Materials Science · Joining, coatings and surface engineering

Joining mechanisms and interfacial bonding

Welding, brazing, soldering, adhesive bonding and mechanical fastening connect materials through different mechanisms. Fusion welding locally melts material; brazing and soldering use a molten filler while the base materials remain solid. Solid-state joining provides other routes.

Useful background: Processing routes and microstructural evolution; Composites, fibres and interfaces; Cracks, fracture and toughness. Each worked case explains the quantities it uses.

Back to the chapter

After reading, you can

Compare joints through bonding mechanism and force path.
Explain why contact, wetting, setting and the resulting interface state matter in different joining routes.
Distinguish nominal shear, local traction and peel-test output using consistent force and area bases.
Use supplied local material states and failure-location evidence to assess what a whole-joint comparison establishes.

A joint creates a new region

Two pieces remain attached, but what holds them together, and which material actually carries the applied force? An adhesive can pass a lap-shear comparison yet peel from an edge. A fusion weld can remain intact while an adjacent softened strip deforms. Both observations are possible because a joint is a newly created material region and force path, rather than a label applied to the original bulk materials.

In fusion welding, a local liquid region subsequently solidifies. It may include filler as well as melted parent material; the weld composition, solidification structure and defects need not reproduce either starting piece. The surrounding parent material can remain solid yet undergo a different thermal history. Joining without an added filler does not mean that the thermal cycle left the original state unchanged.

In brazing and soldering, liquid filler must contact and wet the relevant solid surfaces before it freezes. Wetting means that the liquid spreads over or occupies the contacting surface under the actual interfacial conditions. Capillary flow can help a suitable filler enter a narrow joint, but a liquid filler nearby is not proof of complete filling or bonding. Oxides, contamination, incomplete contact and interfacial reactions can change the result. The process names alone do not establish a useful joint.

A solid-state route instead brings solid surfaces into sufficiently intimate contact. In diffusion bonding, microscopic high points initially carry much of the contact load; deformation and time-dependent flow can enlarge contact, while atomic transport can help shrink remaining voids and develop the interface. A macroscopic gap is not automatically bridged by atoms merely because the pieces are hot. Stable surface films can obstruct metal-to-metal contact. Contact development therefore depends on the material pair, surface state and joining history.

Compare what creates the connection before comparing a mechanical test. The actual route and material pair determine the resulting region.
Route	Connection-forming event	State that still needs evidence
Fusion welding	Local melting followed by solidification	Weld structure, flaws and adjacent thermal changes
Brazing or soldering	Molten filler wets solid parent surfaces and then solidifies	Filling, interfacial reaction and filler state
Solid-state diffusion bonding	Contact development and thermally activated interfacial evolution	Unbonded regions, surface films and local material state
Adhesive bonding	Surface contact followed by a load-carrying adhesive state	Interface attachment, adhesive cohesion and conditioning
Mechanical fastening	Contact, bearing and possibly clamp-induced friction	Fastener, hole region, clamping and force path

Contact is necessary, but setting and load transfer are separate tasks

The pieces joined by an adhesive are called adherends. Adhesion concerns their attachment across the adhesive–adherend interfaces; cohesion concerns resistance to separation within the adhesive. Molecular attraction, suitable chemical interactions and mechanical interlocking can contribute in different systems. There is no one universal molecular bond for every adhesive and surface pair.

A liquid or deformable adhesive must contact the intended surface rather than only a weak contaminant layer. It must also develop a useful internal state. A thermosetting adhesive can cure by chemical reaction and network formation; a hot-melt adhesive can set by cooling; a solvent-based adhesive can set as solvent leaves. These changes are different from establishing the interface itself. A hardened adhesive can still detach from a poorly bonded surface, while a well-attached interface can remain connected as the adhesive tears internally.

A bolted or riveted connection uses another path. Force can pass through bearing contact between a fastener and hole, or through friction when clamping maintains contact, depending on the state and geometry. The fastener can also carry shear or tension. A hole changes the available section and introduces local stresses. Mechanical attachment therefore does not require that the two original surfaces have become one chemically bonded phase.

Choosing a route means matching its connection-forming mechanism and side effects to the material combination and task. An adhesive route may avoid local fusion but can introduce time-dependent or temperature-sensitive response. A thermal route may develop a metallic connection while altering nearby material. A removable fastener can introduce mass, holes and concentrated contact. None is universally strongest without a defined geometry, endpoint and environment.

