Cauchy's Momentum Equation
Statement
For a continuum with mass density \(\rho\), velocity field \(\mathbf{u}(\mathbf{x},t)\), Cauchy stress tensor \(\boldsymbol{\sigma}\) and body force per unit mass \(\mathbf{f}\), conservation of linear momentum for every material volume is equivalent, wherever the fields are smooth, to the pointwise partial differential equation \(\displaystyle \rho\,\frac{D\mathbf{u}}{Dt}=\nabla\!\cdot\!\boldsymbol{\sigma}+\rho\,\mathbf{f}\), where \(\dfrac{D}{Dt}=\dfrac{\partial}{\partial t}+\mathbf{u}\cdot\nabla\) is the material derivative.
Why it matters
Cauchy's momentum equation is Newton's second law written for a deformable continuum. It is the parent equation of essentially all of continuum mechanics: choosing a constitutive law for \(\boldsymbol{\sigma}\) collapses it to the Navier–Stokes equations (Newtonian fluid), the Euler equations (inviscid fluid), the Navier–Cauchy equations (linear elastic solid), or the equations of plasticity and viscoelasticity.
Its power is that it separates universal physics (momentum balance) from material-specific physics (the constitutive relation). The balance below is exact and model-independent; every rheological assumption enters only through \(\boldsymbol{\sigma}\).
Assumptions
Derivation
Result
Reading. Mass-times-acceleration per unit volume of a moving fluid parcel equals the net internal surface force per unit volume (the stress divergence) plus the body force per unit volume. The acceleration on the left is the material acceleration \(D\mathbf{u}/Dt\), which includes the convective part \((\mathbf{u}\cdot\nabla)\mathbf{u}\): even in a steady field a parcel accelerates as it moves through spatial gradients. The divergence of the stress measures the imbalance of tractions on opposite faces of an infinitesimal element; a uniform stress exerts no net force.
Units check. Each term has dimensions of force per volume. \(\rho\,D\mathbf{u}/Dt\): \(\mathrm{kg\,m^{-3}}\cdot\mathrm{m\,s^{-2}}=\mathrm{N\,m^{-3}}\). \(\nabla\!\cdot\!\boldsymbol{\sigma}\): stress \(\mathrm{Pa=N\,m^{-2}}\) divided by length \(\mathrm{m}\) gives \(\mathrm{N\,m^{-3}}\). \(\rho\,\mathbf{f}\): \(\mathrm{kg\,m^{-3}}\cdot\mathrm{m\,s^{-2}}=\mathrm{N\,m^{-3}}\). Consistent.
Limiting cases
- Statics / equilibrium \((\mathbf{u}=0)\): reduces to \(\nabla\!\cdot\!\boldsymbol{\sigma}+\rho\mathbf{f}=\mathbf{0}\), the equation of elastostatics; with \(\boldsymbol{\sigma}=-p\mathbf{I}\) it gives hydrostatics \(\nabla p=\rho\mathbf{f}\).
- Inviscid fluid \((\boldsymbol{\sigma}=-p\mathbf{I})\): since \(\nabla\!\cdot(-p\mathbf{I})=-\nabla p\), one gets the Euler equation \(\rho\,D\mathbf{u}/Dt=-\nabla p+\rho\mathbf{f}\).
- Newtonian fluid \((\sigma_{ij}=-p\delta_{ij}+\mu(\partial_i u_j+\partial_j u_i)+\lambda\,\delta_{ij}\nabla\!\cdot\mathbf{u})\): yields the Navier–Stokes equations; incompressible form \(\rho\,D\mathbf{u}/Dt=-\nabla p+\mu\nabla^2\mathbf{u}+\rho\mathbf{f}\).
- Linear elastic solid \((\boldsymbol{\sigma}=\lambda(\nabla\!\cdot\mathbf{w})\mathbf{I}+2\mu\,\boldsymbol{\varepsilon})\) with displacement \(\mathbf{w}\): gives the Navier–Cauchy elastodynamic equation \(\rho\,\ddot{\mathbf{w}}=(\lambda+\mu)\nabla(\nabla\!\cdot\mathbf{w})+\mu\nabla^2\mathbf{w}+\rho\mathbf{f}\).
