11–Applications of the Divergence Theorem:

The Compressible Euler Equations

MATH 22C

1. Derivation of the Compressible Euler Equations

In this section we use the divergence theorem to derive a physical inter-
pretation of the compressible Euler equations as the continuum version
of Newton’s laws of motion. Reversing the steps then provides a deriva-
tion of the compressible Euler equations from physical principles. The
compressible Euler equations are
ρt + Div(ρu = 0 (1)
(ρu )t + Div(ρui u + pei ) = 0,
i
(2)
where ρ is the density (mass per volume), u = (ui , u2 , u3 ) is the veloc-
ity, and p is the pressure (force per area) in the fluid. For example, and
for simplicity, we can assume the pressure is a known function of the
density through the equation of state p = p(ρ). A solution of the equa-
tions would consist of functions ρ(x, t), u(x, t), x = (x, y, z) that meets
the partial derivative constraints (1) and (2) at every point. The com-
pressible Euler equations describe the motion of a compressible fluid
like air, under the assumption that there is no viscosity or dissipation.
One can view these as the expressioin of Newton’s laws for a continuous
media. That is, in words, the first Euler equation (1), (often referred
to as the continuity equation), expresses conservation of mass, that
mass is nowhere created nor destroyed. The second Euler equation,
(2), is the continuum version of Newton’s force law F = ma = d(mv) dt
.
It expresses that changes of momentum are due solely to gradients in
the pressure. It is interesting that when the pressure depends only on
the density, p = p(ρ), these two Euler equations stand on their own.
But when the pressure depends on the temperature as well, we obtain
the full system of compressible Euler equations by including one final
equation for the energy, written most simply in the form
Et + Div[(E + p)u] = 0, (3)
where E is the energy per volume. For smooth solutions that are shock
free, the energy equation can be replaced by the much simpler entropy
equation
St + Div(Su) = 0. (4)
To make sense of this, we must introduce the second law of thermo-
dynamics and explain the connections between the energy density E,
1
2

the entropy density S and the pressure p. The explanation is most

interesting because it connects up the compressible Euler equations to
thermodynamics, and by this we’ll see that entropy (a measure of re-
versibility) can be given a precise definition in the context of a perfect
fluid, and we’ll see that (4) precisely expresses that entropy does not in-
crease along particle paths when solutions are smooth. As we’ll see, this
is consistent with the fact that we neglect viscosity and heat conduc-
tion in the derivation of (1)-(4). In further developments of the theory
of shock waves, (see e.g. [?]), it follows that the entropy equation (4)
is violated on shock waves, but the condition that entropy increases on
shock waves is sufficient to rule out the unphysical rarefaction shocks
that satisfy the Rankine-Hugoniot jump conditions.
So in this section we first use the divergence theorem to derive the
physical principles expressed by the first two Euler equations (1), (2).
When p = p(ρ), this stands on its own. We next derive the continuum
version of conservation of energy expressed by the energy equation (3).
We then introduce the second law of thermodynamics together with
enough fluid dynamics to derive (4) and interpret it as telling us that
entropy does not increase when shocks are not present. In the final
section we describe the fundamental polytropic equation of state that
connects the pressure p to the energy E and entropy S for noninter-
acting gas of molecules. The polytropic equation of state describes a
gas of identical molecules each consisting of r atoms.
The compressible Euler equations with polytropic equation of state
are a fundamental set of equations. Every term and every constant
in the compressible Euler equations with polytropic equation of state
is derivable from first principles. Nothing is phenomenological (like a
constant whose value is determined by an experiment) or ad hoc (like
a term added or a value assigned to make a numerical experiment fit
the data). For this reason they are fundamental to Applied Mathemat-
ics, Physics and Fluid Mechanics, and they provide the main physical
setting for the Mathematical Theory of Shock Waves. A student who
learns this has the opportunity to connect up thermodynamics, fluid
mechanics, physics, and PDE’s in a unified, self-contained, fundamen-
tal theory. The gain is well worth the effort!

2. The Mass and Momentum Equations

To start, restrict attention first to the first two Euler equations equa-
tions (1) and (2). To derive the physical principles that underly the
equations, choose any fixed volume V, integrate the equations (1), (2)
 3

over V, and apply the divergence theorem. Start with (1), the so called
continuity equation:

Z Z Z Z Z Z Z Z Z
0= ρt + Div(ρu) dV = ρt dV + Div(ρu) dV.
V V V
(5)

Now in the first integral, since the integration is over x not t, we claim
we can pass the partial derivative with respect to t out through the
integral sign to get a regular derivative on the outside,

Z Z Z Z Z Z
d
0= ρt dV = ρ dV. (6)
V dt V

To verify this, use the definition of derivative directly: the definition of

Z Z Z
d
ρ dV
dt V

leads directly to

Z Z Z Z Z Z 
1
lim ρ(x, t + ∆t) dVx − ρ(x, t + ∆t) dVx
∆t→0 ∆t V V
Z Z Z
ρ(x, t + ∆t) − ρ(x, t)
= lim dVx
∆t→0 V ∆t
Z Z Z
= ρt (x, t) dVx .
V

