Adhesion, protection and joint durability

Initial attachment is only a starting condition

A coating still looks attached after a wet exposure, but its pull-off load has fallen. Has water crossed the coating, softened its adhesive, weakened a buried interface, or changed the substrate? The same reduced load can have several causes. A useful durability comparison follows the route into the assembly, the state changed there and the response measured afterwards.

An interface can degrade under moisture, heat, reaction or repeated loading. Surface contamination and preparation affect initial bonding, while transport along edges and defects can change later performance.

The adherends are the bodies joined by an adhesive. Water can enter through a permeable polymer, along exposed edges or through connected defects. Sorption means incorporation of water into the material; transport determines how its distribution changes. A thick section can have a wet outer region while its interior is still changing. The average water content of a whole joint does not locate water at a buried interface.

Absorbed water can plasticise an adhesive: it changes molecular interactions and segment mobility, making the material softer in some conditions. Swelling is a dimensional change. If the adhesive cannot expand as it would freely, restraint can generate stress. These physical changes need not require cleavage of the polymer backbone, and some response may return when water leaves.

Hydrolysis is different: a chemical reaction with water breaks susceptible covalent bonds. Merely forming hydrogen bonds with water is not hydrolysis. Water at an interface can also change the interactions that attach the materials, while hydration or corrosion can alter the substrate surface. Drying need not reconstruct broken chains, remove corrosion products or reconnect an existing separation. Which processes occur depends on chemistry, temperature, load and history; they are alternatives to investigate, not a compulsory sequence for every adhesive.

A useful causal sequence is therefore exposure → entry path → local water and material state → changed force transfer or crack resistance → observed test result. Separate evidence is needed at the links. A wet mass increase alone does not identify broken bonds, and an apparently clean fracture surface alone does not prove that the original interface was free of a thin polymer layer.

A better film can leave an edge route almost unchanged

Consider a separate, fictional protective-film experiment at fixed temperature. A flat homogeneous film has intact area A = 0.0100 m², thickness d = 120 µm = 1.20 × 10⁻⁴ m and constant water diffusivity D = 1.00 × 10⁻¹² m²/s. The water concentrations at its two faces, expressed per volume of film, are maintained at c_w = 1.00 and c_d = 0 kg/m³. These are supplied concentrations inside the film, not ambient relative humidities.

Steady state has already been established, with no reaction, bulk flow, swelling or evolving defects. Let x increase from the wet face to the dry face. Conservation makes the one-dimensional flux constant; constant D then gives a straight concentration profile. Fick’s law J = −D dc/dx gives J = D(c_w − c_d)/d. Here J is water mass crossing unit intact area per time, in kg/(m²·s). Its positive sign means transfer toward the dry face. Diffusion and phase transformations develops the transport law and its limits.

Multiplying by area gives the intact-film delivery rate q_bulk = AJ. The supplied values give J = 8.333… × 10⁻⁹ kg/(m²·s) and q_bulk = 8.333… × 10⁻¹¹ kg/s = 0.300 mg/h. A separate edge bypass delivers q_edge = 0.700 mg/h under the same maintained reservoirs. That edge rate is supplied independently; it is not divided by the intact-film area to invent an edge diffusivity.

The two paths feed one receiver that continuously removes the arriving water, maintaining its dry boundary. Their total rate is q = q_bulk + q_edge = 1.000 mg/h. Doubling only d halves the intact-film rate to 0.150 mg/h but leaves the edge rate unchanged: the total becomes 0.850 mg/h, a 15% reduction. Reducing only the supplied edge rate by a factor of ten instead gives 0.370 mg/h, a 63% reduction. Doing both gives 0.220 mg/h.

These are alternative supplied configurations with unchanged D, face concentrations, intact area and the stated independent edge rate. They are not predictions of what a named coating or sealant will do. During six hours after steady state, the collected masses are 6.00, 5.10, 2.22 and 1.32 mg. Collection in a maintained receiver is not the water stored in a finite joint. No adhesive strength, time to debonding or corrosion life is specified by this transport account.

A real thickness change can alter curing, defects, stress and the time required to approach steady state. Edge treatment can alter the boundary conditions of the intact film. Those changes must be measured or modelled before applying the simple comparisons. Protection requires considering all relevant routes, including the route left unchanged by the proposed improvement.

Use a relevant failure assessment

A pull-off, peel or shear result has a specific geometry and failure path. Identify whether separation occurred within the adhesive, at an interface or in the substrate. Ageing protocols should represent relevant mechanisms before being used for service predictions.

In pull-off loading, a nominal force acts away from a bonded surface; force divided by a declared bonded area gives an apparent tensile stress. Peel bends a flexible adherend and concentrates opening near an advancing front; force per width is often the reported output. Lap shear uses an overlap that can also bend and develop opening stresses. Equal numerical outputs from different geometries do not imply equal intrinsic adhesion. Joining mechanisms and interfacial bonding explains those force paths.

