BlogApplication notes
Why did a bond fail months later?
- Type
- Application notes
- Published
- 20 November 2026
- Prepared by
- Technical team
- Product group
- Adhesives
When a bond fails months later, it is easy to assume the adhesive was weak, but a strength test on a fresh joint is poor at catching weak spots at the interface, and water, heat and load expose them over time. What actually brings it down can be poor surface preparation, water that got in, unequal thermal expansion, a load that never lets up, or incomplete cure, and often two or more at once.
Sometimes a part that passed final inspection and gave no trouble for its first weeks comes back months later with the joint open. On the returned part, the adhesive has come away whole from one face, lifted bit by bit from the edges, or torn within itself. This kind of failure shows up after the part has reached the customer, so it turns into returns, field repairs and safety issues, at a cost far above catching it in process.
Adhesion forms at the interface between the adhesive and the adherend, the part being bonded, through molecular attraction and through mechanical interlocking with a rough surface. For that to work, the adhesive must wet the surface well and the surface itself must be sound (Kinloch, 1987; da Silva et al., 2018). If a weak layer sits on the surface, the load breaks through that layer, which is called a weak boundary layer (Bikerman, 1968).
The trouble is that a weak interface still gives good strength when fresh. It may pass a strength test in which two overlapped strips are pulled apart briefly and dry, yet fail first once water and stress act together. Work on bonded aircraft structures found that durability in humid service separated good and poor surface treatments better than initial strength did (Venables, 1984; ASTM International, 2024c). So the biggest clue to the cause is not the strength on day one but where the joint broke. If the adhesive came away whole from one face, the interface was weak; if it tore within itself and remains on both faces, the interface held and the adhesive gave way to load, heat or moisture. If the adherend broke first, the bond did its job. The international standard names and classifies these failure patterns, and asks for test results to be recorded with the pattern, not as numbers alone (ISO, 2022).
In such cases the technical team sets up the five hypotheses below and rules them out one by one. The start is always the same: keep the failed part and photograph the fracture surface, then note the date and lot it was made in, the materials joined, the temperature and humidity where it served and the load it carried. That one record screens several hypotheses at once.
Hypothesis 1: poor surface prep
When the adhesive stays whole on one face of the fracture and the other face shows clean bare material, look at this hypothesis first. The standard calls this pattern adhesive failure, a break at the interface (ISO, 2022). It shows up on metal parts still carrying cutting fluid or rust preventive, on plastic parts moulded with release agent, on parts bonded long after treatment, and on parts carried bare-handed or in oily gloves. Say injection-moulded plastic covers are bonded to metal frames and shipped. The coupons pulled at shipment were fine, yet on the parts that come back months later the adhesive sits smooth on the frame alone, and the inside of the cover looks as clean as if it had never been bonded.
What the adhesive actually gripped was not the cover but a thin layer on top of it. Metal surfaces carry thin films of cutting fluid, rust preventive or skin oil, and plastics carry release agent or additives that have bloomed out from inside. These layers lower the surface energy, the degree to which a surface attracts a liquid, so the adhesive cannot spread (Ebnesajjad, 2014). The adhesive bonds to the contaminant, the contaminant clings only loosely to the part, and the load breaks through that layer (Bikerman, 1968; Kinloch, 1987). In soft PVC or rubber, the plasticiser added to make the material soft can migrate slowly to the interface after bonding and build a new weak layer months later (Bikerman, 1968).
On metals the oxide matters too. Native oxide or rust is loose and weak, so the joint breaks inside the oxide however well the adhesive grips it. Water makes this worse. Aluminium oxide hydrates in water, swelling into a weak hydroxide, and when this happens at the interface the joint opens up over months (Venables, 1984). That is why the same adhesive lasts very differently in humid service on a surface that was only degreased and on one given a stable oxide by etching or anodising (Venables, 1984; ASTM International, 2024b).
