output	Quickly track energy and iterative trends	Not a complete parameter profile and does not replace detailed convergence checks
CONTCAR	Current or final structure and related continuation information	Does not automatically prove that the structure has been optimized to the force threshold
vasprun.xml / vaspout.h5	Structured data for controlled post-processing	Availability depends on version, build and write completeness
DOSCAR / PROCAR / EIGENVAL	Density of states, projection or band structure related data	Projected charge is not equal to unique atomic charge; line mode Fermi energy is unreliable

It is recommended that each calculation directory be accompanied by a report: structure sources and hashes, software and dataset identification, submission scripts, all key inputs, restart sources, stopping criteria, convergence curves, post-processing versions and result units. POTCAR is licensed and managed, and public reproduction records can write the data set name, generation information, and check values; restricted data files should not be freely distributed as teaching downloads.

For details on restart settings, see ISTART with ICHARG . Compatible target quantities should be obtained by re-running from scratch and reading the old state; if the difference is obvious, first investigate the local magnetic state, method changes, PAW single-center information and convergence, and then talk about "machine error".

I have read and understood this section Back to top ↑

A Results Report Others Can Check

REPORT YOUR RESULT

"The calculated adsorption energy −0.6 eV" lacks necessary context: relative to what reference state, which of several configurations, does it include zero point energy and entropy, and is the unit adsorbed once or per atom? Use the following template to put these questions before starting the calculation, instead of working backwards when writing the paper.

Research Question: Compare two clear candidate configurations A and B.
Target quantity: Write calculation definition, notation convention, and units.
Model: State the unit cell, composition, magnetic state, environment and structural origin.
Method: software version, PAW data, functional, key parameters and basis.
Numerical check: record the convergence test and target quantity changes, not just the final threshold.
Results: Write numerical values, error margins, alternative explanations, and validation ranges.
Reproducible material: where input, output, commit scripts, and post-processing methods are saved.

Write "prediction", "simulation" and "experimental verification" separately. For example, "In a given PBE+U model and supercell, the electron energy of A is lower than that of B" is a conditional calculation conclusion; it does not automatically prove that the actual sample only generates A under limited temperature and atmosphere. Documenting such boundaries does not weaken the results but rather makes the next step of verification clearer.

Exercise: The same input on two machines results in different final results. How to check first?

First confirm that the VASP and PAW data versions, compilation functions, effective NBANDS, actual read settings and final magnetic states are the same; compare whether they all reach sufficient stop thresholds. The order of parallel floating point operations can make a small difference, but it cannot explain changes much larger than the numerical precision scale. Focus the comparison on whether the target energy differences, forces, and properties fall within established error limits.

The full record template is available in the download area of this page. return to Course Home , choose the next exercise based on your own research question.

Extracting a Checkable Summary from OUTCAR

matching inspect_vasp_run.py Only files are read and the calculation directory is not modified. It extracts the last TOTEN, entropy-free energy and sigma→0 energy, the last set of atomic forces, as well as the program's stop message and file SHA-256. Different energy fields have different meanings and must be agreed upon before making a difference.

When using the entire package of examples, run the following command from the unzipped example root directory. The script is located tools subdirectory; if you download the script separately, change the script path in the command to the actual saving location. will /path/to/run/OUTCAR Replace with your own output file.

python tools/inspect_vasp_run.py /path/to/run/OUTCAR > summary.json

The script requires Python 3 and does not rely on additional packages. If the output has not yet been written, some fields will be null. The maximum force modulo all atoms read, including fixed atoms, so it cannot be used indiscriminately as a substitute for optimization criteria under actual constraints.

Abstracts are not automatically accepted.

The presence of an EDIFF stop message in the file does not prove that the last round or each round of electronic iterations has converged; the appearance of a timing footer does not mean that the structural optimization has reached the force threshold. The script always sets automatically_certified_converged to false, please combine OSZICAR, structural constraints, calculation goals and convergence experiments to judge.

I have read and understood this section Back to top ↑

Example Files for This Module

DOWNLOAD & CHECK

Read the file description and corresponding chapters first, and then complete the model, POTCAR and pre-calculation. Files with the suffix .template contain parameters to be changed and cannot be submitted as they are.

Calculation recurrence record template ↓ report/calculation-report.md

Blank research record, including model, method, convergence, operating environment and evidence boundaries; can be filled in according to specific tasks.

Primary Sources & Further Reading

READ THE PRIMARY SOURCES

The main text is an independently organized study instruction. For label meanings and version differences, refer to the official documentation corresponding to your program version.