- Uniform stress, no body force: \(\nabla\!\cdot\!\boldsymbol{\sigma}=0\Rightarrow D\mathbf{u}/Dt=0\); parcels move with constant velocity along their paths.
Breaks when
- Shocks and other discontinuities. Across a shock, contact surface or free interface the fields are not differentiable; the differential form is meaningless and must be replaced by the integral form, giving jump conditions \([\![\rho\mathbf{u}(\mathbf{u}\cdot\mathbf{n}-U_n)-\boldsymbol{\sigma}\mathbf{n}]\!]=0\) (Rankine–Hugoniot).
- Rarefied / non-continuum flow. When the Knudsen number \(\mathrm{Kn}=\lambda/L\gtrsim 0.1\) (upper atmosphere, MEMS, nanoscale channels) the continuum hypothesis fails; the equation must be derived as a moment of the Boltzmann equation, with slip and higher moments no longer closed by a simple \(\boldsymbol{\sigma}\).
- Relativistic flows. At speeds comparable to \(c\) (astrophysical jets, quark–gluon plasma) momentum and energy fluxes mix; one uses \(\partial_\mu T^{\mu\nu}=0\) instead.
- Media with internal structure / body couples. Polar (Cosserat) and liquid-crystal continua carry couple stresses, making \(\boldsymbol{\sigma}\) asymmetric, so an independent angular-momentum balance must be adjoined.
Failure modes
- Using the partial derivative \(\partial\mathbf{u}/\partial t\) instead of the material derivative \(D\mathbf{u}/Dt\). Dropping the convective term \((\mathbf{u}\cdot\nabla)\mathbf{u}\) removes exactly the nonlinearity responsible for turbulence and Bernoulli's theorem; it is legitimate only after linearization or for genuinely spatially uniform fields.
- Writing \(\nabla\!\cdot\!\boldsymbol{\sigma}\) as \(\boldsymbol{\sigma}\cdot\nabla\) or as a scalar. The stress divergence is a vector, \((\nabla\!\cdot\!\boldsymbol{\sigma})_i=\partial_j\sigma_{ij}\); contracting on the wrong index gives the wrong force for non-symmetric or spatially varying \(\boldsymbol{\sigma}\).
- Sign error on the pressure term. Because \(\boldsymbol{\sigma}=-p\mathbf{I}+\dots\), \(\nabla\!\cdot(-p\mathbf{I})=-\nabla p\); students frequently keep \(+\nabla p\) and get flow accelerating up the pressure gradient.
- Forgetting continuity in Step 3. Omitting \(\partial_t\rho+\nabla\!\cdot(\rho\mathbf{u})=0\) leaves the non-conservative form \(\partial_t(\rho\mathbf{u})+\nabla\!\cdot(\rho\mathbf{u}\otimes\mathbf{u})\), which is correct but must not also be equated to \(\rho\,D\mathbf{u}/Dt\) — that double-counts.
- Applying the pointwise PDE across an interface. Using it through a shock or a fluid–fluid boundary instead of the jump conditions produces unphysical, mesh-dependent results in simulations.
Discussion
The equation is a statement about surfaces masquerading as a statement about volumes. The only forces a continuum element feels from its neighbours act on its bounding surface, encoded by the traction field \(\mathbf{t}=\boldsymbol{\sigma}\mathbf{n}\). The divergence theorem is what lets these surface interactions be re-expressed as a local volumetric force \(\nabla\!\cdot\!\boldsymbol{\sigma}\). This is why the existence of the Cauchy stress tensor is a genuine prerequisite: without the linearity \(\mathbf{t}(\mathbf{n})=\boldsymbol{\sigma}\mathbf{n}\) there is no tensor to take the divergence of.
The split of the total force into a body part \(\rho\mathbf{f}\) and a contact part \(\nabla\!\cdot\!\boldsymbol{\sigma}\) mirrors the derivation's two threads. The body force ties to the matter thread — it is proportional to mass and represents long-range fields (gravity, electromagnetism) acting at a distance. The stress divergence is the force thread — short-range molecular interactions homogenized into a field. That both appear as force-per-volume is the content of the continuum idealization.
