Paris Law (Fatigue Crack Growth)

The idea: damage you can't undo, paid in tiny installments

Bend a paper clip back and forth. Nothing happens for the first few flexes — then a faint crease forms, then a crack, then it snaps. You never exceeded the static strength of the wire; you broke it by repetition. That is fatigue, and the crack didn't appear all at once. It advanced a microscopic step on each bend, and every step was permanent. Run the cycles fast enough and a part rated for a 50-year life can fail in an afternoon.

Paul Paris asked the obvious follow-up question in 1961: how big is each step? His answer — initially rejected by three journals before he published it himself — is that the growth per cycle depends almost entirely on one number, the stress-intensity-factor range ΔK, raised to a power. Not the peak stress, not the mean stress, not the crack length alone, but the combination of stress and crack size that fracture mechanics calls K. That single insight turned fatigue from a statistical black box into a quantity an engineer can integrate to a number of cycles.

The practical payoff is enormous. Before Paris' law, the rule for a fatigue-critical part was "retire it before any crack can form" (safe-life). After it, you could say "assume the worst undetected flaw is already in there, prove it grows slowly enough that an inspection will catch it before it gets dangerous" (damage tolerance). That shift is why a commercial airframe can fly 90,000 pressurization cycles with cracks growing in it the whole time — and still be safe.

How Paris' law works: ΔK is the throttle

The stress-intensity factor K is the fracture-mechanics measure of how hard a crack tip is being pried open. For a crack of length a in a body under remote stress σ, it is:

K = Y · σ · √(π·a)

  Y   geometry factor (dimensionless, ~1 for a small edge crack)
  σ   remote applied stress              [MPa]
  a   crack length (or half-length)      [m]
  K   stress-intensity factor            [MPa·√m]

Under cyclic load, σ swings between σ_min and σ_max, so K swings too. Paris' law says the crack growth per cycle tracks that swing:

da/dN = C · (ΔK)^m

  ΔK   = K_max − K_min = Y · Δσ · √(π·a)   [MPa·√m]
  da/dN = crack extension per cycle         [mm/cycle or m/cycle]
  C, m  = material constants (from testing)

  Δσ   = σ_max − σ_min  (stress range)
  R    = K_min / K_max  = σ_min/σ_max (stress ratio — affects C, m, ΔK_th)

The crucial feedback loop: as the crack grows, a increases, so ΔK increases even at constant applied stress range. A bigger ΔK means a bigger da/dN, which grows the crack faster, which raises ΔK again. Crack growth is self-accelerating. Plot crack length against cycles and you get a curve that creeps along nearly flat for most of the life, then sweeps almost vertical at the end. By the time you can see a fatigue crack with the naked eye, most of the fatigue life is already gone.

On a log–log plot of da/dN versus ΔK, the data form an S-curve with three regions:

• Region I (threshold). Just above ΔK_th. Growth drops toward zero; below the threshold the crack effectively does not propagate. The curve bends sharply down — Paris' straight line over-predicts here.
• Region II (Paris régime). The long straight middle. This is where da/dN = C·(ΔK)^m holds, with m the slope of the line in log–log space. Most of the integrable life lives here.
• Region III (fast fracture). K_max nears the fracture toughness K_IC. Growth runs away, the curve sweeps up, and the part is cycles away from rupture. Paris under-predicts here.

The math: integrating to fatigue life

Predicting life means integrating the growth law from the starting flaw a_i to the critical crack a_c, where K_max first reaches the fracture toughness. Treating the geometry factor Y as constant gives a closed form:

da/dN = C·(Y·Δσ·√(π·a))^m

Separate and integrate from a_i to a_c:

         ⌠ a_c   da
  N_f =  │      ─────────────────────
         ⌡ a_i   C·(Y·Δσ·√π)^m · a^(m/2)

For m ≠ 2:

         a_c^(1−m/2) − a_i^(1−m/2)
  N_f = ───────────────────────────────
          C·(Y·Δσ·√π)^m · (1 − m/2)

Critical crack length (set K_max = K_IC):

  a_c = (1/π) · ( K_IC / (Y·σ_max) )^2

Two facts fall out of this integral that every fatigue engineer internalizes:

• Life is dominated by the small-crack stage. Because of the a^−m/2 in the integrand, the early cycles (small a) consume the most cycles per millimetre. For m = 3, doubling a_i (1 → 2 mm) cuts N_f by about a third; quadrupling it (1 → 4 mm) costs roughly 60% — and the sensitivity steepens fast as m rises. This is why "what is the largest flaw inspection could miss?" is the single most important input — it sets a_i.
• Stress range hits life by the m-th power. N_f ∝ Δσ^−m. With m = 3, cutting the stress range by 20% (×0.8) multiplies life by 0.8⁻³ ≈ 1.95 — almost double. A 30% cut nearly triples it. Mean stress barely enters (it shifts R, which nudges C and ΔK_th); the range is what matters.

