Basic hydrodynamics

David Gurarie

Newtonian uids: Euler and Navier-Stokes equations

The basic hydrodynamic equations in the Eulerian form consist of conservation of mass, momentum and energy. We denote by u = u (x; t) velocity eld, and - density eld, with Dt = t + u - derivative (of any material quantity) along the ow direction.

1.1

Mass conservation

The continuity (conservation) of mass requires the mass transfer rate along the ow Dt be balanced by its change due to the volume compression/expansion by the ow u , t + (u) = Dt + ( u) = 0 If density is constant, (1) turns into incompressibility condition u=0 (1)

1.2

Momentum conservation

The change of momentum Dt u of a Newtonian uid is balanced by a combination of body forces Fb (distributed through the volume) and surface forces Fs . The former include gravity, Coriolis (rotation), Coulomb, Lorentz (for electrically conducting uids) etc. The latter have a general form Fs = N dS where is the stress-tensor, N -normal to the surface element dS. The stresses is made of two components: pressure p and viscous (frictional) forces, depending on the rate of (velocity) deformation: u = (i uj ). The latter is made of the ric rate of strain-tensor sij = (i uj + j ui ). Only the latter (strain) would contribute to the viscous stresses. antisymmetric (rotation) component: u =
1 2 i uj j ui 2

, and the symmet-

Assuming isotropy (rotational symmetry) one can derives the following representation of the combined pressure + viscosity stresses 1 ij = p ij + sij + tr (s) ij 3 (3)

Lame coecients , measure the dynamic viscosity, while tr (s) = u gives divergence. Now the momentum conservation: t (u)+[u (u)] = turns into the Navier-Stokes equation t (ui ) + [u (ui )] = i p + ui +i u , or (4) 3 = i p + 2 ui + i ( u) , for const , 3 written here for the i-th component of momentum. Dividing by and using mass conservation (1) it takes the form Dt u = p + u+ ( u) 3 (5)

where = , = are the kinematic viscosities. Two other simplications arise for incompressible uids (second viscous term drops)

Dt (u) = and the ideal Euler uid (no viscosity)

p + u

Dt u = ut + (u ) u +

p =0

(6)

The Euler momentum equation can be rewritten via (15) as 2 u p ut + + u ( u) = 0 2

(7)

If Pn denotes the space of all such matrices, the isotropy of the linear strain-to-stress map : s implies that commutes with the natural action of Lie group SO (n) on space Pn : s U sU (conjugation). Any such map of Pn ( SO (n)-invariant) is well known to be of the form s0 + s0 s0 + s0 So is determined by a pair of scalars , - Lame coecients. In other words, respects the scalar+traceless decomposition (2), hence follows (3).

1 Relation (3) results from the decomposition of a symmetric n n-matrix s into its scalar and traceless components s0 + s0 tr (s) tr (s) I + s I s= (2) n n | {z } | {z } s0 s0

In particular, steady-state potential ow ( u = 0) has Bernoulli form, u2 p W = + conserved along ow-lines 2 u2 p ut = 0 + = Const =u=0 2 Taking div of (7) we get t ( u) + p u2 + 2 | u|2 u u = 0

Hence the ideal uid pressure is a quadratic functional of velocity u o u2 n u2 p 2 = 1 | u| + u u = 1 2 + u u 2 2 t (u ) = t + (u ) ( ) u = 0 (8)

Taking div of (7) we get the Euler equation for the vorticity eld

(9)

Integral conservation laws

Several integral conservation laws are easily veried for Euler uids: Momentum: ZZZ u dx = const is conserved due to u = 0 Energy: E= ZZZ
1 2 2 u dx

= const

Indeed, 2 2 2 u u p p u = u ut = u + u (u ) = u + + t 2 2 2 -complete divergence. Total vorticity ZZZ dx = const

Indeed, incompressibility of elds u; implies (u ) f = (f u) and ( ) f = (f ) for any f . Hence each component k of obeys by (9) t ( k ) = (uk k u) - complete divergence 3

Circulation (Kelvin): for any loop t that moves along the ow I u ds = const C (t) =
t

