CP2K

Getting Started

  • Spack

Methods

  • Density Functional Theory
  • Post Hartree-Fock
  • Semi-Empiricals
  • Machine Learning
  • Embedding
  • QM/MM
  • Sampling
  • Optimization
  • Properties

References

  • Input Reference
    • ATOM
    • DEBUG
    • EXT_RESTART
    • FARMING
    • FORCE_EVAL
      • BSSE
      • DFT
        • ACTIVE_SPACE
        • ALMO_SCF
        • AUXILIARY_DENSITY_MATRIX_METHOD
        • DENSITY_FITTING
        • EFIELD
        • ENERGY_CORRECTION
        • EXCITED_STATES
        • EXTERNAL_DENSITY
        • EXTERNAL_POTENTIAL
        • EXTERNAL_VXC
        • KG_METHOD
        • KPOINTS
        • LOCALIZE
        • LOW_SPIN_ROKS
        • LS_SCF
        • MGRID
        • PERIODIC_EFIELD
        • POISSON
        • PRINT
        • QS
        • REAL_TIME_PROPAGATION
        • RELATIVISTIC
        • SCCS
        • SCF
        • SCRF
        • SIC
        • TDDFPT
        • TRANSPORT
        • XAS
        • XAS_TDP
        • XC
      • EIP
      • EMBED
      • EXTERNAL_POTENTIAL
      • MIXED
      • MM
      • NNP
      • PRINT
      • PROPERTIES
      • PW_DFT
      • QMMM
      • RESCALE_FORCES
      • SUBSYS
    • GLOBAL
    • MOTION
    • MULTIPLE_FORCE_EVALS
    • NEGF
    • OPTIMIZE_BASIS
    • OPTIMIZE_INPUT
    • SWARM
    • TEST
    • VIBRATIONAL_ANALYSIS
  • Bibliography
  • Units

Release Versions

  • 2023.2
  • 2023.1
  • All versions
CP2K
  • Input Reference
  • FORCE_EVAL
  • DFT
  • LS_SCF
  • PAO
  • LINE_SEARCH
  • Edit on GitHub

LINE_SEARCH

Detail settings or linesearch method. [Edit on GitHub]

Subsections

  • PRINT

Keywords

  • EPS_STEP_SIZE

  • INITIAL_STEP_SIZE

  • MAX_STEP_SIZE

  • METHOD

  • TINY_STEP_SIZE

Keyword descriptions

EPS_STEP_SIZE: real = 1.00000000E-001 

Convergence criterion of GOLD method. [Edit on GitHub]

INITIAL_STEP_SIZE: real = 1.00000000E-001 

Initial step length [Edit on GitHub]

MAX_STEP_SIZE: real = 3.00000000E+000 

Maximum step length [Edit on GitHub]

METHOD: enum = ADAPT 

Valid values:

  • ADAPT extrapolates usually based on 3 points, uses additional points on demand, very robust.

  • 3PNT extrapolate based on 3 points

  • 2PNT extrapolate based on 2 points and the slope (super fast, but might get stuck at saddle points)

  • GOLD perform 1D golden section search of the minimum (very expensive)

  • NONE always take steps of fixed INITIAL_STEP_SIZE

Mentions: ⭐PAO-ML

Linesearch method. [Edit on GitHub]

TINY_STEP_SIZE: real = 2.00000000E-003 

Step length taken if negative step is suggested. [Edit on GitHub]

Previous Next

© Copyright 2000-2024, CP2K Developers.

Built with Sphinx using a theme provided by Read the Docs.