Skip to content

5.2 Elastic tensor from finite strains

These are unexecuted teaching inputs and starting settings to test. Original figures are schematics, not computed results. Use licensed VASP and PAW data, replace every placeholder, record the executable version and validate convergence.

5.2.1 Model, units and provenance

Use eV for energy, Å for length and eV/Å for force; 1 kbar = 0.1 GPa. State normalization per atom, molecule, primitive cell or simulation cell. Record PAW identifiers, release, ZVAL, ENMAX and permitted hashes; never redistribute POTCAR. SCF convergence addresses the chosen electronic problem; convergence of the target property requires separate tests.

 Elastic tensor from finite strains — original conceptual schematic

Original schematic. Curves explain concepts; blank data areas await verified learner results. No calculation is claimed.

\[ \boldsymbol\eta=(\epsilon_{xx},\epsilon_{yy},\epsilon_{zz},2\epsilon_{yz},2\epsilon_{xz},2\epsilon_{xy})^{\mathsf T},\qquad C_{ij}=\frac{\partial\sigma_i}{\partial\eta_j} \]

5.2.2 Worked case: procedure, interpretation and checks

What is measured?

A bulk modulus describes uniform compression. An elastic tensor describes how stress responds to each independent small deformation, including stretching and shear. Treat the crystal as a directional spring network. “Stiff along x” does not imply “stiff in shear.” At small strain, σi = Σj Cij ηj. Adopt engineering Voigt strain η = (εxx, εyy, εzz, 2εyz, 2εxz, 2εxy), and stress ordering (σxx, σyy, σzz, σyz, σxz, σxy). These conventions must be stated because a missing factor of two in shear corrupts C44.

A clamped-ion tensor keeps internal fractional positions affine under strain. A relaxed-ion tensor lets atoms adjust inside each strained cell. Internal rearrangement can soften the response, so the two tensors answer different physical questions.

Hands-on sequence

Begin with a zero-pressure cell relaxed to substantially tighter tolerances than the strain-induced stress signal. Record its residual stress. For each of six engineering components j, create both positive and negative ηj, initially trying magnitudes 0.0025, 0.005, and 0.01. Use a deformation gradient F = I+ε with symmetric off-diagonal entries ηj/2 for engineering shear. This is a small-strain construction, not a finite-strain constitutive law.

For each strained cell: (1) run an affine static calculation for clamped-ion response; (2) if relaxed-ion response is desired, relax positions at exactly that cell using ISIF=2; (3) perform a final static stress calculation with the same electronic settings. Fit every stress component as a function of strain. Central differences remove the leading effect of residual stress:

Cij ≈ [σi(+δ ej)−σi(−δ ej)]/(2δ).

Build the full matrix, inspect Cij−Cji before imposing symmetry, and then compare independent entries allowed by the crystal point group. A large asymmetry is diagnostic information, not something to erase immediately by averaging.

VASP routes and sign

A separate built-in finite-difference option is:

This alternative is not the INCAR for every manually strained point. Verify supported settings for the installed version, read the separate elastic contributions in OUTCAR, and identify the relaxed-ion total rather than copying an arbitrary tensor block. VASP's DFPT displacement response does not provide the full strain elastic tensor simply by replacing 6 with 8.

VASP stress follows its own sign convention: positive diagonal values indicate compression. For conventional tension-positive stress-strain fitting, convert signs consistently and convert kbar to GPa. Also explicitly reorder VASP's printed tensor components to the chosen Voigt ordering. Validate with a small isotropic expansion: the physical restoring pressure should become tensile/negative hydrostatic pressure relative to equilibrium.

Read the tensor intelligently

At zero external stress, cubic mechanical stability requires C11−C12>0, C11+2C12>0, and C44>0. These are local elastic conditions, not proof of stability against every finite-wavevector phonon. Under finite pressure the appropriate stress-corrected stability criteria differ. For a general crystal, use the correctly constructed elastic energy quadratic form and relevant boundary conditions; blindly testing every Cij>0 is wrong because off-diagonal entries may be negative.

Converge strain magnitude and electronic accuracy together. Too small δ divides numerical stress noise by a tiny number; too large δ includes nonlinear elasticity. Compare slopes across the three amplitudes, increase the cutoff, and look for consistent symmetry-equivalent entries. Two-dimensional materials require an explicitly declared effective thickness or a 2D stiffness in N/m; vacuum-diluted GPa values are not intrinsic material properties.

Exercise: for a cubic crystal apply a single xx strain and identify which stress responses estimate C11 and C12. Apply a shear and verify the engineering-shear factor. Compare clamped-ion and relaxed-ion tensors without assuming they must differ by the same amount in every component.

5.2.3 Unexecuted inputs and analysis scaffolds

These are unexecuted teaching inputs and starting settings to test. Original figures are schematics, not computed results. Use licensed VASP and PAW data, replace every placeholder, record the executable version and validate convergence.

5.2.3.1 Input block 1

# External finite-strain series: internal coordinates relax, strain is fixed
IBRION = 2
NSW    = 150
ISIF   = 2
EDIFFG = -0.002
# Save final stress using an additional NSW=0 static run

5.2.3.2 Input block 2

# Alternative response calculation on the well-relaxed reference structure
IBRION = 6
ISIF   = 3
NFREE  = 2
POTIM  = 0.015
NSW    = 1

5.2.5 Technical sources