Thermal Expansion — From Bridge Joints to Solder Fatigue

Category: Physics Fundamentals | 2026-03-25 | サイトマップ
NovaSolver Contributors
Table of Contents
  1. Thermal Expansion: Small Effects, Big Consequences
  2. Linear Thermal Expansion: ΔL = αLΔT
  3. CTE of Engineering Materials
  4. Constrained Thermal Stress: σ = EαΔT
  5. Bridge Expansion Joints and Railway Buckling
  6. Bimetal Actuators and Thermostat Strips
  7. CTE Mismatch in Electronics Packaging
  8. FEM Thermo-Structural Coupling
  9. Cross-Topics

1. Thermal Expansion: Small Effects, Big Consequences

Thermal expansion coefficients are tiny numbers — typically $10^{-6}$ per kelvin. Yet these small numbers, multiplied by large structures, large temperature swings, and high stiffness, produce forces and stresses large enough to buckle railway tracks, shatter ceramic coatings, and fatigue solder joints on circuit boards through millions of thermal cycles.

In FEM, thermal loading is often overlooked in initial design phases but becomes critical in detailed design — particularly for structures exposed to temperature cycles (engines, electronics, bridges) where differential expansion between materials causes cumulative fatigue damage.

🧑‍🎓

Bridge expansion joints seem huge — like 10cm wide. How much does a bridge actually expand? It seems like overkill.

🎓

It's not overkill at all. Let's calculate: a 100m steel bridge with CTE α = 12×10⁻⁶/K and a seasonal temperature swing of 50°C: ΔL = αLΔT = 12e-6 × 100 × 50 = 0.06 m = 60 mm. That's 6 centimeters. A long rail bridge of 500m would need 300mm of movement. If you prevented that expansion, the compressive stress would be σ = EαΔT = 210×10⁹ × 12e-6 × 50 = 126 MPa — close to the yield stress of structural steel, and far beyond what's needed to buckle a rail. Railways in hot climates "sun kink" — the rail buckles sideways on a hot day if expansion joints are missing or clips fail. That's exactly what happens before derailments on hot summer days.

2. Linear Thermal Expansion: ΔL = αLΔT

The linear coefficient of thermal expansion (CTE) $\alpha$ characterizes how much a material lengthens per unit length per degree of temperature change:

$$\Delta L = \alpha L_0 \Delta T, \qquad \varepsilon_{\text{thermal}} = \alpha \Delta T$$

For isotropic materials, the same expansion occurs in all directions. For anisotropic materials (carbon fiber composites, single crystals), CTE is a tensor with different values in different directions. For a 3D solid, the thermal strain in FEM is added to the mechanical strain:

$$\boldsymbol{\varepsilon}_{\text{total}} = \boldsymbol{\varepsilon}_{\text{mechanical}} + \boldsymbol{\varepsilon}_{\text{thermal}} = \frac{\boldsymbol{\sigma}}{\mathbf{D}} + \boldsymbol{\alpha}\Delta T$$

The mechanical strain $\boldsymbol{\varepsilon}_{\text{mechanical}} = \boldsymbol{\varepsilon}_{\text{total}} - \boldsymbol{\alpha}\Delta T$ is what drives stress — thermal strain alone (in an unconstrained material) produces no stress.

3. CTE of Engineering Materials

MaterialCTE α (×10⁻⁶/K)E (GPa)EαΔT at 100°C (MPa)
Fused silica (SiO₂)0.55734.0
Invar (Fe-Ni 36%)1.214117
Silicon (Si chip)2.615039
Borosilicate glass (Pyrex)3.36421
Carbon fiber (CFRP 0°)0.0–21350–27
Titanium (Ti-6Al-4V)8.611498
Steel (carbon)11–13210230–273
Copper17117199
Aluminum alloy2369159
Solder (SnAgCu)2150105
FR4 PCB (in-plane)14–172028–34

Note the wide range — from Invar at 1.2 to aluminum at 23 ppm/K. Joining materials with very different CTEs creates stress when they're bonded and subjected to temperature change.