A simple test shows whether the surface was clean. In the water-break test, water is flooded over the prepared metal surface to see whether the film holds; if it breaks into beads, a water-repelling contaminant remains. Sensitivity drops on rough or porous surfaces, though (ASTM International, 2021). On plastics such as polyethylene and polypropylene, use the wetting tension test, which brushes on standard test liquids to see whether they spread. Wetting tension alone does not guarantee adhesion, but it shows when the treatment process has drifted (ASTM International, 2023a). In the records, look at the time from degreasing, abrading or priming to bonding, and whether the parts were touched bare-handed or with oily gloves in between.
If surface preparation turns out to be the cause, fix the order of treatment first. Keep to degrease, abrade or blast, then degrease again, and pick the treatment for each metal from standard guidance (ASTM International, 2024b). Low-energy plastics need flame, corona or plasma treatment or a dedicated primer (Ebnesajjad, 2014). Put the water-break or wetting tension test on the process checklist, and set a maximum time from treatment to bonding, re-treating any part that exceeds it.
Hypothesis 2: water ingress
When water is the cause, the timing and the place give it away. Failures cluster after the rainy season or summer, or occur only on equipment that is washed often, inside housings where day and night temperature swings bring condensation, or on metal joints outdoors. Take a housing with an aluminium panel bonded on, standing beside equipment that is often splashed with wash water. It gets through its first winter without trouble, but after the summer the panel edges begin to lift, and when it is pulled off, the edges show bare, discoloured metal while further in the adhesive is torn and left on both faces.
Even cured adhesive slowly takes up water. Water molecules slip between the polymer chains and let them move more freely, which is called plasticisation. The glass transition temperature, the point where the adhesive turns from rigid to rubbery, falls, and stiffness and strength drop (Kinloch, 1987; da Silva et al., 2018). Damage at this stage is largely reversible. Epoxy lap joints soaked in 50°C water lost strength the longer they soaked, and almost fully recovered once dried (Mubashar et al., 2009).
Irreversible damage arises at the interface and in chemical bonds. Water travels faster along the metal-adhesive interface than through the bulk adhesive. In a study at 70°C and about 100% relative humidity, water diffusion in joints was considerably faster than in the same adhesive in bulk (Zanni-Deffarges & Shanahan, 1995). Water reaching the interface displaces the bonds between adhesive and surface and hydrates the oxide (Venables, 1984). In polyurethane adhesives, long exposure to heat and humidity together can hydrolyse the polyester segments of the chain, breaking them by reaction with water (Dubelley et al., 2018).
To tell whether it is water, dry a still-intact part from the same lot that saw the same conditions and see whether strength comes back. If it almost fully returns, the adhesive was softened by absorbed water; if it does not, the interface or the chemical bonds are probably damaged (Mubashar et al., 2009). Discolouration or corrosion product at the fracture edges marks where water came in. To reproduce the failure, use the wedge test. A wedge is driven into the bondline to hold the interface in tension, the joint is exposed to humidity and crack growth is tracked; it is known to match service performance better than lap-shear or peel tests (ASTM International, 2024c).
For parts in humid service, write strength after humid-heat ageing into the requirement before choosing the adhesive. On metals, go beyond degreasing to a treatment that builds a stable oxide, plus a primer (Venables, 1984; ASTM International, 2024b). Seal or paint over the bondline edges to close the way in for water, and design out places where water can pool.
Hypothesis 3: thermal mismatch
Thermal expansion becomes a problem where unlike materials are joined. It shows up when joints between aluminium and steel, metal and plastic or metal and glass sit where the temperature range is wide, outdoors, near ovens or on refrigeration equipment, and it is worse on long or large parts. A glass panel bonded into an aluminium frame and set up outdoors is one example. After months of sunny days and cool nights, gaps have opened first at the ends and corners of the joint and crept inward bit by bit.
The root of the problem is the coefficient of thermal expansion, the rate at which a material grows when heated and shrinks when cooled, which differs from one material to another. When two materials move by different amounts as temperature changes, the bondline takes up the difference in shear and peel, and this stress peaks at the ends of the joint (da Silva & Adams, 2007; da Silva et al., 2018). A heat-cured adhesive sets at its cure temperature and so starts life with residual stress once it cools (Kinloch, 1987).
