[Intermediate] — The one modelling assumption the whole field rests on, when it holds, and why solids and fluids turn out to be the same theory with different closures.
Source: [reference] — see the domain note on what that means here.
The continuum hypothesis
Matter is atoms with space between them. Continuum mechanics ignores that entirely and treats material as continuously distributed — every point in a body has a density, a velocity, a stress, and these vary smoothly.
That’s obviously false at small enough scale, and it’s the assumption that makes the whole field work.
The justification is separation of scales. Define density at a point as the mass in a small volume around it, divided by that volume:
Shrink and you see three regimes:
density
│ ╱╲ ← too small: individual atoms,
│ ╱ ╲ wild fluctuations
│ ╲╱╲╱ ╲___________________ ← the plateau: the REV
│ ╲___ ← too large: real variation
└──────────────────────────────────────→ ΔV
atomic representative body
elementary volume
The plateau is the representative elementary volume (REV) — large enough to average over many molecules, small enough to be “a point” relative to the body. The theory works when that plateau exists.
Where it breaks:
- Rarefied gases — at high altitude the mean free path approaches the body size. The Knudsen number tells you: and continuum is fine; above ~0.1 you need kinetic theory
- Nanoscale — a 5 nm film is tens of atoms thick. Molecular dynamics territory
- Granular materials — sand is a continuum at the scale of a beach and discrete at the scale of a grain
- Cracks — a crack tip has a stress singularity in continuum theory, which is the theory announcing it’s outside its domain. Fracture mechanics works around this → Failure and Yield
What you gain is enormous: the machinery of calculus. Fields, derivatives, integrals, and PDEs instead of coupled ODEs.
Solids and fluids are the same theory
The structural insight worth having early. Both obey the same conservation laws:
Those give you equations relating stress to motion, and they’re the same equations for steel, water, air, blood, and rock. → Conservation Laws
The equations are also underdetermined — more unknowns than equations. What closes the system is a constitutive model: a statement about how this particular material responds to deformation. → Constitutive Models
| Solid | Fluid | |
|---|---|---|
| Responds to | how much you deform it (strain) | how fast you deform it (strain rate) |
| Constitutive law | ||
| At rest under shear | holds a deformed shape | keeps flowing |
| Simplest model | Hooke’s law | Newtonian viscosity |
That’s the entire distinction. A fluid is a material that cannot resist shear at rest; a solid can. Everything else — the balance laws, the stress tensor, the kinematics — is shared.
Which is why the same course covers Navier–Stokes and beam bending, and why viscoelastic materials (polymers, biological tissue, asphalt) are genuinely awkward: they’re both, depending on timescale. Silly putty bounces and also flows.
The three ingredients
Every problem in this field is assembled from three parts, and knowing which one you’re missing is most of getting unstuck:
1. Kinematics — the geometry of deformation. How do you describe motion and quantify “how much has this deformed”? No physics, just geometry. → 03, 04
2. Balance laws — conservation of mass, momentum, energy. Physics, and material-independent. → 06
3. Constitutive models — how this material relates stress to deformation. Material-specific, and empirical. → 07
kinematics + balance laws + constitutive model = a solvable problem
(geometry) (physics) (the material)
The balance laws are as close to certain as anything in engineering. The constitutive model is where the modelling error lives, and it’s the part chosen rather than derived. When a simulation is wrong by a factor of three, look there first.
Why tensors
You cannot avoid them, so it’s worth knowing why.
Consider stress. At a point inside a loaded body, cut an imaginary plane and ask what force the material on one side exerts on the other. The answer depends on which plane you chose.
A horizontal cut through a beam under bending sees mostly shear. A vertical cut sees mostly normal stress. Same point, different answers.
So stress is not a number and not a vector — it’s an object that takes a plane orientation and returns a traction vector. That’s a linear map from vectors to vectors: a second-order tensor.
Nine components in 3D (six independent, by symmetry). Not because someone wanted it complicated — because the physical quantity genuinely has that many degrees of freedom. → Index Notation and Tensors
Same for strain, and for the deformation gradient.
The two viewpoints
A recurring source of confusion, so it’s worth naming now.
Lagrangian (material) — follow a particle. “Where is this bit of material now?” Natural for solids, where you care about a specific piece of steel.
Eulerian (spatial) — watch a fixed point in space. “What’s the velocity at this location right now?” Natural for fluids, where tracking individual water molecules is absurd.
Both describe the same physics. Solid mechanics is usually Lagrangian, fluid mechanics usually Eulerian, and the translation between them is where a lot of the notation comes from — the material derivative, the two stress measures, the several strain measures. → Kinematics
The simplifying assumptions
Most engineering practice is continuum mechanics plus assumptions that make it tractable. Knowing which are active is the skill:
Small strain — deformations under ~1%. Lets you use the linearised strain tensor and ignore the difference between deformed and undeformed geometry. This one assumption removes most of the difficulty, and it’s why undergraduate solid mechanics is manageable. → Finite Deformation is what happens without it.
Linear elasticity — stress proportional to strain, fully reversible. True for metals below yield, false for rubber, soil, and anything plastic.
Isotropy — same properties in every direction. True for most metals, false for wood, composites, bone, and rolled sheet.
Homogeneity — same properties everywhere.
Incompressibility — density constant. Excellent for liquids and for gases below Mach 0.3.
Every textbook formula you’ve used carries some subset of these. The formula isn’t wrong; the assumptions are. Being able to name them is the difference between using a result and understanding it.
Where this is used
- Structural analysis — will this hold? → beams, plates, FEM
- Fluid dynamics — aerodynamics, pipe flow, weather, blood flow
- Manufacturing — forming, casting, injection moulding, additive
- Geomechanics — soil, rock, earthquakes
- Biomechanics — tissue, implants, cardiovascular
- Computer graphics — cloth, hair, fluid and soft-body simulation are all this, with speed prioritised over accuracy
- Robotics — soft robotics and compliant mechanisms are continuum problems
Reading this track
The order is deliberate: kinematics before stress before balance laws before constitutive models, because each genuinely needs the previous. Notes 02–06 are the foundation and can’t be skipped. After that, 07–13 are applications and you can take what you need.
The prerequisite is multivariable calculus and linear algebra — partial derivatives, divergence, gradient, matrices, eigenvalues. The vault’s maths notes cover these at the level ML needs, which is a floor rather than enough; tensor calculus and PDEs are a genuine gap.
Related
- Index Notation and Tensors — the language
- Constitutive Models — where the modelling error lives
- Hardware: Electricity — the other physical domain here
- Continuum mechanics map
Software engineering
Infrastructure & security
Computer science
Hardware & engineering