4. Constrained Thermal Stress: σ = EαΔT

When thermal expansion is fully constrained (zero displacement allowed), the entire thermal strain is converted to mechanical stress:

$$\sigma_{\text{thermal}} = E\varepsilon_{\text{thermal}} = E\alpha\Delta T$$

For steel heated by 100°C while fully constrained: $\sigma = 210 \times 10^3 \times 12 \times 10^{-6} \times 100 = 252$ MPa — already near yield of mild steel (235–355 MPa). A 200°C heating cycle would exceed yield, causing permanent plastic deformation.

Thermal Shock and Ceramics

Ceramics have low CTE (2–6 ppm/K) and high $E$ but very low fracture toughness. When a ceramic component experiences a sudden temperature gradient (thermal shock), the surface expands or contracts faster than the interior. The resulting thermal stress can exceed the ceramic's tensile fracture strength, causing catastrophic cracking. This is why refractory bricks, ceramic cutting tools, and glass cookware are designed to minimize thermal shock — and why thermal FEM with transient analysis is essential in high-temperature ceramic component design.

5. Bridge Expansion Joints and Railway Buckling

Bridge expansion joints accommodate the seasonal and diurnal thermal movement of bridge decks. Without them, thermal stress would cause either buckling (compression in summer) or tensile fracture (winter shrinkage).

Railway Track Stability

Continuous welded rail (CWR) has no expansion joints. Instead, it's laid at a specific "stress-free temperature" (typically 27°C in temperate climates, 35°C in hot climates) with a pre-tension applied during welding. The rail is stressed in tension at low temperatures and compression at high temperatures. At the maximum rail temperature (~60°C summer), the compressive stress:

$$\sigma = E\alpha\Delta T = 210 \times 10^9 \times 12 \times 10^{-6} \times (60-27) = 83.2 \text{ MPa}$$

If this compressive stress, combined with any horizontal mis-alignment or soft ballast, exceeds the lateral buckling resistance, the track buckles sideways — "sun kink." Railway infrastructure maintenance agencies monitor track temperature and issue speed restrictions on hot days based on exactly this FEM-validated calculation.

6. Bimetal Actuators and Thermostat Strips

A bimetallic strip bonds two metals with different CTEs. When heated, the metal with higher CTE expands more, causing the strip to bend toward the lower-CTE side. The relationship between temperature change and tip deflection for a bimetal strip of length $L$ and total thickness $h$:

$$\delta_{\text{tip}} = \frac{3(\alpha_2 - \alpha_1)(T-T_0)L^2}{2h}$$

For a 100mm strip, $\alpha_2 - \alpha_1 = 10 \times 10^{-6}$/K, $h = 1$ mm, $\Delta T = 50$ K: tip deflection = $3 \times 10^{-5} \times 50 \times 0.01 / 0.002 = 7.5$ mm — easily enough to open or close an electrical contact. Bimetal thermostats, circuit breakers, and HVAC actuators all work on this principle.

🧑‍🎓

I need to model a bimetal actuator in Abaqus. Do I model the two layers separately and join them, or can I use a special element type?

🎓

Model the two layers as separate materials with different CTE values (*EXPANSION), bonded with a *TIE constraint (or shared nodes at the interface, which assumes perfect bonding). Apply a temperature field (*TEMPERATURE or *PREDEFINED FIELD) as the load. The different expansion rates will naturally cause bending — Abaqus will compute the curvature and tip deflection automatically. Use NLGEOM if the deflections are large relative to thickness. One important detail: specify the reference temperature correctly — the temperature at which both layers have zero thermal strain (the manufacturing/bonding temperature, not room temperature). Getting this wrong shifts the entire curvature vs. temperature curve.