To the second integral in (7) we apply the Divergence theorem to ob-

tain:
Z Z Z Z Z
Div(ρu) dV = ρu · n dA. (7)
V ∂V

Putting these together we obtain:

4

� � � � �
d
ρ dv = − ρu · n dA
dt V ∂V

Total Mass Mass Flux

Inside V Out Through ∂V

ρu
n

∆Ai
V
ρu

∂V
n

Conservation of Mass: The time rate of change of the mass in

volume V equals minus the mass flux through the boundary ∂V

Conclude: The continuity equation implies that the total rate of

change of mass inside any volume is equal to minus the flux of mass
out through the boundary, a precise expression of conservation of mass
for that volume. Since the volume V is arbitrary, it follows that mass
is conserved in every volume, hence mass is conserved. But reversing
the steps, we have a derivation the continuity equation as an expression
of conservation of mass. That is, we can start by defining the physical
principle of conservation of momentum as meaning precisely that “the
 5

time rate of change of the total mass in every volume is equal to minus
the flux of mass through the boundary” in the sense that (7) holds
for every volume V , and from this, using the divergence theorem in
reverse, we can conclude that (7) holds in every volume, and from this
(using that the integral of a continuous function is zero in every volume
iff the function is identically zero) we can conclude that conservation
of mass in this sense implies the continuity equation (1). We can now
take this reversed argument as a physical derivation of (1) from a pre-
cise physical expression of the principle of conservation of mass in a
continuous media. (Of course, the derivation technically is only valid
for continuous functions, so shock waves are another matter!)

Consider next the second Euler equation (2),

(ρui )t + Div(ρui u + pei ) = 0,
which is really three equations, one for each i = 1, 2, 3. We now use the
divergence theorem to show that it implies conservation of momentum
in every volume. That is, we show that the time rate of change of
momentum in each volume is minus the flux through the boundary
minus the work done on the boundary by the pressure forces. This is
the physical expression of Newton’s force law for a continuous medium.
For this, integrate (2) over a fixed volume V to obtain
Z Z Z Z Z Z Z Z Z
i i
0= (ρu )t dV + Div(ρu u) dV + Div(p ei ) dV.
V V V
(8)
Taking the time derivative out through the first integral and applying
the divergence theorem to the second two integrals (as we did for the
continuity equation) we obtain
Z Z Z Z Z Z Z
d i i
(ρu ) = − (ρu )u · n dA − p ei · n dA.
dt V ∂V ∂V
(9)
Here ρui u is the i-momentum flux vector, with dimensions
mass × velocity distance i − momentum
(ρui ) · u = · = ,
volume time area time
so minus the first integral on the right hand side of (9) integrates over
the area to give minus the rate at which i-momentum is passing out
through the boundary ∂V of V. Similarly, for the second integral on the
right hand side of (9), the dimensions of p is force per area, ei · n = ni
equals the i’th component of the normal, so minus the second integral
on the right hand side of (9) integrates over the area to give minus the
6

total i-component of the pressure force on the boundary ∂V. Putting

these together we obtain:

� � � � � � �
d i i
ρu dv = − (ρu )u · n dA − p ei · n dA
dt V ∂V ∂V

Total i-Momentum i-Momentum Flux Total i-Component

Inside V Out Through ∂V of the Pressure Force
Exerted on ∂V

ρui u
n

∆Ai
V
ρui u

∂V
n

Conservation of Momentum: The time rate of change of Mo-

mentum in a given volume V is minus the i-momentum flux out
through the boundary ∂V minus the total force of the pressure ex-
erted on ∂V.

Conclude: The i-momentum equation implies that the total rate of

change of i-momentum inside any volume is equal to minus the flux of
i-momentum out through the boundary minus the net force due to the
pressure exerted on the boundary, a precise expression of conservation
of i-momentum for that volume. Note that the continuum version of
Newton’s force law F = ma appears as an integrated form of F = d(mv)
dt
,
 7

that is, on the LHS F is replaced by the net force of the pressure on
the boundary ∂V, and the right hand side is replaced but the total
time rate of change of i-momentum in V. Since the volume V is arbi-
trary, it follows that the balance of i-momentum holds in every volume,
i = 1, 2, 3, and we say that momentum is conserved. Again, we can
reverse the steps, and start by defining the physical principle of con-
servation of momentum as meaning precisely that “the time rate of
change of momentum in every volume is equal to minus the flux of mo-
mentum through the boundary minus the force due to the pressure on
the boundary” in the sense that (9) holds for every volume V , and from
this, using the divergence theorem in reverse, we can conclude that (8)
holds in every volume, and from this (using again that the integral of
a continuous function is zero in every volume iff the function is identi-
cally zero) we can conclude that the momentum equations (2) for each
i = 1, 2, 3, follows from the principle of conservation of momentum as
expressed in (9). We can now take this reversed argument as a physical
derivation of (2) from a precise physical expression for the principle of
conservation of momentum in a continuous media. (Again, the deriva-
tion technically is only valid for continuous functions, so shock waves
are another matter!)