Follow the force through the geometry

Shear traction acts parallel to a chosen interface; normal traction acts perpendicular to it. Traction is force per area on that local plane. Pulling along an overlapped joint often transfers much of the force through adhesive shear. But the external tensile direction does not establish a state of pure shear throughout the adhesive.

In a single-lap arrangement the two adherend force paths are offset. Their eccentricity can cause bending and normal opening near the overlap ends. Adherend stiffness, adhesive thickness, restraints and deformation affect the local distribution. In an ideal symmetric double-lap arrangement with two identical interfaces, symmetry can split the transferred force equally between them. That removes one geometric asymmetry; it does not make their end stresses uniform or eliminate every normal stress.

Peeling bends a relatively flexible adherend away from the other piece and concentrates opening near an advancing separation front. The normal opening and bending work differ from the loading in a lap-shear test. A common peel output is force divided by strip width, p = P/b, in N/mm. Here P is the measured force and b the tested width. It is not the same definition as P divided by bonded area in N/mm², and neither is automatically the intrinsic work needed to break molecular bonds.

For example, a stipulated peel test gives P = 200 N at b = 20 mm, so p = 10 N/mm under its stated angle, rate, strip and conditioning. A different test giving 300 N at 30 mm has the same force-per-width number. That equality alone does not establish equal interface fracture resistance: adherend bending, adhesive deformation, failure location and test conditions can differ. The peel-versus-shear worked case below asks precisely why one coupon result cannot qualify the other loading.

Schematic, not a stress solution or a scale drawing. In the symmetric double-lap arrangement, a centre adherend is pulled right by total force P. Identical upper and lower outer adherends are each pulled left by P/2. The two adhesive layers are hatched. The symmetric external forces have zero resultant and zero resultant moment. Equal force sharing is stipulated by symmetry; neither uniform shear nor zero normal stress is implied. This two-interface diagram is separate from the following one-interface 400 square millimetre force account. Below, a flexible adherend bends upwards from a held substrate. The separation front is marked; the fixture provides the reactions to the applied peel force. The peel output P/b is force per out-of-plane strip width, not force per bonded area or an intrinsic fracture energy.

The average does not locate the largest traction

Take a separate simple force-transfer account with one planar bonded interface of width b = 20 mm and overlap length L = 20 mm. The bonded area is A = bL = 400 mm². Transfer P = 1,600 N parallel to that plane. The nominal mean shear traction is τ_mean = P/A = 4.00 N/mm² = 4.00 MPa. This is an area average required by force balance, not a prediction of a uniform local stress.

The next figure compares two invented distributions. U is uniform at 4 MPa. E has 6 MPa over each end quarter of the length and 2 MPa over the middle half. The end areas are each 100 mm² and the middle area 200 mm², so their transferred forces are 600, 400 and 600 N. Both distributions carry 1,600 N. Their maximum tractions are nevertheless 4 and 6 MPa.

These are deliberately supplied force-balanced profiles, not solutions for the same actual elastic joint or measured traction maps. Equilibrium alone cannot decide which distribution a material combination develops. That requires compatible deformation, constitutive behaviour, geometry and boundary conditions. Sharp steps are bookkeeping idealisations; this account neither resolves an end singularity nor calculates peel stress.

To see why a mean-only screen can mislead, impose an invented rule that local shear must be at most 5 MPa. U passes at the stated load, whereas E does not. If E keeps exactly the same profile shape while all its tractions scale linearly with P, the equality load is 1,600 × 5/6 ≈ 1,333 N. A rule incorrectly applied only to the mean would instead give 5 × 400 = 2,000 N. The difference exposes the assumptions; neither number is a real adhesive allowable or a component approval.

Increasing the overlap length lowers P/(bL) at fixed force and width. It need not lower the actual end traction in the same proportion, because the load-transfer distance and bending can change. Load transfer through fibres and interfaces develops another geometry with spatial transfer. A nominal area-normalised result should remain attached to the joint and test that produced it.