Use the fictional records below to compare one fixed pull-off configuration. Every coupon has the same nominal 10 mm × 10 mm bonded area, fixture, alignment, loading rate and test temperature. Each entry is one exact illustrative record from a separate destructively tested coupon, not a mean or a population estimate. Dry controls are retained for the same elapsed time as their corresponding exposed sets. Initially and after this dry control history, A carries 1000 N and B carries 400 N.

Define retained load R = F_exposed/F_dry-control for each material configuration and matched elapsed time. For A after the wet exposure, R_A = 600/1000 = 0.60. For B, R_B = 360/400 = 0.90. B retains a larger fraction, but A carries the larger wet load: 600 rather than 360 N. Because the areas are 100 mm², their wet apparent stresses are 6.00 and 3.60 MPa; 1 MPa = 1 N/mm². A ranking by retained fraction answers a different question from a ranking by absolute exposed load.

That distinction has an experimental counterpart. A 2006 study of PMMA coatings on differently prepared surfaces found that an octyl-treated surface lost little adhesion upon moisture exposure yet had poor dry and wet attachment. The measured output was a shaft-loaded blister fracture-energy quantity, not this lesson’s pull-off load. The result illustrates why fractional loss and useful exposed performance must be considered separately; it does not prescribe a treatment or transfer a numerical property to another joint.

A separate A coupon exposed wet and then reconditioned to a verified dry state gives 850 N. Its ratio to the matching dry control is 0.85, a rise of 25 percentage points above the wet ratio 0.60. This observation is consistent with a reversible contribution, but it does not show that 85% of molecular bonds were repaired or identify the remaining damage. The original 600 N coupon was destroyed; it was not dried and retested intact.

Cohesive failure lies within the adhesive, interfacial failure appears near an adhesive–adherend boundary, and substrate failure lies within the adherend. Mixed locations can occur. A change from predominantly cohesive to predominantly interfacial separation helps locate a changed limiting region, but cannot alone distinguish water displacement, chemical change, pre-existing defects or altered stress. Failure inside the substrate shows that the connection transmitted the force reached in that configuration; it leaves the interface’s own separation threshold unmeasured.

A net mass gain can conceal solid material lost

Mass measurements can support an ingress assessment, but the numerator must describe what is being counted. For a separate fictional unbonded polymer coupon, let m₀ = 2.000 g be its original dry mass. After exposure its measured wet mass is m_w = 2.030 g. Independent assessment establishes that its residual non-water solid mass at that wet endpoint is m_s = 1.990 g. For this case, subsequent verification removes all water without any further loss or gain of non-water material; that is a supplied validity condition, not a drying instruction.

The original solid has lost m_lost = m₀ − m_s = 0.010 g. Water present at the wet endpoint is m_water = m_w − m_s = 0.040 g. The balance m_w = m₀ − m_lost + m_water gives 2.000 − 0.010 + 0.040 = 2.030 g. Calling the net gain 0.030 g “all the absorbed water” would miss the simultaneous solid loss.

The net gain is 1.50% of original dry mass, whereas the water amount is 2.00% of original dry mass or about 2.01% of residual dry solid mass. The different percentages have different denominators. If drying also removes additives or creates new chemical changes, the final dry mass is not automatically the residual solid mass that was present at the wet endpoint. With only m₀ and m_w, two unknown contributions cannot be separated.

In a joint, substrate uptake or corrosion products can further change the total mass. Even a valid water amount does not locate the water, establish its chemical effect or determine interface fracture resistance. Combine mass balance with the appropriate local-state and mechanical observations rather than treating mass gain as a universal damage meter.

Check the clock at more than one response level

A hotter exposure can shorten a test, but it can also change uptake, chemical reactions, polymer state and failure location. Continuous wet exposure can differ from wet–dry cycles. To infer a service response, the relationship between the accelerated and intended conditions must itself be supported over the intended geometry, loading, environmental history and response range.

One testable hypothesis is a single acceleration factor a: the intended-condition curve has the same retained-load shape as the hotter curve, with every time multiplied by a. At a specified retained fraction r, this requires t_S(r) = a t_H(r), where H denotes the hotter condition and S the intended condition. This is a hypothesis about complete matched responses, not a general law of ageing.

For a separate fictional comparison, stipulate exact level-crossing times obtained from response records with the same retention definition, appropriately time-matched dry controls, geometry and mechanical test. At r = 0.80, H reaches the level in 20 h and S in 100 h, giving a = 5. At r = 0.60, the times are 40 and 120 h, giving a = 3. No one constant factor satisfies both pairs. Using the first pair alone would predict 200 h at the second level instead of the stipulated 120 h.

The contradiction rejects the single-factor hypothesis for these supplied exact records. It does not identify which mechanism changed, and does not prove that a different variable-factor model is valid. Real records also require uncertainty in response and level-crossing time. Similar fracture locations or an apparently straight fitted line are useful evidence, but neither by itself establishes identical chemistry, transport or a valid long extrapolation.