3. The energy equation

To complete the theory of the compressible Euler equations to the case
when p is not a function of the density alone, we consider finally the
energy equation, the fifth and last equation in the Euler system. The
equation couples to (1), (2) through the pressure when the pressure
depends on the density as well as the temperature, specific energy e or
specific entropy s, say p = p(ρ, s). The energy equation is
Et + Div((E + p)u) = 0, (10)
where
1
E = ρe + ρ|u|2 (11)
2
is the total energy per volume, the sum of the internal energy per vol-
ume ρe (e=e(ρ, s)=specific internal energy=the energy per mass stored
in the vibrations of the molecules so that ρe is internal energy per vol-
ume), and the kinetic energy 12 ρ|u|2 , the kinetic energy due to the
motion of the fluid particles. To finish we use the divergence theorem
to show that (10) implies conservation of energy in every volume. That
is, we show that the time rate of change of energy in each volume is
minus the flux of energy out through the boundary minus the net rate
8

at which work is done by the pressure force acting on the boundary.

This is the continuum version of Newton’s principle that the time rate
of change of work should be equal to the energy change per time. That
is, if F is a classical force, then
W ork = F orce × Displacement
gives
   
d d d dv dv ds d 1 2 d
Work = F·ds = m ds = m = mv = Energy
dt dt dt dt dt dt dt 2 dt
but in the continuum case, the work per time done by the pressure
force can be stored in the internal energy of vibration as well as in the
kinetic energy of motion. To see this, integrate (10) over a fixed volume
V to obtain
Z Z Z Z Z Z Z Z Z
0= Et dV + Div(Eu) dV + Div(p u) dV.
V V V
(12)
Taking the time derivative out through the first integral and applying
the divergence theorem to the second two integrals (as we did for the
momentum equation) we obtain
Z Z Z Z Z Z Z
d
E=− Eu · n dA − pu · n dA.
dt V ∂V ∂V
(13)
Now Eu is the Energy flux vector, with dimensions
energy distance energy
Eu = · = ,
volume time area time
so minus the first integral on the right hand side of (9) integrates over
the area to give minus the rate at which Energy is passing out through
the boundary ∂V of V. Consider then the second integral on the right
hand side of (13). The dimensions of p are force per area (as always),
so pu has dimensions
f orce distance f orce × distance work
pu = · = = ,
area time area × time area time
and so minus the second integral on the right hand side of (9) integrates
over the area to give the total work per time done by the pressure force
acting on the boundary ∂V of V. Putting these together we obtain:
 9

� � � � � � �
d
E dv = − Eu · n dA − pu · n dA
dt V ∂V ∂V

Total Energy Energy Flux Rate of Work Done

Inside V Out Through ∂V by Pressure Force
On ∂V

Eu
n

∆Ai
V
Eu

∂V
n
Conservation of Energy: The time rate of change of
energy in each volume is minus the flux of energy out
through the boundary minus the net rate at which work
is done by the pressure force acting on the boundary.
Conclude: The energy equation implies that the total rate of change
of Energy inside any volume is equal to minus the flux of Energy out
through the boundary minus the net rate at which work is done by the
pressure force on the boundary, a precise expression of conservation of
energy for that volume. Note that the continuum version of Newton’s
power law F · u = dW dt
appears with F · u replaced by pn · u where pn
is the force per area, implying pn · u is the work per area time, so that
integrating out the area gives the work per time. Since work is energy,
10

work per time appropriately contributes to balance the time rate of

change of the energy expressed in the other two terms in (13). Thus
the second term on the RHS of (13) represents the pressure forces
doing work on the fluid, and the balance of energy law (13) tells us
that the last Euler equation (10) expresses that the pressure is the
only force doing work in the fluid, thereby completing the continuum
version of Newton’s law of energy conservation. Since the volume V is
arbitrary, it follows that the balance of Energy holds in every volume
as a consequence of (10), and we say that energy is conserved. Again,
we can reverse the steps, and start by defining the physical principle
of conservation of energy as meaning precisely that “the time rate of
change of energy in every fixed volume is equal to minus the flux of
energy out through the boundary minus the rate at which the pressure
force does work on the boundary” in the sense that (13) holds for every
volume V , and from this, using the divergence theorem in reverse, we
can conclude that (10) holds in every volume. From this (using again
that the integral of a continuous function is zero in every volume iff the
function is identically zero) we can conclude that the energy equation
(2) for each i = 1, 2, 3, follows from the principle of conservation of
energy as expressed in (13). We can now take this reversed argument
as a physical derivation of (10) from a precise physical expression for the
principle of conservation of energy for a continuous media. (Again, the
derivation technically is only valid for continuous functions, so shock
waves are another matter!)
Finally, it is interesting that although the derivation of the equations
is extremely interesting for understanding how Newton’s laws correctly
extend to a continuous medium, the derivation has helped us almost
not at all in understanding the mathematical theory of the evolution
of the compressible Euler equations that express these laws.