Worked example: a pressurized fuselage panel

Take an aluminium 2024-T3 fuselage skin with a small through-crack at a rivet hole. Each ground-air-ground cycle pressurizes the cabin, putting the skin in hoop tension.

Material: 2024-T3 aluminium
  C   = 1.6e-11   (da/dN in m/cycle, ΔK in MPa·√m)
  m   = 3.0
  K_IC ≈ 34 MPa·√m

Loading (one flight = one cycle):
  σ_max = 100 MPa hoop stress (cabin pressurized)
  σ_min = 7 MPa  (on the ground)
  Δσ    = 93 MPa
  Y     ≈ 1.0   (small crack, far from edges)

Initial detectable flaw:  a_i = 1.0 mm = 0.001 m
Critical crack length:
  a_c = (1/π)·(K_IC/(Y·σ_max))^2
      = (1/π)·(34/100)^2 m
      = 0.0368 m ≈ 37 mm

Cycles to failure (m=3, so 1−m/2 = −0.5):
  denominator = C·(Y·Δσ·√π)^3·(1−m/2)
  Δσ·√π = 93·1.7725 = 164.8
  (164.8)^3 = 4.48e6
  C·(...) = 1.6e-11 · 4.48e6 = 7.16e-5
  × (−0.5) = −3.58e-5

  numerator = a_c^(-0.5) − a_i^(-0.5)
            = (0.0368)^(-0.5) − (0.001)^(-0.5)
            = 5.21 − 31.62 = −26.41   (units 1/√m)

  N_f = (−26.41) / (−3.58e-5) ≈ 7.4e5 cycles

So about 740,000 flights from a 1 mm crack to fracture — comfortably more than a 90,000-cycle design service life. But now repeat the calculation with a 4 mm initial flaw (a missed inspection): a_i^−0.5 drops to 15.8, the numerator becomes −10.6, and N_f falls to roughly 296,000 cycles — a 60% loss of life from a 3 mm change in assumed flaw size. That sensitivity, not the absolute number, is the real lesson. Inspection intervals are set so that even the largest flaw that could slip past one inspection cannot grow to a_c before the next.

Typical Paris constants for real materials

C and m are measured per ASTM E647 on compact-tension or middle-tension specimens. Values vary with units (be ruthless about whether ΔK is in MPa·√m and da/dN in m/cycle, as tabulated below, versus mm/cycle or in/cycle elsewhere), stress ratio R, environment, and temperature. Representative room-air values at R ≈ 0.1:

Material	m	C (m/cycle, MPa·√m)	ΔKth (MPa·√m)	KIC (MPa·√m)
Ferritic-pearlitic structural steel	~3.0	~6.9 × 10⁻¹²	5 – 8	40 – 80
Martensitic / high-strength steel	~2.2 – 3.3	~1.4 × 10⁻¹¹	2.5 – 5	20 – 60
Austenitic stainless (304/316)	~3.2	~3 × 10⁻¹²	4 – 7	150 – 250 (J-based)
2024-T3 aluminium	~3.0	~1.6 × 10⁻¹¹	2 – 4	~34
7075-T6 aluminium	~3.5 – 4	~5 × 10⁻¹¹	1.5 – 3	~24
Ti-6Al-4V titanium	~3.5 – 4	~1 × 10⁻¹¹	3 – 6	~75
Nickel superalloy (Inconel 718)	~3 – 4	~5 × 10⁻¹²	5 – 10	~90

The exponent m is the more transferable of the two — it clusters near 3 for most engineering metals and is the slope you'd eyeball off a log–log da/dN plot. C is units-bound and material-specific; a higher C shifts the whole Region-II line up, meaning faster growth at any ΔK. A handy sanity check: at ΔK = 10 MPa·√m, a steel with C ≈ 7 × 10⁻¹² and m = 3 grows at 7 × 10⁻¹² × 10³ = 7 × 10⁻⁹ m/cycle ≈ 7 nanometres per cycle — roughly 25 atomic spacings, which matches the striation spacings seen on fatigue fracture surfaces under a microscope.