Residual load after exposure and time to failure under sustained load are also different endpoints. A coupon that carries 600 N in a short destructive test has not thereby demonstrated survival at that load for a year. Lifetime evidence, scatter and extrapolation develops the additional sampling and model requirements.

Choose evidence that can separate the explanations

Start with the question the change is meant to answer. A barrier comparison needs the geometry, exposed routes, boundary conditions and transported amount. An attachment comparison needs the loading mode, area or width basis, test rate and temperature, conditioning state and failure location. A service inference additionally needs relevant histories, variability and a validated relation between test and use.

For the two-path model, separately measuring a receiver with and without the edge route can test whether a lower film flux leaves the bypass dominant. For the matched pull-off records, time-matched dry controls distinguish exposure from changes that occur during storage; wet and independently reconditioned sets probe state dependence. Examination of both mating surfaces, with suitable microscopy and chemical analysis, can distinguish a thin retained adhesive layer from a genuinely different fracture path. No single one of those observations uniquely identifies every mechanism.

Keep the conclusions attached to their observations. A coating may slow delivery yet detach because attachment or mismatch remains limiting. A joint may retain most of a poor initial load. A substrate may fail before the interface is tested to separation. Improved performance under one exposure is useful evidence within that comparison; durability of the complete material combination still depends on the intended conditions.

Worked examples

Try the reasoning

1. Why locate the fracture path?

   Show the answer and reasoning

   Different regions can limit the test.

   A number alone may not measure the intended interface.

2. Can accelerated ageing change the mechanism?

   Show the answer and reasoning

   Yes.

   Temperature or chemistry can introduce a different failure process.

3. In a new configuration of the steady film model, double the intact area but keep d, D, face concentrations and the separately measured 0.700 mg/h edge rate fixed. What is the new six-hour collected mass? Why is this edge condition essential?

   Show the answer and reasoning

   7.80 mg: intact-film rate 0.600 plus edge rate 0.700 gives 1.300 mg/h.

   Intact flux per area is unchanged, so total intact delivery doubles. A real change of specimen size can change perimeter and its leak rate; holding the edge rate fixed is an explicit new supplied condition, not a geometric law.

4. Keep the 0.700 mg/h bypass unchanged and use the same six-hour screen of at most 3.00 mg. Could any finite increase in intact-film thickness alone meet it under the steady model?

   Show the answer and reasoning

   No. The bypass alone collects 4.20 mg in six hours.

   For positive finite d, q_bulk is positive and approaches zero as d increases without bound. Even that limiting total is 0.700 mg/h, above the allowed 0.500 mg/h. An unchanged parallel route sets a floor.

5. A separately tested pull-off coupon with the same 100 mm² area fails entirely inside its substrate at 900 N. Can its intrinsic interface strength be ranked above or below A’s 600 N wet record?

   Show the answer and reasoning

   No intrinsic interface ranking follows. The tested connection transmitted 900 N before the substrate failed.

   The interface did not separate in that record, so its own threshold was not measured. The higher assembly force and nominal 9.00 MPa are valid observations for that configuration, not an intrinsic interface-fracture value.

6. Another coupon starts at 2.000 g and also weighs 2.030 g wet. Its independently established residual solid mass is 2.000 g. Compare its water amount with the 1.990 g residual-solid case.

   Show the answer and reasoning

   The new coupon contains 0.030 g of water, compared with 0.040 g in the original case.

   Equal net gains can conceal different water amounts when solid loss differs. The original lost 0.010 g; the new case stipulates none. Neither total amount specifies a buried-interface concentration.

7. Change only the second intended-condition time in the clock table from 120 to 200 h. Would the two endpoint pairs now establish a = 5 over the whole ageing process?

   Show the answer and reasoning

   They would both be consistent with a = 5, but would not establish it between or beyond those endpoints.

   The second ratio would be 200/40 = 5. Agreement at two levels is a limited test, not proof of a complete common curve, unchanged mechanisms or an extrapolation to a safe service life.

8. A new wet-aged coupon carries 600 N. Its time-matched dry control carries 800 N, although its original initial record was 1000 N. What is retained load on the defined matched-control basis, and why also keep the initial record?

   Show the answer and reasoning

   R = 600/800 = 0.75. The wet load is also 0.60 of the initial 1000 N.

   The two ratios answer different comparisons. The dry control’s change shows that elapsed storage/history matters; replacing it by the initial value would confound that change with the wet exposure. No unique cause follows from these three values alone.

What to carry forward

• Measure durability of the interface and surroundings.
• Retain failure location and exposure history.
• A lower intact-film flux can leave a parallel edge route dominant.
• Retained fraction, absolute exposed load and fracture location answer different questions.
• Separate water gained from non-water mass lost, with a declared denominator and valid state measurement.
• An accelerated response needs a tested relation to intended conditions; one fitted endpoint is not a lifetime model.