Volume 1
Edited by C. Truesdell
CoEditors: L. Collatz . G. Fichera
M. Fixman . P. Germain . J. Keller . A. Seeger
Linearized Analysis of
One  Dimensional
Magnetohydrodynamic Flows
Roy M. Gundersen
SpringerVerlag
Berlin. Gottingen . Heidelberg. New York
1964
Dr. Roy M. Gundersen
Professor of Mathematics
University of WisconsinMilwaukee
TitelNr. 6729
Dedicated to my Parents
Preface
Magnetohydrodynamics is concerned with the motion of electrically
conducting fluids in the presence of electric or magnetic fields. Un
fortunately, the subject has a rather poorly developed experimental
basis and because of the difficulties inherent in carrying out controlled
laboratory experiments, the theoretical developments, in large measure,
have been concerned with finding solutions to rather idealized problems.
This lack of experimental basis need not become, however, a multi
megohm impedance in the line of progress in the development of a
satisfactory scientific theory. While it is true that ultimately a scientific
theory must agree with and, in actuality, predict physical phenomena
with a reasonable degree of accuracy, such a theory must be sanctioned
by its mathematical validity and consistency. Physical phenomena may
be expressed precisely and quite comprehensively through the use of
differential equations, and the equations formulated by LUNDQUIST and
discussed by FRIEDRICHS belong to a class of equations particularly
wellunderstood and extensively studied. This class includes, in fact,
many other eminent members, the solutions of which have led to results
of farreaching scientific and technological application. Frequently, the
mathematical analysis has provided the foundations and guidance
necessary for further developments, and, reciprocally, the physical
problems have provided, in many cases, the impetus for the development
of new mathematical theories which often have evolved to an a priori
unpredictable extent.
In the fall of 1961, the author became seriously interested in the
subject of magnetohydrodynamics, and this interest increased expo
nentially after reading the remarkable paper by FRIEDRICHS. Therein,
FRIEDRICHS showed the amazingly close parallelism between the theory
governing the motion of a compressible fluid whose electrical conduc
tivity may be assumed to be infinite and the theory of conventional gas
dynamics. Specifically, the hydromagnetic equations belong to the class
of symmetric hyperbolic equations. Because these equations are also
reducible, it is possible to develop a theory of shock waves and simple
waves in a manner quite similar to that employed in ordinary gas
dynamics. In this monograph, the author has used FRIEDRICHS' results
as a basis for a more detailed examination of the mathematical structure
of the equations governing a quite specialized class of hydromagnetic
flows; namely, those which may be considered effectively onedimensional
VIII Preface
General Theory
1.1 Introduction
In common with the equations of conventional gas dynamics, the
equations governing the motion of a perfectly conducting compressible
fluid belong to the class of reducible hyperbolic equations. Because of
this structural similarity, many of the techniques which have been
applied so successfully to gas dynamic flows may be extended to magneto
hydrodynamic flows, and, in particular, it is possible to build up a
theory of shock waves and simple waves remarkably parallel to that of
conventional gas dynamics. This was pointed out quite elegantly by
FRIEDRICHS [1] and amplified by others.
The equations of magnetohydrodynamics form a system of first order
nonlinear partial differential equations. Because these equations can be
solved exactly only in certain quite specialized cases, alternative
approaches are mandatory. Whenever problems involving nonuniform
shock waves are under consideration, entropy variations are involved
and the differential equations for nonisentropic flow must be employed.
Although nonisentropic flows are seldom readily amenable to exact
mathematical analysis and the added effects of electrical conductivity
only compound the difficulties, various perturbation techniques have
led to important results and a better understanding of nonisentropic
conventional gas dynamic flows. Therein lies the reason for this investi
gation.
The primary purpose of this monograph is the presentation of a
method for discussing weakly nonisentropic quasionedimensional
flows of an ideal, inviscid, perfectly conducting compressible fluid
subjected to a transverse magnetic field. It is the simplicity of application
and the ease of extension which make the method attractive and, in
fact, many of the solutions to the corresponding problems in con
ventional gas dynamic flows are contained as special cases, and this
provides a valuable check on the theory. The analysis is based on cha
racteristic perturbations developed by PAUL GERMAIN and the author
for discussing similar problems for nonconducting fluids and the
author's extensions of this work to magnetohydrodynamic flows.