Beyond Paris: Forman, Walker, and NASGRO

The bare power law ignores three real effects: the threshold, the toughness ceiling, and the stress ratio. Engineering codes use extended forms that bolt these back on:

	Paris	Walker	Forman	NASGRO (Forman–Newman)
Equation core	C·ΔKm	C·(ΔK/(1−R)1−γ)m	C·ΔKm / ((1−R)KIC−ΔK)	full sigmoidal w/ thresh + closure
Captures Region I (threshold)	No	No	No	Yes (ΔKth term)
Captures Region III (fracture)	No	No	Yes	Yes
Stress-ratio R effect	No (single R)	Yes	Yes	Yes (closure-based)
Crack closure	No	Implicit	No	Yes (explicit f-function)
Number of constants	2 (C, m)	3	3 – 4	~7+
Typical use	Hand calcs, Region-II life	Variable-R spectra	Life near fracture	AFGROW / NASGRO software, certification

One physical effect threading through all of these is crack closure (Elber, 1970): the crack faces touch before the load reaches zero, because plastically stretched material in the wake props the crack open from below. Only the portion of the cycle above the opening load, ΔK_eff, actually drives growth. Closure explains why low-R loading grows slower than the nominal ΔK predicts and why an overload can retard subsequent growth — it leaves a plastic zone the crack must chew back through.

Where it's used: damage-tolerant design in the field

Domain	What Paris' law sets	Real driver
Commercial aircraft (FAA/EASA damage tolerance)	Inspection intervals between detectable flaw and critical crack	Mandated after the 1988 Aloha Airlines 737 fuselage rip; multi-site fatigue cracking at rivet rows
Pressure vessels & piping (ASME BPVC, API 579)	Remaining life of a found flaw, fitness-for-service "run or repair"	Pressure cycling, thermal transients, weld defects
Steel bridges (AASHTO, Eurocode 3 fatigue)	Fatigue category & inspection of welded details	Truck-pass stress cycles; weld toes are crack starters
Railway axles & wheels	Crack-arrest hold and replacement mileage	Rotating-bending cycles; rare but catastrophic (Eschede, 1998)
Offshore & wind-turbine welds	Sea-state spectrum life; weld toe crack growth	Millions of low-amplitude wave/gust cycles
Gas-turbine discs & blades	Cyclic life of fan/disc bores, life-limited parts	Start-stop + steady cycles; tracked by NASGRO/AFGROW
Nuclear reactor components	ASME Section XI flaw-acceptance and growth	Thermal fatigue, environmentally assisted cracking

In every one of these, the workflow is the same triangle: (1) the largest flaw a non-destructive inspection might miss sets a_i; (2) the fracture toughness and peak stress set a_c; (3) Paris' law integrates the cycles between them, and the inspection interval is set to a fraction of that — typically half — so a flaw is guaranteed to be caught before it grows critical. Tools like NASA's NASGRO and the USAF's AFGROW automate the integration over realistic load spectra, but the engine inside them is still da/dN = C·(ΔK)^m.

Common misconceptions and pitfalls

• "Paris' law covers all of fatigue." It covers only crack propagation, and only in Region II. It says nothing about crack initiation — the cycles spent nucleating a crack at a smooth surface, which can be most of the life for a polished, defect-free part. For initiation you need S-N curves or strain-life (ε-N). Paris assumes a crack already exists.
• Mixing up units of C. C is meaningless without its units, and they're a minefield: ΔK in MPa·√m vs ksi·√in, da/dN in mm/cycle vs m/cycle vs in/cycle. A published C can differ by orders of magnitude purely from unit convention. Always confirm the unit set before plugging a tabulated C into your integral.
• Treating Y as constant when it isn't. The geometry factor Y depends on crack length relative to part size. For a crack approaching a free edge or a finite-width plate, Y climbs steeply and the closed-form N_f integral breaks down — you must integrate numerically with Y(a) updated each step.
• Ignoring stress ratio and overloads. A single tensile overload can retard growth for thousands of cycles (compressive residual stress in its plastic zone); a compressive underload can accelerate it. Constant-amplitude Paris constants systematically mis-predict life under a real variable-amplitude spectrum unless retardation/closure models are added.
• Forgetting the environment. C and m for "room air" can be wildly optimistic in seawater, hydrogen, or at high temperature. Corrosion-fatigue and environmentally assisted cracking can raise growth rates by 10× or more and lower ΔK_th. Offshore and nuclear codes carry separate, harsher constants for this reason.
• Extrapolating past Region III. Once K_max nears K_IC the crack is in fast fracture and the power law badly under-predicts. The integral must stop at a_c — pushing it further invents life that doesn't exist.