Invented one-interface force account, constant across width. The supplied local profiles are alternatives satisfying resultant force balance, not an elasticity solution or material dataset.
Profile/region	Area (mm²)	Shear traction (MPa)	Transferred force (N)
U: whole overlap	400	4	1,600
E: first end quarter	100	6	600
E: middle half	200	2	400
E: second end quarter	100	6	600

Invented force account for one interface, width 20 millimetres and overlap length 20 millimetres. Position along the overlap is horizontal; local shear traction in megapascals is vertical. Dashed U is 4 megapascals over all 20 millimetres. Solid stepped E is 6 over 0 to 5 millimetres, 2 over 5 to 15, and 6 over 15 to 20. A dotted horizontal line marks the invented inclusive local rule of 5 megapascals. Each profile has area 80 megapascals times millimetres under its curve; multiplying by the width gives 1600 newtons. Their maxima are 4 and 6 megapascals. The step transitions do not specify a local end or peel solution; these supplied equilibrium alternatives are not predictions for the same elastic joint. At the same load U passes the toy rule and E does not.

The surrounding state matters

Assess interface chemistry, thermal history, geometry, residual stress and load transfer. A joint can fail in the bond, surrounding material or fastener. Its strength is not simply the bulk strength of either starting material.

Consider fusion joining a previously cold-worked metal. The fusion region melts and resolidifies. Outside it, a heat-affected zone, abbreviated HAZ, remains unmelted but experiences a local heating-and-cooling path. Recovery and recrystallisation can remove part of the dislocation state responsible for the original work hardening, leaving a softer neighbouring region. In another alloy, phase transformations or changes in precipitates can instead produce a different response. HAZ is a location affected by heat, not the name of one universal phase.

Use a separate invented one-dimensional axial screen to isolate that consequence. Three representative regions lie in series along a load path. Each carries the same tensile force P, with no bypass. Their areas and locally characterised 0.2% proof stresses are given in the table. Proof stress is the engineering stress defined by a 0.002 strain-offset construction; 1 MPa = 1 N/mm². The model excludes bending, stress concentrations, residual stress, cracks and load redistribution. It asks only whether P/A exceeds a supplied local proof-stress value.

At P = 3,000 N, the base and HAZ sections both carry 150 MPa; the smaller weld section carries 187.5 MPa. The base and weld pass their local screens, but the HAZ exceeds 120 MPa. Multiplying each area by its proof stress gives equality forces of 4,800, 2,400 and 4,800 N. The smallest, 2,400 N, governs this idealised first-threshold comparison. It is not the ultimate joint breaking load.

Raising only the weld-region proof stress from 300 to 350 MPa raises that section’s equality force to 5,600 N but leaves the governing 2,400 N unchanged. Raising only the HAZ value from 120 to 180 MPa instead raises the model minimum to 3,600 N. These changed values are separately supplied state data, not results predicted from a named treatment. A better weld-metal number can leave the whole-path screen unchanged.

Real heating also creates incompatible expansion, contraction and structural changes that can leave residual stress after external loads are removed. The simple zero-residual-stress axial screen does not estimate those stresses or their consequences. Actual section shape, defects, thermal history and local properties must be examined before transferring a coupon or section comparison to a finished joint.

Invented final-state data, tested on a common temperature/rate/proof-stress basis. Areas are the stated load-bearing cross-sections. Three regions are represented in a serial axial model, not a full weld stress field or material datasheet.
Representative region	Area A (mm²)	Local 0.2% proof stress (MPa)	P = A × proof stress (N)
Unaffected base	20.0	240	4,800
Heat-affected region	20.0	120	2,400
Weld region	16.0	300	4,800

Invented serial axial proof-stress screen. The upper schematic represents three sections on one force path, with equal and opposite applied tensile forces P. Each section carries the same P. These are representative sections rather than a literal weld layout, and the drawn lengths do not represent material volumes. Baseline equality forces A times local 0.2 percent proof stress are 4800 newtons for base, 2400 for the heat-affected region and 4800 for weld. In the lower bar plot these appear at 4.8, 2.4 and 4.8 kilonewtons, with different endpoint shapes and hatching. A vertical dashed line marks an applied 3.0 kilonewton force, exceeding only the heat-affected threshold. The minimum is 2.4 kilonewtons. Raising only weld proof stress leaves that minimum unchanged; a separately supplied heat-affected improvement can raise it. The zero-residual-stress model excludes bending, cracks, local stress concentration and redistribution; these equality forces are not breaking loads or component ratings.

