4. Damage initiation criteria for composites

Progressive damage frameworks require a precise answer to a deceptively simple question: at what stress state does a composite lamina begin to fail, and in which mode? A single scalar failure index is insufficient because carbon/epoxy plies exhibit fundamentally different strengths in fiber tension, fiber compression, matrix tension, matrix compression, and out-of-plane shear. The four criteria discussed here — Hashin, Puck, LaRC, and Christensen — each encode a different physical hypothesis about how those modes interact, and each produces a different set of mode-specific damage variables that drive stiffness degradation in the subsequent progressive-damage step.

Hashin: Mode-Separated Quadratic Stress Criteria

The Hashin criterion (1980) is the most widely implemented baseline in commercial FE codes. It separates failure into four modes and evaluates each with an independent quadratic expression. Using the standard convention where 1 is the fiber direction, 2 is the in-plane transverse direction, and 3 is the through-thickness direction, the fiber tension mode is:

F_ft = (σ11 / X_T)² + α · (τ12² + τ13²) / S12²  ≥ 1  → fiber tensile failure

where X_T is longitudinal tensile strength, S12 is in-plane shear strength, and α is a shear contribution factor typically set to 1.0 (some implementations use 0 for a purely fiber-dominated check). Matrix tension is evaluated as:

F_mt = (σ22 / Y_T)² + (τ12 / S12)² + (τ23 / S23)²  ≥ 1  → matrix tensile failure

Compression modes use the same quadratic structure but substitute X_C and Y_C, and the fiber compression expression typically omits the shear term because fiber kinking is governed by initial fiber misalignment rather than by shear interaction alone. The practical appeal of Hashin is its low calibration cost: only five strength values (X_T, X_C, Y_T, Y_C, S12) are required, all available from standard ASTM D3039/D6641 coupons. Its principal limitation is that it ignores the physical mechanism of fiber compression failure — it cannot predict fiber kinking, and it treats matrix compression as a simple strength reduction rather than a fracture-plane problem.

Puck: Physically Based Fracture-Plane Criteria

Puck's theory (1996) was developed from the observation that matrix failure in compression occurs on a inclined fracture plane, not on the ply plane. The criterion introduces the fracture plane angle θ_fp (measured from the 2–3 plane) as an explicit unknown. For matrix failure, Puck defines two modes:

• Mode A — transverse tension or moderate shear, fracture plane at θ_fp = 0° (the ply plane).
• Mode B — transverse compression with high shear, fracture plane at θ_fp ≈ 45°–55°.
• Mode C — pure transverse compression, fracture plane at θ_fp ≈ 45°.

The Puck matrix criterion takes the form of a quadratic interaction on the fracture plane:

(τ_nt / S_nt)² + (τ_n1 / S_n1)² + (σ_n / Y_T)²  ≥ 1   (tension, σ_n > 0)
(τ_nt / S_nt)² + (τ_n1 / S_n1)² + (σ_n / Y_C)²  ≥ 1   (compression, σ_n < 0)

where n is the fracture-plane normal, t is the in-plane transverse direction on that plane, and 1 is the fiber direction. The shear strengths S_nt and S_n1 are derived from S12 and S23 through a Mohr's-circle transformation that depends on θ_fp. In practice, the criterion is evaluated by sweeping θ_fp from −90° to +90° in 1° increments and taking the maximum failure index. This makes Puck more computationally expensive than Hashin but far more accurate for compression-dominated matrix failure, which is the critical mode in many structural applications.

LaRC: NASA Langley's Physically Based Criteria

The LaRC series (LaRC03, LaRC04) was developed at NASA Langley to address specific deficiencies in Hashin and Puck, particularly for matrix compression and fiber kinking. LaRC03 introduces a modified matrix compression criterion that accounts for the effect of through-thickness stress σ33 on the fracture plane, and a fiber kinking criterion based on the initial fiber misalignment angle φ0:

F_fiber_kinking = (τ_m / S_L)² + (σ_m / Y_C)²  ≥ 1

where τ_m and σ_m are the shear and normal stresses on the kink plane, computed from the applied stress state rotated by φ0. The misalignment angle is typically calibrated from compression tests and ranges from 1° to 3° for aerospace-grade carbon/epoxy. LaRC04 extends this to three-dimensional stress states and adds a criterion for matrix failure under combined transverse compression and shear that is more accurate than Puck's for high hydrostatic pressure. The trade-off is calibration complexity: LaRC requires the fracture toughness values G_Ic and G_IIc in addition to the strength values, because the criterion is formulated in terms of energy release rates for the kinking and matrix modes.