The symmetry thread enters through a companion balance: conservation of angular momentum for the same material volume, worked through identically, forces \(\boldsymbol{\sigma}=\boldsymbol{\sigma}^{\mathsf T}\) in the absence of body couples. Momentum balance thus fixes the divergence of \(\boldsymbol{\sigma}\) while angular-momentum balance fixes its symmetry; together they leave six independent stress components, exactly the number a symmetric constitutive law must supply.
Structurally, Cauchy's equation is a local conservation law \(\partial_t(\rho u_i)+\partial_j\Pi_{ij}=\rho f_i\) with momentum-flux tensor \(\Pi_{ij}=\rho u_i u_j-\sigma_{ij}\): the Reynolds (convective) flux \(\rho u_i u_j\) plus the stress flux \(-\sigma_{ij}\). In this conservative form it is the spatial part of the vanishing four-divergence of the non-relativistic stress–energy tensor, and it is the form finite-volume solvers integrate to guarantee discrete momentum conservation. Common misconceptions: that \(D\mathbf{u}/Dt\) is a total time derivative of a fixed particle's velocity in the naive sense (it is, but only when computed as the Eulerian field derivative following the flow); that \(\nabla\!\cdot\!\boldsymbol{\sigma}=\mathbf{0}\) means no stress (it means no net stress force — a body can be highly stressed yet in equilibrium); and that pressure is a separate physical entity from stress rather than the isotropic part \(p=-\tfrac13\mathrm{tr}\,\boldsymbol{\sigma}\).
Worked examples
Example 1 — Hydrostatic pressure in a still fluid. A tank of water is at rest under gravity \(\mathbf{f}=-g\hat{\mathbf{z}}\). Find the pressure at depth \(h=10\ \mathrm{m}\); take \(\rho=1000\ \mathrm{kg\,m^{-3}}\), \(g=9.81\ \mathrm{m\,s^{-2}}\), surface pressure \(p_0=1.013\times10^{5}\ \mathrm{Pa}\).
Reading. Ten metres of water roughly doubles the absolute pressure, adding about one atmosphere of gauge pressure — the familiar "one atmosphere per 10 m" diving rule falls straight out of Cauchy's equation.
Example 2 — Plane Poiseuille flow (wall-driven balance). Steady, fully developed, incompressible flow of water between parallel plates at \(y=\pm h\), \(h=1.0\ \mathrm{mm}\), driven by a constant pressure gradient \(-\,dp/dx=G=200\ \mathrm{Pa\,m^{-1}}\); viscosity \(\mu=1.0\times10^{-3}\ \mathrm{Pa\,s}\), no body force in \(x\). Find the centreline speed and the wall shear stress.
Reading. The parabolic profile is the signature of a pressure-driven viscous flow, and the wall shear \(\tau_w=Gh\) is fixed purely by geometry and the driving gradient — the viscosity cancels from \(\tau_w\) but sets the speed. Both follow directly from the \(x\)-component of Cauchy's equation.
Units check. \(u_{\max}\): \(\mathrm{Pa\,m^{-1}\cdot m^2/(Pa\,s)}=\mathrm{m\,s^{-1}}\). \(\tau_w\): \(\mathrm{Pa\,m^{-1}\cdot m=Pa}\). Consistent.
Problems
- (A) A fluid parcel in a steady one-dimensional flow has velocity field \(u(x)=U_0(1+x/L)\hat{\mathbf x}\) with \(U_0=2\ \mathrm{m\,s^{-1}}\), \(L=0.5\ \mathrm{m}\). Find the material acceleration at \(x=0\).
Solution
\(Du/Dt=\partial_t u+u\,\partial_x u=0+U_0(1+x/L)\cdot(U_0/L)\). At \(x=0\): \(=U_0^2/L=(2)^2/0.5=8\ \mathrm{m\,s^{-2}}\). Note the parcel accelerates despite the field being steady — purely convective acceleration. - (A) In a static atmosphere of uniform density \(\rho=1.2\ \mathrm{kg\,m^{-3}}\), by how much does the pressure drop over a height rise of \(100\ \mathrm{m}\)? Take \(g=9.81\ \mathrm{m\,s^{-2}}\).
