How Science Works | How Tiny Cracks Turn into Material Failure

For a simple tensile specimen, nominal stress is force divided by the chosen cross-sectional area. It is useful for comparing loads across different specimen sizes. But it is an average description, not a map of local stress.

A hole, notch or crack changes the way forces pass through the material. Near a sharp crack tip, local stresses can be much larger than the nominal value. The geometry matters: a long, narrow crack is not equivalent to a rounded hole with the same missing area.

Think of pedestrians trying to pass around a barrier in a corridor. The average number of people per square metre does not tell you where the crowd is most compressed. Likewise, an average stress can conceal a severe local concentration. The analogy helps locate the issue; it is not a quantitative model of a solid.

A high tensile-strength value therefore cannot, by itself, establish the resistance of a cracked component. The existing Materials Science guide compares strength, hardness, stiffness and toughness. Here the narrower task is to understand what makes an existing crack extend.

Contents · Next chapter

At the tip, bonds can separate and surrounding material may deform. Some materials dissipate substantial energy through plastic deformation, damage or other local processes before the crack advances. Others provide less opportunity for such dissipation under the same type of loading.

This is why “harder” does not automatically mean “more resistant to cracking”. Hardness concerns resistance to indentation or related local deformation. Fracture resistance concerns the conditions required for a crack to advance. They are different questions even when processing influences both properties.

In an ideal elastic model, stress increases without bound as distance to a perfectly sharp tip approaches zero. Real materials do not sustain literal infinite stress. The mathematical singularity tells us that the idealisation cannot describe the immediate tip at every scale. A finite process region develops there.

The useful scientific move is not to discard every model containing an idealisation. It is to identify where the approximation still predicts observations. A description outside the tiny tip region may remain useful even though the atom-by-atom behaviour at the tip requires another model.

Contents · Previous chapter · Next chapter

Extending a crack makes new surfaces and can require irreversible deformation around the tip. Those processes need energy. Meanwhile, parts of a loaded specimen relax as the crack grows, and the loading system can supply or release energy. Growth depends on the balance.

The energy release rate, commonly called G, expresses the available change in mechanical energy per increment of crack area under specified conditions. It is compared with an appropriate resistance to extension. It is not simply the total energy stored anywhere in the object.

Crack initiation and continued propagation are also different events. A crack may begin moving and then slow or stop if its driving force falls or resistance rises. Alternatively, extension may make the next increment easier and lead to unstable growth.

Loading arrangements matter. A test machine holding displacement approximately fixed does not exchange energy with a specimen in exactly the same way as an ideal system holding force fixed. A photograph of the final fracture surface cannot tell the whole loading history.

David Roylance’s MIT introduction to fracture mechanics develops the connection between crack size, energy and fracture resistance. This guide uses that framework qualitatively rather than treating a classroom formula as a design approval.

Contents · Previous chapter · Next chapter

For suitable linear-elastic opening-mode problems, a common form is K_I = Yσ√(πa). Here σ is a specified nominal stress, a is the crack-size parameter defined for that geometry, and Y is a dimensionless geometry factor. K_I is the Mode I stress-intensity factor.

Do not guess what a means from a sketch. It may be a depth or a half-length depending on the solution. Likewise, Y is not a universal constant. The specimen shape, boundaries and loading arrangement determine which expression is appropriate.

A critical value can characterise resistance under qualified conditions. The notation K_Ic has a particular plane-strain fracture-toughness meaning; a convenient failure number measured in an arbitrary specimen is not automatically K_Ic.

The units make the distinction visible. If stress is in MPa and crack length is in metres, K has units MPa√m, not MPa. It cannot be substituted directly for tensile strength in a calculation. For translation and notation details, use the existing fracture-toughness terminology guide; this article focuses on the physical sequence.

Contents · Previous chapter · Next chapter

Illustrative model, not a safety calculation. Assume two specimens have the same material resistance and approximately the same geometry factor over the comparison. One has crack parameter a; the other has crack parameter 4a. Compare the nominal stresses at which each reaches the same chosen critical K value.