Springer Tracts, Vol. 1: Gundersen 1
2 1. General Theory
B t + uB + BUill = 0
Ill (1.2.4)
(1.2.5)
where w, the limiting case of a fast wave, is the true speed of sound, i.e.,
small disturbances propagate with speed W relative to the fluid. It is
most convenient to deal with the characteristic form of the system where
derivatives in only one direction appear in each equation. A complete
derivation is given in Appendix B, and it is seen that for an arbitrary
flow in a uniform duct, the system may be written in the following
characteristic form
Xo  (u + w) t{J = 0 (1.2.6)
x'"  (u w) t", = 0 (1.2.7)
x~ ut€ =0 (1.2.8)
(00 2  c2 ) B{J 2cc{J c2 s{J
B +wu{J+ yt" y(yl)c. =0 (1.2.9)
B€ 2c~ =0 (1.2.11)
B (yl)c
Se= o. (1.2.12)
2wc", _ 0
 U'" + ,'=
(yl) C  •
(1.2.14)
1.2 The Linearized Equations 5
so that
co 2 = c2 [1 + kc'1*]. (1.2.15)
For a monatomic gas, y = 5/3, and Eqs. (1.2.13) and (1.2.14) may be
integrated explicitly to give
U2 + 1
(1'1)
J(1 + kC'1*)1/2 dc = IX (1.2.18)
 U2 + 1
(1'1)
J(1 + kC'1*)1/2 dc = {J (1.2.19)
where use has been made of the relation ~ = (I' ~;) e and Eq. (1.2.4)
omitted. Adding and subtracting Eqs. (1.2.20) and (1.2.21) gives the
6 1. General Theory
')'1 +.!!.!.+
~ 2 (u + w) [~+~]
2 ')'1
uwA z wAt c2 s z
=~2A+ 2')'(')'I)c~ (1.2.22)
~.!!.!.+
')'1 2
(u w) [~+~]
2 ')'1
uwA z wAt c2 s z
=~2A 2')'(')'I)c~ (1.2.23)
St + us", = O. (1.2.24)
The close analogy of this system with the basic equations of conventional
gas dynamics should be noted, viz., a hyperbolic system of three non
linear first order partial differential equations with Riemann invariants
which are linear relations between the dependent variables. This system
includes the gas dynamic equations as the special case of vanishing
magnetic field, i.e., for k = 0, W = c, w = c.
A formal linearization of Eqs. (1.2.22)(1.2.24) in the neighborhood
of a known (isentropic) state, i.e., a uniform or simple wave flow, denoted
by the subscript zero, leads to the following system of linear equations,
which has the same characteristic surfaces as Eqs. (1.2.22)(1.2.24),
for the terms of first order, denoted by the subscript one
Rt + (uo + wo) R", + (~ + WI) (x'"
Uo
2
+~=
')'1
(X
'
~+~=
2 ')'1
P
are the characteristic parameters of the base flow.
In the linearized form of the basic equations, the entropy equation may
be solved directly and independently. From Eq. (1.2.27), it is clear that Sl
remains constant along the particle paths of the base flow, i.e., along
dx/dt = U o' Further, since flo(dx  uodt) is the exact differential of a
function 'Po such that 'Po equated to a constant defines the trajectories,
the solution of Eq. (1.2.27) may be written as
Sl = r('Po)
1.2 The Linearized Equations 7
r
with an arbitrary differentiable function. It is convenient to define a
function To (x, t) by
C~SI'" = c~eor'CI'o) = y(y  1) Cl1 To .
Since it follows from Eq. (1.2.15) that
2tl +WI _ [3 w~ + 22
(I'  2) Cg] R
+ [ (I'  2) c~ 
22
W~] 5
Wo Wo
_ =[W~+(2y)cg]R_[3W&+(Y2)cg]5
UI WI 2w& 2w~·
Rt + (U o+ Wo )R '" + {[ 3W~+(Y2)C~]R
2w~ +
[(Y2)c~wg]5}
2w~ IX",
(1.2.28)
{[wg+(2 y)c&]R_[3wg+(Y2)Cg]5}R
5t + (U o  ) 5
Wo '" + 2w~ 2w& /,,,,
(1.2.29)
Although Eqs. (1.2.28) and (1.2.29) form a system of two coupled equa
tions for the quantities Rand 5, these equations uncouple in each of the
cases to be considered and may be solved separately. Since the equations
are linear, solutions may be superposed. Further, the solutions may be
obtained by solving the linear homogeneous system associated with
Eqs. (1.2.28)  (1.2.29) and then adding a particular solution to the com
plete system in order to include the effects of crosssectional and/or
entropy perturbations.
