A bonded joint that has come apart carries its own history on the two faces it left behind. Before any testing, before any records are pulled, the single most informative observation is simply where the separation ran: cleanly off one substrate, through the middle of the adhesive, or into the material being bonded. Those three outcomes point toward different causes and different responsible parties, and they are distinguishable by eye long before a laboratory is involved. ASTM D5573 exists to make that observation repeatable rather than impressionistic, and it is usually the first exhibit produced in any dispute over a bonded assembly.
Three outcomes, three different stories
Adhesive failure, sometimes called interfacial failure, describes a separation that runs along the boundary between the adhesive and the substrate, leaving one face essentially bare. Cohesive failure runs through the bulk of the adhesive itself, leaving residue on both faces. Substrate failure occurs when the material being bonded gives way first, tearing fibres out of a composite or yielding in a metal while the bondline stays intact.
The three are not gradations of the same problem. They point in genuinely different directions, which is why the classification is done before rather than after a theory of the case has formed.
What a bare substrate implies
A joint that peels off a clean, unmarked surface with little or no adhesive transferred is the classic signature of an interface that never achieved proper adhesion. The usual candidates are inadequate surface preparation, a contaminant such as mould release or machining oil, an incompatible or improperly applied primer, or an open time exceeded before the parts were mated. What the observation does not do on its own is name which of those occurred — it narrows the inquiry to the interface and to the process that created it.
What a cohesive failure implies
When the crack runs through the adhesive and both faces carry a reasonably uniform coating, the interface itself performed. That is often the best available outcome for a bonded joint, because it means the assembly reached the adhesive's own strength rather than failing at a preventable defect. Whether it is good news depends entirely on what that strength was and what it was supposed to be. A joint that reached its rated capacity and still failed raises a design-allowable or adhesive-selection question. A joint whose cure data shows it never reached full strength raises a process question instead.
Substrate failure is not automatically a clean result
Fibre tear in a composite or yielding in a metal adherend means the bond outperformed the part it was holding. That normally exonerates the bonding process. It also relocates the question: either the joint was over-designed relative to the structure, or the structure was under-designed for the load it actually saw. Either way the inquiry moves off the adhesive and onto the design or the loading condition, which is a materially different matter than a bondline defect.
Real joints are mixed, and the percentages carry the finding
Few failures are purely one mode. A large bonded area commonly shows interfacial release over part of its surface, cohesive fracture elsewhere, and sometimes substrate involvement at the most highly loaded edge. D5573 accommodates this by asking for the percentage of each mode across the bonded area rather than a single label, and that distribution is frequently more informative than any one classification.
A joint that is ninety per cent cohesive with a small interfacial patch at one corner tells a different story than one that is ninety per cent interfacial with a cohesive island in the middle.
Where the interfacial region sits matters
The location of interfacial release within the joint is diagnostic. Release concentrated at an edge is consistent with peel stress or with moisture ingress working inward from the perimeter over time. Release concentrated at a specific area away from any edge more often reflects localised contamination or a region the surface treatment did not reach. Mapping the modes onto the joint geometry, rather than tallying them in aggregate, is what converts a classification into an argument.
What the classification cannot settle by itself
Visual classification identifies where the joint separated. It does not establish why. A bare substrate is consistent with several distinct process faults, and distinguishing them requires the analytical work that follows: spectroscopy on both faces to identify contamination or confirm adhesive chemistry, thermal analysis for residual cure exotherm and glass transition, and contact-angle measurement on exemplar parts to assess whether the surface treatment was producing an adequately wettable surface at the time.
Why the faces have to survive the handling
Everything above depends on the fracture surfaces reaching a laboratory in the condition they separated in. Solvent-wiping removes the contamination that would have identified the cause. Re-mating the halves to see how they fit transfers material between faces and destroys the percentages. Handling the surfaces directly adds skin oils to a chemistry that is about to be analysed for exactly that kind of species. Photograph both faces as separated, and preserve them without cleaning, touching or reassembly.
Where the classification is contested
Disputes over bonded assemblies frequently turn on the mode distribution, because it maps so directly onto responsibility — the interface belongs to whoever prepared and assembled it, the bulk adhesive belongs to whoever formulated and specified it, and the substrate belongs to the designer. That is precisely why the classification should be documented before any party has cleaned, sectioned or retested anything, and why the percentages are worth recording carefully rather than summarising as a single word.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.