Because K scales with σ√a, increasing a fourfold doubles √a. To keep K unchanged, σ must be halved. The specimen with the longer crack reaches the same model threshold at half the nominal stress.

For example, if the first hypothetical case reaches the threshold at 120 MPa, the second reaches it at 60 MPa under those assumptions. These numbers are invented to illustrate the square-root relationship. They are not material-property data and should not be used to judge a real component.

Now challenge the comparison: would Y remain constant as the crack approached an edge? Is the specimen still predominantly elastic? Does the crack have the assumed orientation? Are residual stresses present? Each answer can change the prediction.

NASA’s experimental–modelling study of a pinned-end notched specimen illustrates why a familiar stress-intensity solution still needs validation for the actual geometry and loading. A correct-looking equation is not evidence that its conditions have been met.

Contents · Previous chapter · Next chapter

A component can survive one load cycle and still accumulate crack extension across many cycles. Fatigue crack growth need not wait until the peak load reaches the threshold for a single rapid fracture event.

The important history includes the loading range, mean level, sequence and environment. A larger overload does not necessarily have a simple, permanent multiplying effect on all later growth: changes near the tip and crack closure can complicate the subsequent response.

The final separation may be sudden, while the preceding growth was gradual. Calling the whole event “instantaneous failure” can therefore hide the stage where evidence accumulated. Conversely, a visible crack does not tell an unqualified observer exactly how much life remains.

NASA research on crack growth after overloads and underloads connects loading history with the effective driving force. Its existence is a reminder to ask for the load sequence rather than only the largest number in a log.

This article explains the mechanism, not inspection intervals or repair decisions. If a real load-bearing object appears cracked, do not test it by applying more load; take it out of use where safe and seek qualified assessment.

Contents · Previous chapter · Next chapter

Start with a precise question: are we testing crack initiation, growth rate, final fracture resistance or the origin of a particular failure? These require different observations. A specimen breaking once is not a complete answer to all four.

For a controlled comparison, record material condition, specimen geometry, crack dimensions and orientation, loading arrangement, temperature and loading history. Measure crack extension alongside force and displacement. State the resolution of the crack measurement; changes smaller than that resolution should not be reported as confidently observed growth.

Surface examination can help locate an origin and distinguish regions formed under different processes. It should be combined with records and mechanical analysis, not treated as a magical photograph that explains everything. Corrosion, residual stress and manufacturing history may be relevant alternative or interacting causes.

When a model disagrees with a test, examine its validity before changing a material constant until the curve fits. NASA’s discussion of fracture-mechanics validity limits explicitly distinguishes assumptions behind linear-elastic parameters from more demanding fracture behaviour.

For school learning, annotated test videos and supplied datasets are sufficient. Do not deliberately fracture glass, blades, pressurised vessels or structural parts. The learning goal is to connect evidence with a model, not to reproduce an uncontrolled failure.

Contents · Previous chapter · Next chapter

“The average stress is low, so the crack cannot matter.” Incorrect. The average does not describe the local crack-tip field or the energy available for extension.

“A material is hard, so it must resist crack growth well.” Not established. Hardness and fracture resistance describe different responses and need appropriate tests.

“The crack is twice as long, so its driving force must be twice as large.” Not a general rule. In the stated K model, with stress and Y held fixed, doubling a multiplies K by √2. Other measures and changing geometry must be treated separately.

“A fitted equation proves the failure mechanism.” No. More than one model can fit a limited dataset. Independent measurements, valid assumptions and predictions outside the fitting set strengthen the argument.

The central chain is flaw geometry → local loading and energy release → material resistance → extension history → possible final failure. Return to Materials Science for the wider structure–processing–properties relationship, or use the linked MIT and NASA sources to examine the underlying mechanics.

Contents · Previous chapter · Continue to Materials Science