4. The material derivative and the entropy equation

Entropy enters the theory of the compressible Euler equations through
the Second Law of Thermodynamics. In this section we derive the
entropy equation (4) using the collection of results we already have,
together with the Second Law of Thermodynamics. Entropy plays a
fundamental role in the theory of shock waves because shock waves
introduce loss of information and increase of entropy, and entropy con-
siderations are required to pick out the physical vs the unphysical shock
waves. The specific entropy s, the entropy per mass of the fluid, is the
fifth thermodynamic variable that can be taken in place of any one of
ρ, p, e and T . The starting assumption of thermodynamics is that all
 11

of these five state variables can be written as a function of any two of

them. For example, the pressure is often taken to be a function of the
specific volume v and entropy s, p = p(v, s), v = 1/ρ.
The entropy as a state variable enters through the Second Law of
Thermodynamics, which for our purposes states simply that
de = T ds − pdv (14)
is an exact differential, or, solving for ds, that
de pdv
ds = + (15)
T T
is an exact differential. This (15) is exact means no more or less than
that the right hand side comes from the gradient of a function, which
means there is a function s(e, v) such that
∂s ∂s
= 1/T, = p/T. (16)
∂e ∂v
That is, recall from vector calculus that the line integral of a differential
like de
T
+ pdv
T
is independent of path if and only if the differential is
exact, and T + pdv
de
T
is exact just means there exists a function s(e, v)
such that (16) holds. Now a class in thermodynamics would show by
Carnot cycles that if de T
+ pdv
T
were not exact, then line integrals around
closed curves would not all be zero, and by this one could construct
a Carnot cycle around which energy would be created from nothing.
Thus, that de
T
+ pdv
T
is exact is equivalent to saying there are no perpetual
motion machines. But for us, we can start by taking the second law as
simply saying that ds = de T
+ pdv
T
is an exact differential, or equivalently,
solving for ds. To derive (4) from the second law, we first introduce
the material derivative.
Definition 1. Let u(x, t) be a given velocity field, and let f (x, t) be any
other function of x = (x1 , x2 , x3 ) and t. Then the material derivative
of f associated with velocity u is defined to be
Df
= ft (x, t) + ∇x f (x, t). (17)
Dt
The material derivative represents the derivative of f along the particle
path. That is, if the curve x(t) is a particle path in the sense that it
solves the ODE
ẋ = u(x(t), t),
then
d Df
f (x(t), t) = ∇x f · ẋ + ft = .
dt Dt
12

The main identity used to connect up the material derivative with the
compressible Euler equations is the following (easily verified)
Div(f u) = f Div(u) + ∇f · u, (18)
which holds for any vector field u. Using these we have the following
useful theorem:
Theorem 2. Assume ρ = ρ(x, t) and u = u(x, t) solve the the conti-
nuity equation (1), and let f = f (x, t) be any smooth function. Then
Df
(ρf )t + Div(ρf u) = ρ . (19)
Dt
Proof: Using the product rule for partial derivatives on the first term
and (18) on the second term on left of (19) gives
Df
(ρf )t + Div(ρf u) = [ρt + Div(ρu)]f + ρ (ft + ∇f · u) = ρ
, (20)
Dt
as claimed, where the first term vanishes by the continuity equation.


Theorem 3. For smooth solutions, the continuity equation (1) is equiv-

alent to either of the following two equations:
1 Dρ
= −Div(u), (21)
ρ Dt
or
1 Dv
= Div(u). (22)
v Dt
and the momentum equation (2) is equivalent to
Du
ρ = −∇p. (23)
Dt
Note that (22) gives meaning to the Divergence of a vector field: Namely,
the divergence of a vector field gives the rate at which volumes change
per volume along the flow of the vector field. Equation (24) gives a
continuum version of Newton’s force law: Namely, “mass times accel-
eration along particle paths is equal to the gradient of the pressure”,
telling us that only the gradient of the pressure contributes to changes
in the momentum.
Proof: For (21) use (19) to write
Dρ
0 = ρt + Div(ρu) = ρt + ∇ρ · u + ρDiv(u) = + ρDiv(u)
Dt
 13

which gives (21). To obtain (22) from this write

Dv D ρ1 1 Dρ 1
= =− 2 = Div(u) = vDiv(u).
Dt Dt ρ Dt ρ
This is a complete proof. 
Theorem 4. Assuming the continuity equation (1), the momentum
equation (2) is equivalent to
Du
ρ = −∇p. (24)
Dt