A joint may survive one temperature change, but day and night, the seasons and machine start-stop cycles turn that stress into a fatigue load. A study that measured strain in heated joints of dissimilar materials saw temperature change alone, with no outside load, produce large strain in the bondline and drive it to failure (Bagiatis et al., 2021). A rigid adhesive cannot take the strain and cracks start at the ends, while a flexible one takes the strain but is weaker when hot, so joints used over a wide temperature range need a balance of the two (da Silva & Adams, 2007). Sealant joints in buildings and equipment follow the same principle, with adhesion and cohesion tested together when water, repeated movement and temperature change combine (ASTM International, 2022).
Look first at where the debonding began. If it started at the ends or corners and spread inward, and the same adhesive gives no trouble between like materials, this hypothesis gains weight. Tabulate the expansion coefficients of the two materials beside the real service temperature range, and reproduce the failure by comparing strength before and after thermal cycling.
The mismatch itself cannot be removed, so let the adhesive and the joint shape absorb it. For combinations with a large mismatch, look first at room-temperature-curing or flexible adhesives, and do not make the bondline too thin. Taper the joint ends, or leave a fillet, a sloped bead of adhesive at the end, to lower the stress there (da Silva et al., 2018). Check strength across the whole service temperature range, and for sealant joints pick a grade rated for the expected movement (ASTM International, 2022).
Hypothesis 4: sustained load
This hypothesis applies where the joint is under load all the time. Typical cases are brackets and hangers carrying weight, fastened joints under clamping force, joints that hold a bent sheet flat against its spring-back, and structures where off-centre loading bends the joint. A bracket bonded to a wall with a heavy part hung from it makes the picture clear. At first it does not budge, then over months the gap between the top of the bracket and the wall opens almost unseen, until one hot day the whole thing falls.
Load on a lap joint does not spread evenly over the bonded area. Because the adherends stretch by different amounts along the overlap, shear stress concentrates at both ends, and when off-centre loading bends the joint, peel stress also arises, prising the bondline apart at the ends (Goland & Reissner, 1944; da Silva et al., 2018). This is why a longer overlap does not add strength in proportion. Adhesives are strong in shear and compression and weak in peel and cleavage, so a design that creates peel fails at the ends even at a low average stress (Kinloch, 1987).
Add time, and creep follows. Creep is the slow stretching of a polymer under a steady load, and it speeds up as heat or absorbed water brings the adhesive closer to its glass transition temperature (Kinloch, 1987; ASTM International, 2024a). A sustained load that caused no trouble at first, such as the part's own weight, a clamping force or a spring force, grows the strain at the ends over months, and that is where the crack starts. When moisture and stress act together, damage comes faster than with either alone (Venables, 1984; ASTM International, 2024c).
On the floor, begin by writing down what load acted at that spot and for how long. A gap that opened slowly before the failure points to creep. The fracture leans towards cohesive failure, torn within the adhesive (ISO, 2022), sometimes with stretched, whitened marks near the ends. Reproduce it with a creep test that holds a constant load at the real service temperature and logs deformation (ASTM International, 2024a). Where peel loads exist, do not judge by shear data alone.
In design, the priority is to turn peel and cleavage into shear or compression. Set overlap length and adherend thickness so the joint does not bend (Goland & Reissner, 1944; da Silva et al., 2018). Under sustained load, choose an adhesive whose glass transition temperature sits well above the service temperature, and add a mechanical fastener if needed. Do not apply a datasheet's short-term room-temperature strength to a load that lasts.
Hypothesis 5: incomplete cure
The last hypothesis is cure. It arises where two-part reactive adhesives are metered and mixed by hand, in workshops with weak heating in winter, and on lines that pass parts on before the full cure time, and it shows as failures clustered in lots made on certain days or shifts, just after a mixing nozzle change, or in winter. Consider a lot bonded with hand-mixed two-part epoxy on a Monday morning after a weekend with the heating off. It passed the room-temperature test at shipment, yet that lot alone comes back failed after the summer, and the adhesive on the fracture keeps a dent when pressed with a fingernail.
Reactive adhesives get their strength from crosslinking, a reaction that ties the chains together, driven by two components, heat or moisture. With a wrong mix ratio, uneven mixing or a cold workshop that slows the reaction, crosslinking stays incomplete. A study of an epoxy adhesive shows that a lower degree of cure brings a lower glass transition temperature and lower strength (Lapique & Redford, 2002).