Keep the failure location with the measured load

A separation near an adhesive–adherend interface, rupture inside the adhesive and tearing of the adherend are different observations. Examine both mating fracture surfaces with appropriate resolution; a very thin retained layer can make an apparently bare interface misleading. A single joint can show mixed locations. The largest measured force by itself does not establish which molecular bond was strongest.

Failure in an adherend shows that the connection transmitted the force reached in that tested configuration. It does not measure the unused capacity of the adhesive, prove that every interface is flawless, or establish a peel result from a tensile-shear test. A fracture surface and a load–displacement record supply complementary evidence. Fracture resistance also differs from the proof-stress screen above, which concerns a stated permanent-deformation endpoint.

Contact formation deserves its own evidence. In a copper diffusion-bonding study, changing the surface asperity geometry changed the observed bonded area and joint response. The comparison illustrates why nominal contact area or one roughness number cannot fully specify the developed interface. A bond-area measurement, a strength result and a mechanism model remain different kinds of evidence; no universal numerical conversion between them is supplied here.

Initial dry attachment does not establish persistence in another environment. Water can alter the adhesive, the interface or the adherend, with the controlling route depending on the system. Adhesion, protection and joint durability develops exposure and time dependence. For the present comparison, retain material pair and surface state, actual joining/setting history, bond geometry, load direction, rate, temperature, conditioning and failure location. Use a changed comparison that could distinguish the proposed weak region rather than ranking process names alone.

Worked examples

Work through the case

An adhesive joint fails by peeling although its shear coupon performs well. Is that inconsistent?

Peel and shear create different interfacial loading.
Joint geometry changes stress concentration.
Use a test relevant to the actual force path.

No. A shear result does not establish peel performance.

Two identical resultants do not give the same local screen

For the supplied 20 mm by 20 mm one-interface account, verify both profiles carry 1,600 N. Apply the 5 MPa local rule, then find E’s equality load under the additional fixed-shape linear-scaling assumption.

The area is 400 mm². U transfers (4 MPa)(400 mm²) = 1,600 N, because MPa × mm² = N.
E transfers (6)(100) + (2)(200) + (6)(100) = 600 + 400 + 600 = 1,600 N. Its average is still 4 MPa, but its maximum is 6 MPa.
U meets 4 ≤ 5; E fails 6 ≤ 5 at this load. For E’s same shape, multiply every traction and P by 5/6. The equality force is 1,333.33… N; its mean is then 3.333… MPa and its middle traction 1.666… MPa.
Using only the mean would permit 2,000 N, 50% above 1,333.33… N. The stipulated local profile and scaling are necessary assumptions. Neither equilibrium nor this illustrative stress rule establishes a real failure load.

Both carry 1,600 N. Only U meets the supplied local rule at that force. E’s conditional equality load is about 1,333 N, not the mean-only 2,000 N.

An improvement can leave the governing region unchanged

Apply the serial axial screen to the three-region table at 3,000 N. Compare changing only weld proof stress to 350 MPa with changing only HAZ proof stress to 180 MPa.

Use P/A: base and HAZ stresses are 3,000/20 = 150 MPa; weld stress is 3,000/16 = 187.5 MPa. Only the HAZ exceeds its given local proof stress.
Baseline equality forces are 20 × 240 = 4,800 N, 20 × 120 = 2,400 N and 16 × 300 = 4,800 N. Their minimum is 2,400 N.
With only the weld changed, its equality force becomes 16 × 350 = 5,600 N. The HAZ still sets 2,400 N, so the whole-path screen is unchanged.
With only the HAZ changed, its equality force is 20 × 180 = 3,600 N. The other two remain 4,800 N; the new model minimum is 3,600 N, a 50% rise. No treatment prediction, crack assessment or ultimate breaking load follows.

The initial 3,000 N fails the HAZ screen. Improving weld proof stress alone leaves the minimum at 2,400 N; the supplied HAZ improvement raises it to 3,600 N.

Equal force per width still needs a matched experiment

Peel record A gives 200 N for a 20 mm strip; B gives 300 N for a 30 mm strip. A colleague divides A’s force by a 20 mm by 20 mm overlap and calls the resulting number a peel property. What do the three normalisations mean?

