[Intermediate → Advanced] — Describing motion without any physics. Pure geometry, and the source of most of the field’s notational baggage.

The two configurations

A body occupies a region of space. It moves and deforms. To describe that, you need two pictures:

   REFERENCE (undeformed)              CURRENT (deformed)
        t = 0                              t = t
    ┌─────────┐                          ╱────────╲
    │  •  X   │        χ(X, t)          ╱   • x    ╲
    │         │      ──────────→       │            │
    └─────────┘                         ╲__________╱

    material coordinates X              spatial coordinates x

labels a material particle — its position in the reference configuration. It’s a name for a bit of material, and it never changes.

is where that particle is now.

The motion is the map between them:

Everything in kinematics is a statement about this map.

The displacement is the difference:

Lagrangian vs Eulerian

The same field can be written as a function of or of , and the choice is the difference between two whole traditions.

Lagrangian (material) description — quantities as functions of . “What is the temperature of this particular particle?”

Natural for solids: you care about a specific piece of steel, the reference configuration is a real remembered state, and the material has a history.

Eulerian (spatial) description — quantities as functions of . “What is the temperature at this location?”

Natural for fluids: tracking individual water molecules is absurd, there’s no meaningful reference configuration, and you care about the flow field.

Solid mechanics is Lagrangian. Fluid mechanics is Eulerian. That single fact explains most of the notational difference between the two fields, and why they look like separate subjects despite sharing the same balance laws.

The material derivative

The bridge between the two descriptions, and the first genuinely subtle idea.

Suppose you want the rate of change of temperature for a given particle, but temperature is written as a spatial field . The particle is moving, so it samples different locations:

Local term — the field is changing at this fixed location.
Convective term — the particle is moving into a region where the field differs.

The standard physical example: stand in a river with a thermometer. is the water at your position warming. is you drifting downstream into warmer water. The material derivative is what the thermometer actually reads.

Acceleration is the case that matters:

That second term is nonlinear in velocity, and it is the reason fluid dynamics is hard. Navier–Stokes would be a manageable linear PDE without it; with it you get turbulence, and a Millennium Prize problem. → Navier–Stokes

Steady flow does not mean zero acceleration. In a converging nozzle at steady state, everywhere and fluid still accelerates, because it moves into a region of higher velocity. That’s the convective term, and it catches people every time.

The deformation gradient

The central object of finite-deformation kinematics. Everything else is built from it.

It maps an infinitesimal material line element from the reference configuration to the current one:

reference                     current
   dX  ────────→  F  ────────→  dx
  (a tiny fibre                (the same fibre,
   of material)                 stretched and rotated)

Note the mixed indices — lowercase for current, uppercase for reference. is a two-point tensor, with one foot in each configuration. That’s unusual and it’s why it doesn’t behave like the tensors in note 02 under a single rotation.

In terms of displacement:

The Jacobian is its determinant:

the ratio of deformed to undeformed volume. Two facts follow immediately:

  • always. means a volume collapsed to nothing; means the material turned inside out. Both are physically impossible, and a finite-element solver reporting a negative Jacobian is telling you an element has inverted — usually a sign the mesh or the load step is wrong
  • Incompressible ⟺ . That’s the exact statement of incompressibility, and it becomes a constraint you enforce

Polar decomposition

contains both stretch and rotation, mixed together. Separating them is essential, because rotation must not produce stress — spinning a body doesn’t deform it.

Every with decomposes uniquely:

  • is a rotation (orthogonal, )
  • is the right stretch tensor (symmetric, positive definite) — stretch first, then rotate
  • is the left stretch tensor — rotate first, then stretch
      U (stretch)          R (rotate)
 □ ──────────────→ ▭ ──────────────→ ◊
        or
      R (rotate)           V (stretch)
 □ ──────────────→ ◇ ──────────────→ ◊

Same final state, different order.

This decomposition is why strain measures are built from rather than . Since , the rotation cancels — leaving pure stretch. A constitutive law written in terms of automatically produces no stress under rigid rotation, which is the requirement of material frame indifference. → Strain Measures

The eigenvalues of are the principal stretches — the stretch ratios along three orthogonal material directions. A stretch of 1 means unchanged; 1.5 means 50% longer.

Computing and requires a matrix square root, which is why in practice you work with directly and avoid the decomposition.

The velocity gradient

The rate-based counterpart, and the one fluids need:

Split it, using the symmetric/antisymmetric decomposition from note 02:

is actual deformation rate — stretching and shearing. is rigid rotation rate and causes no deformation at all.

This is the split that makes Newtonian fluids work. Viscous stress depends on and not on — a fluid in solid-body rotation (a bucket of water spun up to steady state) has and experiences no viscous stress, despite . If viscous stress depended on , rotating a bucket would heat the water, which it doesn’t.

The vorticity is the axial vector of :

and the local angular velocity. Vorticity is the central object of much of fluid dynamics — it’s what tornadoes, wingtip vortices and turbulent eddies are made of.

Compatibility

A subtlety worth knowing exists.

If you’re given a displacement field , you can always compute a strain field. The reverse is not true — an arbitrary symmetric tensor field is not necessarily the strain of any continuous displacement field.

Six strain components come from three displacement components, so they’re over-determined: there are constraints. The compatibility conditions (in small strain, ) are what a strain field must satisfy to correspond to a real, gap-free, overlap-free deformation.

Physically: an incompatible strain field would tear the material or make it overlap itself.

This matters when you solve in terms of stress rather than displacement, and it’s the reason displacement-based finite elements are the default — assume a continuous displacement field, and compatibility is satisfied automatically. → FEM

The small-strain simplification

Almost everything above collapses when displacement gradients are small (, in practice under ~1%):

  • The reference and current configurations become interchangeable — you can write equilibrium on the undeformed geometry
  • , and all strain measures coincide
  • The material derivative’s convective term becomes negligible for solids
  • Everything becomes linear, so superposition works

That last point is why linear elasticity is tractable at all — you can solve for one load case and add. → Linear Elasticity

When it fails: rubber, biological tissue, metal forming, buckling, and anything with large rotations. A cantilever tip deflecting 30% of its length has small strains and large rotations, and the small-strain tensor gets it wrong — a rigid rotation registers as spurious strain. That’s the case for finite deformation. → Finite Deformation


↑ Back to Continuum Mechanics