Solution
Hydrostatic balance \(dp/dz=-\rho g\Rightarrow \Delta p=-\rho g\,\Delta z=-(1.2)(9.81)(100)=-1177\ \mathrm{Pa}\approx-1.18\ \mathrm{kPa}\). Pressure falls about 1.2 kPa (≈1.2% of an atmosphere) over 100 m. - (B) Starting from \(\rho\,D\mathbf u/Dt=\nabla\!\cdot\!\boldsymbol\sigma+\rho\mathbf f\), show that for an inviscid fluid \((\boldsymbol\sigma=-p\mathbf I)\) in steady flow with \(\mathbf f=-\nabla\Phi\) the quantity \(\tfrac12|\mathbf u|^2+\int dp/\rho+\Phi\) is constant along a streamline (Bernoulli).
Solution
With \(\boldsymbol\sigma=-p\mathbf I\), \(\nabla\!\cdot\!\boldsymbol\sigma=-\nabla p\), giving Euler \(\rho(\mathbf u\cdot\nabla)\mathbf u=-\nabla p-\rho\nabla\Phi\) (steady). Use the identity \((\mathbf u\cdot\nabla)\mathbf u=\nabla(\tfrac12|\mathbf u|^2)-\mathbf u\times(\nabla\times\mathbf u)\). Dot with the unit tangent \(\hat{\mathbf t}=\mathbf u/|\mathbf u|\); the vorticity term is orthogonal to \(\mathbf u\) and drops. Then \(\hat{\mathbf t}\cdot\nabla\bigl[\tfrac12|\mathbf u|^2+\int dp/\rho+\Phi\bigr]=0\), so the bracket is constant along the streamline. \(\square\) - (B) A Newtonian fluid \((\mu=8.9\times10^{-4}\ \mathrm{Pa\,s})\) flows in a pipe of radius \(R=5\ \mathrm{mm}\) driven by \(-dp/dx=G=500\ \mathrm{Pa\,m^{-1}}\). Using the axial Cauchy balance \(0=G+\mu\,\tfrac1r\tfrac{d}{dr}(r\,du/dr)\), find the centreline velocity and volumetric flow rate.
Solution
Integrate \(\tfrac1r(r u')'=-G/\mu\) with \(u(R)=0\), \(u'(0)=0\): \(u(r)=\tfrac{G}{4\mu}(R^2-r^2)\). Centreline \(u_{\max}=GR^2/4\mu=(500)(5\times10^{-3})^2/(4\cdot8.9\times10^{-4})=(500)(2.5\times10^{-5})/(3.56\times10^{-3})\approx3.51\ \mathrm{m\,s^{-1}}\). Flow rate (Hagen–Poiseuille) \(Q=\pi G R^4/8\mu=\pi(500)(6.25\times10^{-10})/(8\cdot8.9\times10^{-4})\approx1.38\times10^{-4}\ \mathrm{m^3\,s^{-1}}\). - (C) Show that the conservative (divergence) form \(\partial_t(\rho\mathbf u)+\nabla\!\cdot(\rho\mathbf u\otimes\mathbf u-\boldsymbol\sigma)=\rho\mathbf f\) is equivalent to \(\rho\,D\mathbf u/Dt=\nabla\!\cdot\!\boldsymbol\sigma+\rho\mathbf f\), and explain why finite-volume solvers prefer the conservative form.
Solution
Expand componentwise: \(\partial_t(\rho u_i)+\partial_j(\rho u_i u_j)=u_i[\partial_t\rho+\partial_j(\rho u_j)]+\rho[\partial_t u_i+u_j\partial_j u_i]\). The first bracket is zero by continuity; the second is \(\rho\,Du_i/Dt\). Adding \(-\partial_j\sigma_{ij}=-(\nabla\!\cdot\!\boldsymbol\sigma)_i\) reproduces the stated equivalence. Finite-volume schemes integrate the conservative form over each cell and apply the divergence theorem, so interior fluxes cancel telescopically between neighbouring cells and total momentum is conserved to machine precision — and, crucially, the weak (integral) form remains valid across shocks where the non-conservative \(D\mathbf u/Dt\) form is undefined, giving correct shock speeds via the built-in Rankine–Hugoniot conditions.