A gives p_A = 200/20 = 10 N/mm. B gives p_B = 300/30 = 10 N/mm. The measured forces differ, but these width-normalised numbers agree.
Dividing A’s force by 400 mm² instead gives 0.500 N/mm² = 0.500 MPa. That is a chosen nominal force/area number, not the same definition or units as p.
Peel angle, strip bending and stretching, adhesive response, rate, conditioning and fracture location can affect the width-normalised output. Equal p is not a demonstrated equality of intrinsic interfacial fracture energy.

Both peel records give 10 N/mm on their own width basis. The 0.500 MPa area-normalised value answers a different question and cannot replace them.

Try the reasoning

Does every joining route melt both base materials?
Show the answer and reasoning
No.
Different mechanisms create different joint regions.
Why inspect adjacent material?
Show the answer and reasoning
The process can change it.
A heat-affected region may govern performance.
Keep P = 1,600 N and L = 20 mm, but increase a one-interface width from 20 to 25 mm. Find the new nominal mean. Has the local maximum necessarily decreased by the same percentage?
Show the answer and reasoning
The mean falls from 4.00 to 3.20 MPa, a 20% decrease. No local-maximum change is established.
A becomes 25 × 20 = 500 mm², so P/A = 3.20 N/mm². A local stress distribution needs compatibility, stiffness and boundary information; fixed resultant and area are insufficient.
In the ideal symmetric double-lap sketch, two identical interfaces each have b = 20 mm and L = 20 mm. The centre carries total P = 1,600 N. If symmetry splits the force equally, what is each interface’s nominal mean? Does doubling the interface count establish twice the real joint capacity?
Show the answer and reasoning
Each transfers 800 N and has nominal mean 2.00 MPa. Twice the actual joint capacity is not established.
The total bonded area is 800 mm² on two planes. Equal force sharing is a symmetry assumption, and local end stresses, adherend failure and changed geometry can still govern. Do not apply the one-interface 400 mm² denominator to the total as though it were each interface’s force.
A different supplied distribution has 5 MPa on each end quarter and 3 MPa over the middle half of the original 400 mm² interface. Find its resultant and compare its maximum with U and E.
Show the answer and reasoning
It also transfers 1,600 N, with maximum 5 MPa: above U’s 4 MPa and below E’s 6 MPa.
5 × 100 + 3 × 200 + 5 × 100 = 1,600 N. The same average can accompany several peaks. This supplied profile meets the inclusive 5 MPa illustrative rule exactly, without becoming an elasticity solution.
Start from the serial case with the HAZ proof stress improved to 180 MPa. A geometry change reduces the weld area to 10.0 mm², with weld proof stress still 300 MPa. Which region now sets the first equality force?
Show the answer and reasoning
The weld region sets 3,000 N, below the HAZ’s 3,600 N and base’s 4,800 N.
A stronger neighbouring state does not preserve the ranking when a different area changes. The minimum of the three local P = Aσ values changes; the model remains an axial proof-stress screen, not a joint breaking-load prediction.
A bonded coupon tears in its adherend at 900 N while the visible bond line remains attached. May 900 N be reported as the adhesive’s intrinsic shear strength, and may that coupon establish peel durability?
Show the answer and reasoning
No to both. The assembled connection transmitted the reached 900 N in that configuration; the adhesive’s unused capacity and other loading/exposure behaviour remain unknown.
The test ended through a competing region. Retain the adherend failure rather than converting force into a universal adhesive property or extrapolating to a different mode and time history.
Two bonded specimens have the same bulk adhesive cure indication. One separates near a contaminated adherend surface; the other tears within the adhesive. Must the cure measurement be wrong, or does equal cure prove equal interface attachment?
Show the answer and reasoning
Neither conclusion follows. Equal bulk setting evidence does not establish the same interface state.
Interfacial contact and chemistry, adhesive cohesion and the local loading are separate. Inspect both fracture surfaces and surface/preparation history at sufficient resolution, while keeping the mechanical test conditions matched.

What to carry forward

Describe the joining mechanism and loading.
Evaluate the joint and affected surroundings.
Bond creation, adhesive setting and mechanical force transfer are different parts of the explanation.
Keep nominal shear, local traction, peel force per width and proof stress on their own definitions.
A changed weld or adhesive number can leave another region governing the tested load path; retain failure location and relevant conditions.

References

Teaching resources and research supporting this lesson’s explanations and examples.