7. CTE Mismatch in Electronics Packaging

Electronic packaging suffers the most severe CTE mismatch problems in engineering. A silicon chip (CTE = 2.6 ppm/K) soldered to a copper PCB pad (CTE = 17 ppm/K) experiences repeated differential thermal strain each power cycle:

$$\Delta\varepsilon = (\alpha_{\text{Cu}} - \alpha_{\text{Si}})\Delta T = (17 - 2.6) \times 10^{-6} \times 80 = 1.15 \times 10^{-3}$$

This 0.115% strain amplitude per thermal cycle, accumulated over 10,000–100,000 power cycles, fatigue-cracks the solder joint connecting chip to board. This is the dominant failure mode of flip-chip BGA (Ball Grid Array) packages in automotive electronics, where temperature swings from -40°C to +125°C are required for 15+ years.

Coffin-Manson Solder Fatigue Life

The Coffin-Manson relation predicts solder joint fatigue life $N_f$ from plastic shear strain amplitude $\Delta\gamma_p$:

$$N_f = \frac{1}{2}\left(\frac{\Delta\gamma_p}{2\varepsilon_f'}\right)^{1/c}$$

FEM thermo-mechanical analysis computes the plastic strain distribution in each solder joint, feeding directly into this life prediction. The FEM-Coffin-Manson methodology is now the industry standard for BGA package qualification, with Ansys, Abaqus, and COMSOL all having validated implementations.

8. FEM Thermo-Structural Coupling

Thermo-structural FEM requires two analyses in sequence (or simultaneous coupling):

  1. Thermal analysis: Solve the heat equation to get temperature field $T(x,y,z)$ or $T(x,y,z,t)$
  2. Structural analysis: Apply the temperature field as a load; nodes expand according to their local temperature and CTE

In Abaqus, the thermal strain is automatically included in the constitutive model when *EXPANSION is defined:

*MATERIAL, NAME=STEEL
*ELASTIC
210000., 0.3
*EXPANSION
12.E-6,   ! CTE in ppm/K (coefficient per degree C)

*STEP
*STATIC
*TEMPERATURE, FILE=THERMAL_RESULTS.odb
*END STEP

Fully Coupled vs. Sequential Approach

For problems where structural deformation changes heat transfer (e.g., contact opens/closes under thermal loading), fully coupled thermal-structural analysis is needed. Use Abaqus coupled temp-displacement elements (C3D8T, S4T). For most engineering problems where the temperature field doesn't depend on the deformation, the sequential approach is adequate and significantly faster.

🧑‍🎓

I ran a thermal-structural FEM on an exhaust manifold and the thermal stress is higher than yield at every hot section. Does that mean the part will fail?

🎓

Not necessarily — high thermal stress in linear elastic analysis often indicates plastic relaxation in real operation. Exhaust manifolds routinely "yield" on the first heat cycle, accumulating small plastic strains that reduce the stress (called "shakedown"). The correct analysis for hot components is an elastic-plastic thermo-structural simulation over several thermal cycles, looking at whether the plastic strain range (not peak stress) stabilizes. If the plastic strain range per cycle is below about 0.2–0.5%, the manifold will typically achieve shakedown and survive thermally. If plastic strain keeps accumulating cycle after cycle (ratcheting), then you have a real failure problem. Use *PLASTIC with temperature-dependent yield strength in Abaqus to capture this correctly.

9. Cross-Topics

TopicConnectionLink
Heat & TemperatureTemperature field from thermal FEM drives thermal expansionHeat and Temperature
Stress & StrainConstrained thermal strain produces thermal stress σ=EαΔTStress and Strain Basics
Hooke's LawLinear elastic material connects thermal strain to thermal stressSprings & Hooke's Law
StaticsThermal equilibrium analysis uses static FEM with temperature loadsStatics and Equilibrium
Heat Transfer TheoryAdvanced thermo-structural coupling and creep at high temperatureHeat Transfer