Under-cured adhesive can still pass a room-temperature test at shipment. But with a low glass transition temperature it softens easily in summer heat or near hot equipment, and unreacted components draw in more water, speeding plasticisation and creep (Kinloch, 1987; Lapique & Redford, 2002). So incomplete cure can be less a cause on its own than a weakness that makes failures from water, heat and sustained stress come sooner.
Pressing the adhesive on the fracture with a fingernail is the first check. If it is tacky, dents easily, or shows streaks and uneven colour, mixing or cure fell short. For a number, compare the glass transition temperature of failed and sound parts by differential scanning calorimetry, or DSC, which measures heat flow (ASTM International, 2023b); a clearly lower value means incomplete cure. Absorbed water also lowers the glass transition temperature, so dry samples from humid service before measuring. Match the dates, shifts and workshop temperatures of the failed lots in the production records as well.
On the cure side, the remedy is to pin down metering and temperature in the records. Check mix ratio and mixing quality every shift, and hold workshop and part temperature and cure time within the manufacturer's recommended range. In winter, consider preheating parts or a heated post-cure. A cure-check coupon made with each lot lets any later failure be traced to cure, or ruled out, at once.
Testing the hypotheses
Five hypotheses do not call for five separate investigations. Read the clues the failed part left behind, then rule them out one at a time, starting with what is easy to measure.
The first step is to sort the fracture into interfacial, cohesive, substrate or mixed failure and photograph it (ISO, 2022). If it is interfacial, start with surface preparation, hypothesis 1: check the treatment records, the water-break or wetting tension results, and the wait from treatment to bonding (ASTM International, 2021, 2023a). This one classification already splits the interface hypotheses from the adhesive ones.
Next come the shape and timing of the spread. If it crept in from the edges, with discolouration or corrosion product, and coincides with humid periods, it is water, hypothesis 2, and the check is whether strength returns after drying (Mubashar et al., 2009). If it started at the ends or corners, between unlike materials or where the temperature range is wide, it points to thermal mismatch, hypothesis 3, to be reproduced by thermal cycling (Bagiatis et al., 2021; da Silva & Adams, 2007).
If the failure is cohesive, look at the adhesive. Measuring the glass transition temperature against sound parts settles incomplete cure, hypothesis 5 (ASTM International, 2023b), and if the design carries sustained or peel loads, confirm hypothesis 4 with a creep test (ASTM International, 2024a). If the glass transition temperature is normal and the load was small, go back to the chance that water softened the adhesive, hypothesis 2.
If that still does not settle it, make control coupons. Compare coupons with only the suspected cause changed against coupons made the current way, side by side, using the wedge test or strength after humid-heat ageing (ASTM International, 2024c). Change one thing at a time, or it will be unclear what worked. Once the cause is found and fixed, the last step is confirmation. It is done only when coupons that have been through humid-heat ageing or thermal cycling reach the expected strength and break within the adhesive or the adherend rather than at the interface (ISO, 2022).
One more thing to remember: the hypotheses overlap. A poorly prepared joint gives way to water first, and under-cured adhesive brings creep and plasticisation forward. In such cases, fix what sits earliest in the chain, surface preparation or cure, first, and the other causes weigh less.
Limits and cautions
Numerical limits differ by adhesive and substrate, so follow the manufacturer's technical data and the relevant standards. For structural members, vehicle and aircraft parts, pressure equipment or any joint tied to human safety, do not decide on floor judgement alone; involve the design authority and the adhesive manufacturer's technical support.
Safety matters too. Many solvents, acids and primers used in surface preparation, and many reactive adhesives, carry inhalation, skin sensitisation or fire hazards, and polyurethanes containing isocyanates can cause respiratory sensitisation. Check the product's safety data sheet, the SDS, before work, and provide ventilation and protective equipment. Where the cause stays unclear or a recall is at stake, sending fracture surface and thermal analysis to a specialist laboratory is faster.
