Calibration of Toughness and Strength Parameters

Learning objectives

By the end of this lesson, you should be able to:

  • Distinguish toughness calibration from strength calibration.
  • Explain why \(G_c\), tensile strength, and compressive strength are separate material inputs.
  • Understand why classical phase-field models can tie apparent strength to \(\ell\).
  • Describe the calibration philosophy behind generalized nucleation models.

1. Why calibration matters

A fracture model should use parameters that correspond to measurable material properties.

For classical brittle phase-field fracture, the common inputs are

\[ E,\quad \nu,\quad G_c,\quad \ell. \]

These are often enough for crack propagation from a pre-existing crack.

For fracture nucleation, we also need strength information, such as

\[ \sigma_{ts},\qquad \sigma_{cs}, \]

or more generally a strength surface

\[ F(\sigma)=0. \]

The key distinction is:

Toughness controls crack propagation.
Strength controls crack nucleation.

2. Toughness calibration

Toughness is represented by the critical energy release rate

\[ G_c. \]

It has units

\[ [G_c]=\text{J/m}^2=\text{N/m}. \]

For a large pre-existing crack, propagation occurs when

\[ G=G_c. \]

Therefore, \(G_c\) should be calibrated from fracture tests involving large cracks, not from smooth tensile tests.

3. Strength calibration

Strength controls when a new crack nucleates.

For uniaxial tension,

\[ \sigma=\sigma_{ts}. \]

For uniaxial compression,

\[ \sigma=-\sigma_{cs}. \]

For general stress states, nucleation is described by

\[ F(\sigma)=0. \]

Strength parameters should be calibrated from failure-stress experiments such as tension, compression, shear, biaxial, or triaxial tests.

4. Why classical phase-field calibration is limited

In classical phase-field fracture, the apparent nucleation stress often depends on

\[ E,\quad G_c,\quad \ell. \]

A typical scaling is

\[ \sigma_c^{\text{app}}\sim\sqrt{\frac{E G_c}{\ell}}. \]

This means that changing \(\ell\) can change the apparent strength.

That is problematic if \(\ell\) was intended only as a regularization length.

5. Calibrating \(\ell\) from tensile strength

Some classical formulations choose \(\ell\) or \(\varepsilon\) so that a one-dimensional tensile test fails at the measured tensile strength.

One relation discussed by Kumar, Bourdin, Francfort, and Lopez-Pamies is

\[ \varepsilon=\frac{3G_cE}{8\sigma_{ts}^2}. \]

This uses measurable quantities \(E\), \(G_c\), and \(\sigma_{ts}\).

However, it only matches one point on the full strength surface: uniaxial tension.

6. Why tensile calibration alone is insufficient

A material may fail differently under:

  • uniaxial tension,
  • compression,
  • shear,
  • biaxial loading,
  • pressure-dependent stress states.

Matching only \(\sigma_{ts}\) does not ensure that the model matches \(\sigma_{cs}\), shear strength, or multiaxial failure data.

For example, a material may be much stronger in compression than in tension:

\[ \sigma_{cs}\gg\sigma_{ts}. \]

A one-point tensile calibration cannot automatically capture this asymmetry.

7. Strength surface calibration

A better calibration strategy prescribes a strength surface directly:

\[ F(\sigma)=0. \]

For isotropic materials, this can be written using stress invariants:

\[ I_1=\operatorname{tr}(\sigma), \qquad J_2=\frac12\sigma_D:\sigma_D, \qquad J_3=\frac13\operatorname{tr}(\sigma_D^3). \]

A Drucker-Prager-type surface can be calibrated using tensile and compressive strengths:

\[ F(\sigma) = \sqrt{J_2} + \frac{\sigma_{cs}-\sigma_{ts}} {\sqrt{3}(\sigma_{cs}+\sigma_{ts})}I_1 - \frac{2\sigma_{cs}\sigma_{ts}} {\sqrt{3}(\sigma_{cs}+\sigma_{ts})} =0. \]

This surface distinguishes tension from compression and gives a simple pressure-sensitive failure criterion.

8. Correct calibration philosophy

The clean calibration logic is:

Elastic constants E, ν → elastic tests

Toughness Gc → large-crack fracture tests

Strength surface F(σ)=0 → failure-stress tests

Length scale ℓ → numerical/localization parameter or calibrated material length

Do not force one parameter to do the job of another.

A complete brittle fracture model should separately represent:

\[ W(\varepsilon),\qquad G_c, \qquad F(\sigma)=0. \]

9. Three calibration regimes

A good model should be checked in three regimes.

Large cracks

Propagation should satisfy

\[ G=G_c. \]

Smooth bulk nucleation

Failure should occur when

\[ F(\sigma)=0. \]

Small cracks and notches

The model should transition between strength-controlled nucleation and toughness-controlled propagation.

Small flaws are often strength-dominated; large cracks are Griffith-dominated; intermediate flaws test whether the model handles both correctly.

10. Main takeaway

Classical phase-field fracture is strong for propagation, but its nucleation strength may be controlled indirectly by \(G_c\) and \(\ell\).

Generalized nucleation models aim to keep:

\[ G_c\quad\text{for propagation} \]

and

\[ F(\sigma)=0\quad\text{for nucleation} \]

as independent material inputs.