8 1. General Theory: 1.2. The Linearized Equations
For an initially unifonn flow, IX and {3 are constants, and the general
solution of Eqs. (1.2.28)  (1.2.29) is
Chapter 2
quantity over the crosssectional area is what the hydraulic theory pre
dicts! Thus, it must be possible to bypass CHESTER'S detailed analysis
and the same results must be obtainable by investigating the problem
from the very beginning by a onedimensional formulation. This extremely
important simplification, which provides the basis for many extensions
of the theory, was observed by PAUL GERMAIN and the applicability of a
quasionedimensional analysis for discussing flows in nonuniform ducts
noted in a preliminary presentation given by GERMAIN and GUNDER
SEN [2]. A more detailed presentation, including a derivation of CHESTER'S
results by the use of a purely onedimensional analysis, was given by
GUNDERSEN [5], [6].
A quite unexpected premium was the somewhat surprising result that
the more simple onedimensional theory led to an important improvement;
namely, the term [Lf A] in CHESTER'S work was replaced by ..4, the per
turbed area distribution. Consequently, instead of the pressure pertur
bation being related to the total change in crosssectional area, a direct
first order differential relation between changes in area and shock strength
was valid at any point in the nonuniform section. The usefulness of this
differential relation was exploited by CHIS NELL [9], who was led to it
differently under the assumption that the asymptotic form of CHESTER'S
steady state solution could be employed. CHISNELL integrated the dif
ferential relation in closed form and used the resultant shock strength
area relationship to give an approximate description of the motion of the
shock in terms of the area of the duct. By suitable choices of the area
distribution, a description of converging cylindrical and spherical shocks
was given, and the results checked by comparison with previous simi
larity solutions, valid in the neighborhood of the points of collapse of the
shocks. The comparison showed the remarkable accuracy of CHISNELL'S
work.
Later, WHITHAM [10] discussed these problems and their extensions
and, in particular, rederived CHESTER'S results by the use of a onedimen
sional approach, virtually identical (the only difference being the choice of
dependent variables) to that previously presented by GERMAIN and
GUNDERSEN [2] and GUNDERSEN [5], [6]. ROSCISZEWSKI [11] also re
derived CHESTER'S results, but made the erroneous statement in his paper
that the onedimensional approach to the problem was developed by
WHITHAM. ROSCISZEWSKI limited credit to the present author's contri
bution to: "the influence of a slow perturbation of the otherwise uniform
piston motion of a shock wave" though it seems somewhat curious that
reference was made to the wrong paper, i.e., [12] instead of [6].
The aforementioned theories suffered from the common defect that
they ceased to be valid for nearsonic flow behind the incident shock. By
the use of techniques developed by STOCKER [13], this defect was removed
2.1 Introductory Comments 11
case of nonstationary flow in front of the shock in [22]. Next, it was shown
that the limitation of a monatomic fluid was unnecessary, in spite of
the fact that generalized Riemann invariants could not be determined
explicitly, at least in terms of elementary functions, and the theory of
[20] and [21] was extended to an arbitrary value ofthe adiabatic index [4 J.
The extension of the results of [22] for arbitrary values of the adiabatic
index is contained in the present work as a special case.
Before proceeding to the determination of the propagation of shocks in
nonuniform ducts, it is of value to give a brief discussion of normal
magnetohydrodynamic shock waves and the transition relations relating
the flow parameters in front of and behind the shock. This is given in the
next section.
yPl + v~ + B~ = yP + v~ + I B~ (2.2.4)
(yI) (>t 2 (>tIL (yI)J!. 2 flaIL
(2.2.5)
(2.2.6)
2.2. Transition Relations Across Normal Shocks 13
The basic set to be discussed is Eqs. (2.2.1), (2.2.2), (2.2.5) and (2.2.6).
It is assumed that the state one is completely known and further that the
shock strength is prescribed. It is immediately apparent that
(2.2.7)
Eq. (2.2.7) may be used to express all quantities with sUbscript two in
Eqs. (2.2.5) and (2.2.6) in terms of those with subscript one by the use of
Eq. (2.2.7). This yields
(1 ~)
2
~?l VIa + P [1 c + ym~ (Ia)2]
 2 1
 0
,  (2.2.8)
+
J
C2
= 1+ m~
~~~~
= 1 + ..."._,=....c~,,::==
Oa  1  ym~ (1  a) 3/2
(2.2.13)
w~ (1 + m~)"C (2.2.14)
wi (1 + mil a
Expressing all quantities in Eqs. (2.2.5) and (2.2.6) in terms of those
with subscript two gives
The secular equation for the system of Eqs. (2.2.15) and (2.2.16) gives the
known result m~/mr = a 2/c.