Proof: Write (2) in the vector form

(ρu)t + Div(ρuuT + pI) = 0,
where I is the 3 × 3 identity matrix and treating u as a column vector
and its transpose uT as a row vector, uuT is the 3 × 3 rank-1 matrix
with row column entries (ui uj ). Then distributing the Div gives
Du
0 = (ρu)t + Div(ρuuT ) + Div(pI) = ρ + ∇p,
Dt
where we have applied (19) and rewritten DivpI = ∇p. This confirms
(24). 
Theorem 5. Assuming the continuity equation (1) together with the
momentum equation (2), the energy equation (3) is equivalent to the
equation
De Dv
= −p = 0. (25)
Dt Dt

Proof: Now by (11) and (18),

DE De Du
Et + Div(Eu) = ρ =ρ + ρu , (26)
Dt Dt Dt
and
Div(pu) = ∇p · u + pDiv(u).
Putting these together and using (24) to cancel the ∇p · u term gives
De 1 Dv
ρ = −pDiv(u) = −p ,
Dt v Dt
which readily gives (26). 
14

Theorem 6. Assuming the continuity equation (1) together with the

momentum equation (2), the energy equation (3) is equivalent to the
entropy equation (4), namely,
St + Div(Su) = 0, (27)
which is equivalent to
Ds
=0 (28)
Dt
Proof: Starting with the Second Law in the form (15),
T ds = de + pdv,
it is not difficult to show that
Ds De Dv
T = +p = 0,
Dt Dt Dt
where we have applied the energy equation in the form (26). This
establishes (28). But since S = ρs is the entropy density, we can write
Ds
St + Div(Su) = (ρs)t + Div(ρsu) = ρ = 0,
Dt
as claimed in (28). 

5. The compressible Euler Equations with Polytropic

Equation of State
We end this section by summarizing the full compressible Euler equa-
tions for a polytropic equation of state. The complete system of com-
pressible Euler equations takes the following conservation form:
ρt + Div(ρu = 0 (29)
(ρu )t + Div(ρui u + pei ) = 0,
i
(30)
Et + Div[(E + p)u] = 0, (31)
with
1
E = ρe + u2 , (32)
2
u2 ≡ |u|2 = (u1 )2 + (u2 )2 + (u3 )2 .
The system (29)-(30) is a system of five equations in the six unknown
functions of (x, t)
u1 , u2 , u3 , ρ, p, e,
consisting of the three components of the velocity vector u = (u1 , u2 , u3 )
(we use superscript indices to be consistent with Einstein summation
convention whereby vector components are alway up, c.f. [?]), and the
 15

the remaining three variables, the density ρ, pressure p and specific in-
ternal energy e are the so called thermodynamic variables. The other
two important thermodynamic variables in shock wave theory are the
temperature T and the specific entropy s (we’ll discuss the entropy in
the next section). A principle of thermodynamics is that all of the five
thermodynamic variables can be expressed as a function of any two
of them. Since by any choice of the two independent thermodynamic
variables, there remain six unknowns and five equations, an equation
of state which gives the pressure in terms of two other thermodynamic
variables, must be given to reduce the number of unknowns by one and
thereby close the equations.
The most fundamental equation of state is the so called polytropic equa-
tion of state given by
p = p(ρ, e) = (γ − 1)ρe, (33)
where γ is the so called adiabatic constant of the gas. Writing p =
p(ρ, e) in (29)-(31) closes the compressible Euler equations into a sys-
tem of five equations in the five unknowns
(u1 , u2 , u3 , ρ, e).
To write the system in the form of a system of conservation laws
Ut + f (U )x = 0,
define the conserved quantities
U = (ρ, G, E),
using
G = (G1 , G2 , G3 ) = (ρu1 , ρu2 , ρu3 ) = ρu,
and find expressions for (p, e) in terms of U . For example, by (32),
 2
E 1 2 U5 1  G 
e = − |u| = − ,
ρ 2 U1 2  U1 
and so  2 !
U5 1  G 
p = (γ − 1)ρe = (γ − 1)U1 − .
U1 2  U1 
A final word on the adiabatic gas constant. The polytropic equation of
state describes a gas of identical molecules each consisting of r atoms.
In this case, assuming the ideal gas law
pv = RT, (34)
with v = 1/ρ the specific volume and R the universal gas constant,
together with the assumption that the internal energy e distributes
16