We use the transport along the ow I 2 I u d p Dt u ds = C= + u ds dt 2 t t H RR The grad-term gives 0 along any closed loop, while the (u ) = t (u ) N dS = 0, due to the zero integrand (16). Helicity (Mofatt): ZZZ (u ) dx = const One writes 2 ut + u + p u = 0 2 t (u ) = 0

multiplies the rst equation by the second one by u and sums together 2 u p + (u ) u { (u )} + ( u)t = 2 The 2-nd term in the r.h.s. is 0, while the 3-rd is transformed into u ( R) = (R u) R ( u) = (R u) where R = u and R ( u) = R = 0 vanishes. Hence 2 u p ( u)t = R u - complete divergence + 2

System of conservation laws

The basic Euler gas-uid dynamics written as mass-momentum conservation for (, u, p) is incomplete unless an equation of state is imposed, that would link pressure p to , u, and/or other parameters of state, s.a. entropy, temperature, etc. The simplest case is isentropic (barotropic) gas, where p = p () - a given function of , e.g. C . One example is provided by the shallow water equations for (u, h), viewed as 2D compressible uid of density equal to height h and pressure p = 1 gh2 , 2 Dt u = gh t h + (uh) = 0 4 (10)

More generally the internal state is described by specic entropy (per unit mass) s, that obeys the conservation law t (s) + (us) = 0 -active tracer, and p = p (, s) - a function of pressure and density. The combined system for (, u, s) has the conservation law form u v 2 u + x u + p + y uv = 0 t v uv v 2 + p s us vs (11)

(12)

for 2D, and similar expression for 3D. In the isentropic case the last equation drops, so the system order is reduced by 1. Euler system (12) has another conserved quantity, the total energy E, made of the kinetic one K = 1 u2 , and the potential (internal) one E (, s). Indeed, 2 2 u t K = u + p + p ( u) (13) 2 t E = [uE] + ( u) (E E sEs ) Hence, the total energy t (K + E) + [u (K + E + p)] = ( u) (p + E E sEs ) (14)

-conserved provided the r.h.s. is zero. For incompressible uids u = 0, the kinetic energy is conserved, along with any function E (, s), as , s are passive scalars. In the compressible case we get the dierential relation between p and E p = ( + ss ) E E In particular,E = s yields p = ( + 1) s . For shallow water, E = 1 1 2 2 2 gh yields p = 2 gh .

4
I

Appendix: vector algebra

(u v) = v ( u) u ( v) (a u) = a ( u) , for a - const (f u) = f u + f u (u v) = (v ) u (u ) v + u ( v) v ( u) (a v) = +a ( v) (a ) v, for a - const ( u) = ( u) u t u ( v) = u (v) (u ) v u ( v) + v ( u) = (u v) (u ) v (v ) u (u ) u =
u2 2

Commonly used vector identities for scalar/vector elds:

II III IV V VI VII

(15)

u ( u) 5

Here v = (i vj ) denotes the complete derivative (Jacobian matrix) of v.f. v. Relation (I) follows from the formal determinant expansion by swapping rows 1 2 3 v1 v2 v3 u1 u2 u3 u1 u2 u3 = 1 2 3 1 2 3 v1 v2 v3 u1 u2 u3 v1 v2 v3 u v = (bh cg, cf ah, ag bf ) and the rst component of the curl becomes (ay g by f ) (cz f az h) (fy b gy a) + (hz a fz c) Here the second pair of brackets is obtained by interchanging elds u, v in the anti-symmetric expression (u v). Adding and subtracting afx the rst pair is brought into the form f ax + gay + haz (ax + by + cz ) f = 1-st comp. (v ) u ( u) v The second pair gives a similar expression with u, v interchanged, Q.E.D. Relations (VI-VII) are consequences of (V). To verify the latter we write u = (a; b; c) and v = (f ; g; h). then u ( v) is equal to i j k = (b (gx fy ) c (fz hx ) ; ...; ...) a b c hy gz fz hx gx fy Adding and subtracting afx we arrange the 1-st component to be afx + bgx + chx (afx + bfy + cfz ) = u x v (u ) f By the same pattern the 2-nd and 3-rd components are u y v (u ) g u z v (u ) h whence follows (V). Corollary 1 If u denotes incompressible velocity and = u -vorticity, then one has (u ) = 2 + u u (u ) = ( ) u (u ) (16)

To check (III) we call the components of elds u = (a, b, c) ; v = (f, g, h). Then