14 2. Shock Propagation in NonUniform Ducts
= {I' 2 1 + m1
2[ I' I' 1 +
(2I')a]}
I'  1
a
(0  a)
(2.2.17)
M2 M~ (2.2.18)
2  aT
Further, there is the result
v~ ~
w~ 1 + m~
v~ ~
w~ l+m~
From Eq. (2.2.1),
U. U  U, + U, M,
C, C, a
so that
n2 = :: = [nl + (1 _ ~) Ml](;) t
or
Finally
u, n,
w, (1 + ml)2
I
u. n.
1
W2
(1 +~)2
Eq. (2.2.10) gives, in fact, the magnetic Hugoniot function for poly
tropic fluids, i.e.,
ym~
P 2(e2 0el) [ el Oe2+2!h ( /?2 )3] P1=0.
le; (2.2.20)
Along the Hugoniot curve, l' varies from 0 to 00, while a varies between the
limits amin = 0 and amax , where amax is determined from the equation
()a  1  Tym 2
(1  a)3 = 0 ;
U. = n 1 (~)* uU t1 + [n 2 _ n1 (.!!.)*] c +i
r] m~[
C, T T C1
.1
{[n n (;)f]
2 1 X
X { naq• .I.
+
(n s qs)A.
+n1 (a)t~+[
T U n2 n1 T
I
(a)t]cII m~
;'0:
C (yI)q. lit
QI}_
f] :~
1
 [n2  n 1 ( ; X
X [{2(0+1) 0/y+2 (0+1) (02) a/y+mHO (1 +0)+(1 + 0) (02)a + (03) (201)a 2 + (30)aB] _
2(0a) (aI) {2(0 + 1)/y + mW + 0 + a(03)]}
m~} ymW  a)3 (4  a)
 qs(yl) (0 + + ym¥(al)3/2 +
1) (aI)
[1 + 0+a(03)]m~
+ 2(0 + I)/y + mW + 0 + a(03)] X
{ (0'1) a + m~ ya[30160a + 3(0 + 1) a s 2a3]/2}]
X (0 + 1) (a 1) + ym~(a 1)3/2
1 _
+
a)2] PI [ymWa)3/2+10a]
[
+ n2n1 ( T
 Ii; (4  a) (0 + I) (Ia) + ym~(Ia)3/2 X
3
{ 20(0+1)/y+2(0+1) (02)a/y+mHO(1 +0) +(1 + 0) (02)a + (03) (20I)aa + (30)a ]
X 2(0a) (aI) {2(0 + 1)/y + mW + 0 + a(03)]} 
m~ }
q.(yl)
Springer Tracts, Vol. 1: Gundersen 2
18 2. Shock Propagation in NonUniform Ducts
2 J 5 u
Fig. 2. A graph of the parameter K (G, m" 0) '" K. vs. G for m, = 0, the ordinary gas dynamic case, and for
m,=2and6
For timeindependent ducts, i.e., 1'2 = 1'2 (x) alone, Eq. (2.3.5) con
tains as a special case the results given [22J for the propagation of a non
uniform hydromagnetic shock wave into a moving monatomic fluid; with
n l = 0 and ¢i = 0, i = 1  5, Eq. (2.3.5) reduces to the solution given [4J for
the propagation of a nonuniform shock wave into a stationary fluid. In the
limit of vanishing magnetic field, Eq. (2.3.5) reduces exactly to the cor
responding solution for the propagation of a conventional gas dynamic
shock wave in a duct with timedependent area variations [7J.
In magnetohydrodynamics, strong shocks can occur in two ways, viz.,
for (1 close to e, i.e., 6 for y = 7/5, or for a very strong applied field for any
(1 > 1, i.e., ml is then large.
For stationary flow in front of the incident shock wave, it is seen that
for y = 7/5 and all ml> lim K ((1, ml> 0) = 0.5 and lim K ((1, ml , 0) = 0.39414,
G>1 + a>6
and these are exactly the limits for the ordinary gas dynamic case which
corresponds to ml = O. Consequently, there is the result:
For very strong or very weak shock waves [in an asymptotic sense], the
propagation of the shock is independent of the applied field and given by
conventional gas dynamic theory.
2.3 Solution behind the Incident Shock 19
~
050
If 5 CT 5
Fig. 3. Agraph of tbe parameter K, vs. a for tn, _ 1,3 and 10
J(.