equally among all the vibrational degrees of freedom, leads logically to

a dervivation of (33) together with the relations, c.f. [?].
R
e= T, (35)
γ−1

2
γ =1+ . (36)
3r
In particular, for a polytropic gas (33), equation (35) tells us that the
internal energy is proportional to the temperature, with proportionality
constant CV = R/(γ − 1) called the specific heat at constant volume,
(the heat required to raise a unit mass one degree); and (36) gives the
adiabatic gas constant as a function of the number of atoms r in the
gas molecules. Taking the limits r = 1 and r → ∞ gives the bounds
5
1<γ≤ ,
3
the value γ = 5/3 applying to a mono-tonic gas, and γ → 1 in the
limit of very heavy molecules. In particular, air is mostly Nitrogen N2 ,
giving a value of gamma equal to
γ = 4/3.
Conclude: Every term and constant in the compressible Euler equa-
tions with polytropic equation of state is derivable from first principles.
Nothing is phenomenological (like a constant whose value is determined
by an experiment) or ad hoc (like a term added or a value assigned to
make a numerical experiment fit the data). For this reason the com-
pressible Euler equations with polytropic equation of state are a fun-
damental set of equations for Applied Mathematics and Physics, they
anchor the subject of PDE’s by marking the starting point for Fluid
Mechanics, and as such they provide the main physical setting for the
Mathematical Theory of Shock Waves.

6. Entropy of a Polytropic gas

We now show how to integrate the second law to find a formula for
s as a function of v and T , and thereby show how entropy enters the
formulas for a polytropic gas. By this we can find s as a function of
any other two thermodynamical variables among ρ, p, e, T .
Theorem 7. Assume (33)-(36) for a polytropic gas, and assume (14)
is an exact differential, so (16) holds. Then
s = cv ln v γ−1 T . (37)
 17

Proof: To integrate (14), (meaning to find a function e(s, v) that meets

conditions (16)), we introduce a clever change of variables. For this,
define what has come to be known as the free energy.
ψ = e − sT.
For us the free energy is important simply because we can solve the
second law when expressed in terms of ψ. For this purpose, take dif-
ferentials on both sides to obtain
dψ = de − sdT − T ds.
Using the second law T ds = de + pdv, continue
dψ = de − sdT − T ds = de − sdT − de − pdv = −sdT − pdv.
or
dψ = −sdT − pdv. (38)
We now solve for the value of ψ that makes (38) exact, namely, we find
ψ such that
∂ ∂
ψ(T, v) = −s, and ψ(T, v) = −p. (39)
∂T ∂v
But using the idea gas law (34)
RT
p(T, v) = ,
v
the second equation in (39) becomes
∂ RT
ψ(T, v) = − .
∂v v
Remarkably, holding we can anti-differentiate this with respect to v
holding T fixed to obtain
ψ(T, v) = −RT ln v + g(T ),
where g(T ) is the arbitrary function of integration. To determine g(T ),
use the first equation in (39) to write
∂
s=− ψ(T, v) = − (R ln v + g 0 (T )) . (40)
∂T
But by the definition of ψ we also have
cv T = e = ψ(T, v) + sT = −RT ln v + g(T ) + RT ln v − T g 0 (T ), (41)
the latter two terms coming from
∂ψ
sT = −T .
∂T
18

Canceling the first and third term on the RHS of (41) gives
cV T = g(T ) − T g 0 (T ),
which upon differentiating both sides with respect to T gives
cv = −T g 00 (T ).
By this we obtain a formula for g 00 (T ), namely
cv
g 00 (T ) = − ,
T
which upon integrating once gives
g 0 (T ) = −cv ln T + const.
Using this in (40), and setting const. equal to zero, (only changes in
entropy are measurable anyway), we obtain
s = cv ln (v γ−1 T ),
as claimed. 
Finally, solving s = cv ln (v γ−1 T ) for T gives
s
T = v 1−γ exp ( );
cv
and using this in e = cv T yields the formula
1 s
e = cv γ−1 exp ( ) ≡ e(s, v).
v cv
This together with the second law gives the equation of state of a
polytropic gas in terms of (s, v):
∂e 1 s
p = − (s, v) = cv (γ − 1) γ exp ( ) ≡ p(s, v).
∂v v cv
This is the form of the equation of state of a polytropic gas often
quoted in the literature. In particular, replacing v = 1/ρ gives primar-
ily because the speed of sound c is given by the formula
s
p(ρ, s) = cv (γ − 1)ργ exp ( ),
cv
and it turns out the correct generalization of the speed of sound σ when
p depends on s as well as ρ is
s
∂p
σ= (ρ, s).
∂ρ
We could obtain this by linearizing the equations just like we did for
the barotropic equation of state p = p(ρ) before.
The following important theorem gives the equation for the entropy:
 19

Theorem 8. Assume the compressible Euler equations with polytropic

equation of state. Then the entropy satifies the additional conservation
law
St + Div(Su) = 0,
where
S = ρs
is the entropy per volume, or entropy density. (Note that multiplica-
tion by ρ = mass/vol converts s=specific entropy=entropy/mass to
S=entropy/vol.)