Christensen: Isotropic Tensor-Based Failure Theory

Christensen's failure theory (1997) takes a different approach: it treats the composite as a transversely isotropic continuum and derives failure criteria from tensor invariants. The theory defines two failure modes — fiber-dominated and matrix-dominated — using stress invariants that are independent of coordinate rotation. For transverse isotropy with the fiber direction as the axis of symmetry, the matrix failure criterion is:

(σ22 + σ33)² / Y_T² + (σ22 - σ33)² / S23² + τ23² / S23² + τ12² / S12² + τ13² / S12²  ≥ 1

The Christensen theory is mathematically elegant and requires only four independent strength parameters, but it is less commonly implemented in commercial progressive-damage codes because its invariant formulation does not map naturally onto the mode-specific damage variables that drive stiffness degradation. It is most useful as a screening criterion for preliminary sizing, where its low calibration cost and smooth failure surface are advantageous.

Selecting and Implementing a Criterion

The choice of criterion is not merely academic — it determines which damage variables are activated, which in turn controls the post-initiation stiffness degradation and the predicted failure sequence. The following table summarizes the practical trade-offs:

In a progressive-damage framework, the criterion is evaluated at each integration point at every increment. When a failure index reaches 1.0, the corresponding damage variable (e.g., d_ft, d_fc, d_mt, d_mc) is activated and the stiffness matrix is degraded according to a user-defined degradation law — typically a sudden reduction to a small residual value or a continuum damage mechanics (CDM) evolution law. The critical implementation detail is that the criterion must be evaluated in the material coordinate system, and the stress state must be updated after each degradation event to avoid spurious stress redistribution.

Worked Example: Hashin Initiation in a [0/90]s Laminate

Consider a 16-ply [0/90]s carbon/epoxy laminate with X_T = 1500 MPa, X_C = 1200 MPa, Y_T = 50 MPa, Y_C = 200 MPa, and S12 = 70 MPa. Under uniaxial tension in the 0° direction, the 90° plies experience transverse tension. The Hashin matrix tension index for a 90° ply at applied strain ε11 = 0.5% is:

σ22 = E22 · ε11 = 9000 MPa · 0.005 = 45 MPa
F_mt = (45 / 50)² = 0.81  → no initiation

At ε11 = 0.56%, σ22 = 50.4 MPa and F_mt = 1.016, so matrix tension initiates in the 90° plies. The 0° plies remain intact because their fiber-direction stress is far below X_T. This example illustrates the mode-separation benefit: the criterion correctly identifies that the first damage event is transverse matrix cracking, not fiber failure, which matches experimental observation for this layup.

The diagram above maps the evaluation flow from the applied stress state through criterion selection to mode-specific damage variables and stiffness degradation. The feedback loop is essential: after degradation, the stress state is redistributed, and the criterion must be re-evaluated at the next increment. This is what distinguishes a progressive-damage analysis from a simple first-ply-failure check.

Practical Implementation Notes

When implementing these criteria in a UMAT or VUMAT subroutine, three details consistently cause problems. First, the fracture-plane search in Puck and LaRC must be performed in the material coordinate system, and the transformation must use the full 3D stress tensor — neglecting σ33 or τ23 will produce incorrect fracture angles for thick laminates. Second, the Hashin criterion's shear interaction term α should be set to 1.0 for matrix modes and 0.0 for fiber compression modes in most implementations; using a single value for all modes is a common error. Third, the degradation law must be regularized to avoid mesh-dependent results — a characteristic length or fracture-energy-based regularization is strongly recommended when using CDM evolution rather than instantaneous degradation.

For a first implementation, Hashin with mode-specific instantaneous degradation is the most robust starting point. Once the analysis converges and the failure sequence is physically plausible, upgrading to Puck or LaRC for compression-critical plies will improve accuracy without requiring a complete reformulation of the progressive-damage framework.