050
J 5 5 IT
Fig. 4. A graph of the parameter K, vs. a for tn. ~ 1000
but the curves are concave upward with those for greater mllying beneath
those for lesser ml , and all curves issue from the point (G, K) = (1, 0.5)
and terminate at the point (6, 0.39414). Graphs are presented in Figs. 2,
3 and 4.
For fixed incident shock strength G, K (G, ml , 0) decreases monotoni
cally with increasing ml , so that in a diverging (converging) duct, the
pressure decrement (increment) is decreased (decreased) by increasing the
2*
20 2. Shock Propagation in NonUniform Ducts
iia + 2m~ c + Pa = _
i nsql A [ (n. +qa)!]
1:. (yl) q. I:s yp.q. (n. + qs)A. 2 X, n.
_ iia + 2m~ c + PI = _
i nlqz A [x (n.q.) t] +
2 (yl)q. 1:. yp.q. (nsqa)As 2' na
+ ~ [x  (us  co.) t] •
1:.
it follows that:
Pa [71 1 + 71m~
p. 71(71l)q.
+ naql 2A
Az [x, (n2  ql) tlnaJ = G [x  (U 2  W2) t]/C2 .
nl  qz a
Evaluating this on the shock, x = Ut, and replacing P2/ p 2 by its value
from Eq. (2.3.5) evaluates G so that the final solution is:
5
+E Di(a,m1 , n1)CPi[<5A., CA.] +
i=1
where
Oa  1  ym~ (1  a)3/2
"1 = (0 + 1) (aI) + ym~(al)3/2
A= x  (U 2  W2) t
t.
10
(2.4.2)
where <1>4 denotes the first four terms on the righthand side of Eq. (2.4.2)
Tables of Y and Z are given in Appendix C and graphs of Z (0', ?nt, 0) = ~2
and !5 (0', m1 , 0) are given in Figs. 5 and 6.
It should be noted that the expression for the pressure in the reflected
pulse as given by Eq. (2.4.1) or Eq. (2.4.2) is singular for truesonic flow
2.5 Modified Perturbation Theory 23
speed behind the incident shock, i. e., for n 2 = Q2' This singularity masks
the formation of secondary shocks and further comments and a modified
perturbation theory valid for all flow speeds will be given in the next
section.
15
2.0
~
""2 + WI +~~
(y2) c l
[~~~]
+ (u+ro) 2 + w B + (y2) cl
(2.5.4)
St + us., = o. (2.5.5)
(2.5.7)
retained while the remaining terms may be linearized directly. Thus, the
resultant equation is:
n R., _ C~81z
Dt +Q ~~'" + [W22 + (2  y)
2
C2] 
WI

Y ( Y  1) c. +
+ [W~ (Y  2) c~J [A + u A ]
+2A.w. 2t 2 2", •
Substituting from Eqs. (2.5.6) (2.5.7) gives the following equation valid
along the negative characteristics:
[2 2) 2] E' [x  Us t]
n
~~t + ~~n ~~'"
n
= W2 + (y  C2
W2
2 +
(2.5.8)
e. _ 1 Po 1 PI + 1 iiil + 121
e;L;"p'L;p; L; m1 e,' (2.5.10)
where the arguments of the last three terms are the same as in the sum
mation term. For convenience, let Eq. (2.5.11) be rewritten with a new
26 2. Shock Propagation in NonUniform Ducts
so that
E'[xuat] {3 (Ma + n a)Aa.l + {3 (na + ql) As.s +
c~ 1 MsAa 1 M.nsc s A.
(2.5.12)
5
+ X; !5i[~i.I(M2 + n2)/M2 + ~i.2/M2C2]
i=l
where the subscripts after the commas refer to derivatives with respect
to the first or second argument of the pertinent function. Substitution
of Eq. (2.5.12) into Eq. (2.5.8) gives the final differential equation. viz ..
Q +QQ = (q~ +Y 2) Plc~(M. +n l) A ••1 +
t III q~M.As
+ (n2 + q2) (q~ + Yz 2) Pl c• A M + (q~ + Y ;) c~(M. + n.) l;!5 i ~i.l +
qzM.nzA z qa M ! i=l
where the arguments have been omitted in the functions arising from
Eq. (2.5.12). For u 2 not close to £02' Eq. (2.5.13) may be solved directly
and the solution so obtained agrees with Eq. (1.2.31) after a redefinition
of the arbitrary function G, and this leads to the solution presented in
section 2.4.