Proof: Write
(ρs)t + Div(ρsu) = ρt s + ρst + sDiv(ρu) + ρ∇s · u
= s(ρt + div(ρu) + ρ

Conclude: The entropy is the state variable that is defined by assum-

ing the second law de = T ds − pdv is an exact differential.

****************************************************

7. Derivation of the Rankine-Hugoniot Jump Conditions

In this section we apply the Divergence Theorem to derive the Rankine-
Hugoniot (RH) jump conditions, which we already used to derive the
shock curves of the p-system. So consider a system of conservation laws
Ut + f (U )x = 0, (42)
where U and f (U ) are vectors in Rn ,
U = (U1 , ..., Un ),
f (U ) = (f1 (U ), ..., fn (U )).
The RH-jump conditions apply to solutions U (x, t) that are smooth
on either side a curve (x(t), t) in the (x, t)-plane, but suffers a jump
discontinuity across the curve. So assume the shock curve (x(t), t) has
finite speed s = ẋ(t), and assume that U (x, t) = UL (x, t) to the left
of this curve x < x(t), and U (x, t) = UR (x, t) to the right of this
curve x > x(t), where UL and UR are smooth functions that solve the
conservation law (42). Now U cannot satisfy (42) on the shock curve
because it suffers a jump discontinuity there, so the derivatives are not
defined. Even so, the RH condition says that the jump in a solution
20

across a shock wave must be related to the speed of the shock by the
condition
s[U ] = [f (U )]. (43)
Here, [·] around a quantity denotes the jump in that quantity from left
to right across the shock wave, so [U ] = UR − UL , [f (U )] = f (UR ) −
f (UL ), etc. Recall that for the p-system, U = (v, u), and for fixed state
UL and constant s, the jump condition was satisfied by two curves
S1 (UL ) and S2 (UL ) that had C 2 contact with the rarefaction curves
R+ +
1 (UL ) and R2 (UL ) at UL . We then defined the wave curves Wi (UL ) =
S1− (UL ) ∪ S1− (UL ), i = 1, 1, consisting the the states UR that could be
connected to UL by a shock wave of speed s = s(UL , UR ) determined by
the analysis. So the question remains: where did the Rankine-Hugoniot
jump conditions come from?

We now derive the RH conditions starting with the notion of a weak or

distributional solution of the conservation laws (42). In words, we show
that any solution that solves (42) on either side of a shock curve (x(t), t)
will be a weak solution if and only if the RH conditions (43) hold. To
start then we must define the notion of weak solution. For this, we look
to construct a condition that is equivalent to (42) for smooth solutions,
but also applies to solutions that have jump discontinuities. The idea
is to multiply the equation through by a smooth “test function” φ(x, t),
integrate over any region, and integrate by parts to get the derivatives
off the unknown functions u and f (u), and onto the test function. The
condition for a weak solution is then the condition that this integrated
equation holds for all smooth test functions, and all three dimensional
regions we integrate over. Now when we integrate by parts, there will be
boundary terms, and to make these vanish, we assume the test function
and all of its derivatives vanish outside the volume we integrate over.
To make this precise, define:
Definition 9. The support of a function φ(x, t), denoted Suppφ, is the
set of all values of (x, t) where the function φ is nonzero.
Thus outside of any set that contains the support of φ, φ ≡ 0, and
so phi and all derivatives of φ vanish in the complement of that set.
We use this to define a test function.
Definition 10. A test function φ(x, t) is a smooth solution such that
Suppφ is contained within a bounded set in (x, t) in −∞ < x < ∞,
t > 0, that is some positive distance from t = 0.
by smooth we mean that a test function φ can be differentiated any
number of times, and the support condition is sufficient to guarantee
 21

that the function, together with all of its derivatives, vanish on the
boundary of any set in −∞ < x < ∞, t > 0 that contains Suppφ.
To get a condition for shock wave solutions u(x, t), multiply equation
(42) by a test function φ(x, t),
ut φ + f (u)x φ = 0,
integrate over (x, t)
Z ∞Z ∞ Z Z
ut φ + f (u)x φ dxdt = ut φ + f (u)x φ dxdt, (44)
0 −∞ K
where we used the fact that φ(x, t) vanishes for (x, t) on the boundary
of K, because K contains the suppport of φ. Denoting the boundary of
K by ∂K and assuming without loss of generality that K is contained
within −∞ < x < ∞, t > 0, we can integrate (45) by parts to obtain
Z Z Z Z
(ut φ + f (u)x φ) dxdt = (uφ)t + (f φ)x φ dxdt
K K
Z Z
− uφt + f (u)φx dxdt.
K
where we have applied the Leibniz product rule to write the integral
as a divergence, plus the integration by parts term. By the divergence
theorem, the first term reduces to an integral on the boundary ∂K
where φ and all its derivatives vanish, so this term is zero, i.e.,
Z Z Z Z
(ut φ + f (u)x φ) dxdt = Divx,t (f φ, uφ) dxdt
K
ZK
− (f φ, uφ) · n dxdt = 0.
∂K
Thus we conclude that for any test function φ and solution u of (42)
we have
Z ∞Z ∞ Z ∞Z ∞
ut φ + f (u)x φ dxdt = − uφt + f (u)φx dxdt, (45)
0 −∞ 0 −∞
so long as u is smooth enough so that the derivatives ut and f (u)x exist.
For shock wave solutions, the right hand side of (45) makes sense, but
the left hand side of (45) does not.
Definition 11. We call u(x, t) a weak or distributional solution of the
conservation law (42) if
Z ∞Z ∞
uφt + f (u)φx dxdt = 0 (46)
0 −∞
for all smooth test function φ(x, t).
22