Sources
- ASTM International. (2021). ASTM F22-21, Standard test method for hydrophobic surface films by the water-break test. https://store.astm.org/f0022-21.html
- ASTM International. (2022). ASTM C719-22, Standard test method for adhesion and cohesion of elastomeric joint sealants under cyclic movement (Hockman cycle). https://webstore.ansi.org/standards/astm/astmc71922
- ASTM International. (2023a). ASTM D2578-23, Standard test method for wetting tension of polyethylene and polypropylene films. https://store.astm.org/d2578-23.html
- ASTM International. (2023b). ASTM E1356-23, Standard test method for assignment of the glass transition temperatures by differential scanning calorimetry. https://store.astm.org/e1356-23.html
- ASTM International. (2024a). ASTM D2294-96(2024), Standard test method for creep properties of adhesives in shear by tension loading (metal-to-metal). https://store.astm.org/d2294-96r24.html
- ASTM International. (2024b). ASTM D2651-01(2024), Standard guide for preparation of metal surfaces for adhesive bonding. https://store.astm.org/d2651-01r24.html
- ASTM International. (2024c). ASTM D3762-24, Standard test method for adhesive-bonded surface durability of aluminum (wedge test). https://store.astm.org/d3762-24.html
- Bagiatis, V., et al. (2021). Thermally induced strain in joints with dissimilar adherends bonded with a flexible adhesive. International Journal of Adhesion and Adhesives, 107, 102853. https://doi.org/10.1016/j.ijadhadh.2021.102853
- Bikerman, J. J. (1968). The science of adhesive joints (2nd ed.). Academic Press. https://shop.elsevier.com/books/the-science-of-adhesive-joints/bikerman/978-1-4832-3309-3
- da Silva, L. F. M., & Adams, R. D. (2007). Joint strength predictions for adhesive joints to be used over a wide temperature range. International Journal of Adhesion and Adhesives, 27. https://doi.org/10.1016/j.ijadhadh.2006.09.007
- da Silva, L. F. M., Öchsner, A., & Adams, R. D. (Eds.). (2018). Handbook of adhesion technology (2nd ed.). Springer. https://doi.org/10.1007/978-3-319-55411-2
- Dubelley, F., et al. (2018). Durability of polymer metal multilayer: Focus on the adhesive chemical degradation. Frontiers in Chemistry, 6. https://doi.org/10.3389/fchem.2018.00459
- Ebnesajjad, S. (2014). Surface treatment of materials for adhesive bonding (2nd ed.). William Andrew. https://shop.elsevier.com/books/surface-treatment-of-materials-for-adhesive-bonding/ebnesajjad/978-0-323-26435-8
- Goland, M., & Reissner, E. (1944). The stresses in cemented joints. Journal of Applied Mechanics, 11, A17–A27. https://doi.org/10.1115/1.4009336
- International Organization for Standardization [ISO]. (2022). ISO 10365:2022, Adhesives — Designation of main failure patterns. https://www.iso.org/standard/80800.html
- Kinloch, A. J. (1987). Adhesion and adhesives: Science and technology. Chapman and Hall. https://doi.org/10.1007/978-94-015-7764-9
- Lapique, F., & Redford, K. (2002). Curing effects on viscosity and mechanical properties of a commercial epoxy resin adhesive. International Journal of Adhesion and Adhesives, 22, 337–346. https://doi.org/10.1016/S0143-7496(02)00013-1
- Mubashar, A., Ashcroft, I. A., Critchlow, G. W., & Crocombe, A. D. (2009). Moisture absorption–desorption effects in adhesive joints. International Journal of Adhesion and Adhesives, 29(8), 751–760. https://doi.org/10.1016/j.ijadhadh.2009.05.001
- Venables, J. D. (1984). Adhesion and durability of metal-polymer bonds. Journal of Materials Science, 19, 2431–2453. https://doi.org/10.1007/BF00550796
- Zanni-Deffarges, M. P., & Shanahan, M. E. R. (1995). Diffusion of water into an epoxy adhesive: Comparison between bulk behaviour and adhesive joints. International Journal of Adhesion and Adhesives, 15(3), 137–142. https://doi.org/10.1016/0143-7496(95)91624-F