The purpose of the section is to to use the Divergence Theorem to

prove the following theorem which, in words, states that a shock wave
solution is a weak solution of (42) if and only if the RH-jump conditions
hold across the shock.
Theorem 12. Assume u(x, t) is a smooth solution of (42) on either
side of a smooth shock curve (x(t), t), but discontinuous across it. Then
u is a weak solution if and only if s[u] = [f ] holds at each point of the
shock. That is, if and only if
ẋ(uR − uL ) = (f (uR ) − f (uL )).
Proof: Assume the φ is a smooth test function with support K, and
assume the values uL (x, t) of u on the left and uR (x, t) on the right of
the shock curve (x(t), t), both solve (42). Let K = KL ∪ KR decompose
K into the part left and right of the shock curve (x(t), t), respectively,
(c.f. Figure 1.) Thus we can write the weak condition as
Z ∞Z ∞ Z Z
uφt + f (u)φx dxdt = uφt + f (u)φx dxdt
0 −∞ K
Z Z Z Z
= uL φt + f (uL )φx dxdt + uR φt + f (uR )φx dxdt, (47)
KL KR
where uL and uR are smooth in K. Thus we can apply the divergence
theorem in the derivation of the weak conditions in reverse to get the
derivative back onto u and f (u), and then apply (42). But the bound-
ary condition at the shock, where φ need not vanish, will produce the
RH condition. Starting on the left, we get
Z Z Z Z
uL φt + f (uL )φx dxdt = (uL φ)t + (f (uL )φ)x dxdt
KL KL
Z Z
− (uL )t φ + f (uL )x φ dxdt.
KL
But by the divergence theorem,
Z Z Z
−−−−−→
(uL φ)t + (f (uL )φ)x dxdt = (fL , uL ) · nL φ ds, (48)
KL Γ
and Z Z
ut φ + f (u)x φ dxdt = 0,
KL

because u solves (42) in KL , so we obtain

Z Z Z
−−−−−→
uL φt + f (uL )φx dxdt = (fL , uL ) · nL φ ds,
KL Γ
 23

where Γ dentoes the shock curve (x(t), t). Note that ut + f (u)x makes
sense and vanishes in KL and KR separately because we don’t have to
take derivatives at the shock itself. (We can take the derivatives all
the way up to the shock curve by continuity of u and its derivatives on
either side of the shock.) Note that the boundary of KL where φ need
not vanish is exactly the shock curve (x(t), t), which we have denoted
by Γ, and nL denotes the outer normal to KL .
Similarly on the right side K of the shock curve we obtain
Z Z Z
−−−−−→
uR φt + f (uR )φx dxdt = (fR , uR ) · nR φ ds.
Γ Γ
Putting (49) and (49) together we obtain
Z Z
−−−−−→ −−−−−→
(fL , uL ) · nL φ ds + (fR , uR ) · nR φ ds
Γ Γ
Z
−−−−−−−−−−−−−→
= (fR − fL , uR − uL ) · nR φ ds = 0, (49)
Γ
where we used that nL = −nR . But on the shock curve (x(t), t) the
normal vector nR = −i + ẋj. That is, the parametrization of the shock
curve Γ with respect to t is
r(t) = x(t)i + tj,
so
v0 (t) = ẋi + j,
and the unit tangent vector T = |v|/|v| is
ẋi + j
T= √ .
ẋ2 + 1
Hence the outer normal nR to γ is
−i + ẋj
nR = √ .
ẋ2 + 1
Using this in (50) gives the result
Z
{s[u] + [f ]} φ ds = 0,
Γ
The result, then, is that a weak solution u(x, t) of the conservation law
(42) that is a smooth solution on either side of a shock curve, must
satisfy (50) for every smooth test function φ. It follows that we must
have s[u] + [f ] = 0 at each point of the shock, for if it were nonzero at
some point on the shock curve, then we could cook up a test function
with support near that point such that (50) was nonzero. Conversely,
if s[u] + [f ] = 0 all along the shock curve, and u is a strong solution on
24

either side, then (50) is zero, and hence working backwards we would
find that u is a weak solution as well. This completes the proof the
theorem, and the derivation of the Rankine-Hugoniot jump conditions.