For u 2 near £0 2, it is a first order nonlinear partial differential equation
with prescribed inhomogeneous term and is to be integrated along the
negative characteristics. The initial condition at the shock is easily
determined. Since
WB = [q~ (2 + 11*)  11*] Cz Cz = ~ +~_ ~
al. 2q~c. 'c. C1 2. 20'
and ii/a and i/T are known on the shock, viz., Eqs. (2.3.5) and (2.5.10),
w.=[q~(2+1):)1)*][~3(Ut,t)_ t/J&(Ut,t) + (IL 1)t/J4(Ut,t) +
ala 2q2 2LI 2LI
5
+ (L~~ 1) (.~ I) i~ Dda, ml , nl ) ~i(Ut, t)] +
+ 1)*) 1)*] (1  + qa) tIn.]
+
[q~ (2
4 zL A
qz 1.
L 1 ) (. 
a
1) K (0', ml> n l ) Aa [Ut, (nz
(2.5.14)
2.6 Criterion for the Particle Velocity to be Unaffected by Entropy Perturbations 27
n
on the shock. Since = u  W = U2  W2 + u 2  W 2 , with W2 given by
Eq. (2.5.14) and U2 by Eq. (2.3.5), the initial condition may be written as:
5 
no _
.:.~  U2  W2
+ ~
~ Qi 'f'i
,I. (Ut t)
, 
Q8 A
A. [Ut, (n. + q.) tIn.] (2.5.15)
i=l •
(W W ) 
[w~ + (1'  2) c~] eoc~T' ('['0)
(2.6.5)
o 1 '"  (1'  1) w~ 2 yc~
dPo = eo [dx  uodt] (2.6.6)
where Po = constant defines the particle paths. Consequently, the problem
is reduced to expressing the compatibility of Eqs. (2.6.4) and (2.6.5), and
the conditions on the functions uo(x, t), Co (x, t) and wo(x, t) which allow
this compatibility will be determined.
The characteristics of Eq. (2.6.4) are
dt dx dW I
T = u;; = [w~ + (1'  2) c~ ] (2.6.7)
 2w~ UO~WI
One first integral of Eq. (2.6.7) is Po = constant, i.e., the particle paths.
Since WI = Wo is a particular solution of Eq. (2.6.4), the following method
of solution is suggested. From Eq. (2.6.7)
dw o
_ [ w~ + (1'  2) cg ] uoz
2 Wo
(2.6.8)
(2.6.14)
30 3. The PistonDriven Shock Wave
Chapter 3
To = 2E' [x  u 2 t] . (3.1.3)
3.1 Arbitrary Area Variations 31
(3.1.4)
(3.1.5)
= i:
v=O
[N 3 c./u 2 q.  N2/u 2]V(I)V ] [t{( _
[N 1 /u 2  c 2 N 3 /u 2 q 2 ]V+1 q2
M )/(q +M )}vJ_
2 2 2
(3.1.6)
00
and
r., _ (Iq~)
1 2(yl)n.q. X
For 1nt = 0, Eq. (3.1.6) reduces to the result given in [19] for the
corresponding problem for nonconducting fluids. The series, Eq. (3.1.6),
converges rapidly, and a few terms suffice for a reasonable approximation.
An alternative approach is to approximate the area perturbation by a
Taylor series, and then the resultant functional equation may be solved
term by term.
(3.2.1)
(3.2.2)
3.2 The PistonDriven Cylindrical Shock Wave 33
The subsidiary conditions are again Eq. (3.1.4), and this allows the deter
mination of F and G in terms of fiT. With the assumption that there is no
perturbation of the piston, the following functional equation is obtained:
monotonicity is lost, but the curves are concave upward with curves for
greater mllying beneath those for lesser~, i.e., for fixed G, K. decreases
monotonically with increasing ml'
1.01
for y = 7/5. The pistondriven cylindrical shock wave is obtained by
letting K = Kc and the resultant function is called Fe (0', ~). The piston
driven spherical shock wave is obtained by letting K = K. and the re
sultant function is calledF.(O', ~). Data are presented for several values
of ~, the measure of the applied field, including the conventional gas
dynamic case which corresponds to ml = O. For convenience of reference,
a table of values of the function F (0', ~) for the case previously derived
[4J, i.e., when there was no influence of a piston, is given in Appendix D.
This function, called Fo(O', ~), is obtained from Eq. (3.4.4) when
K = K (0', ~, 0) as obtained from Eq. (2.3.5).
The tables may be used in several ways, e.g., suppose there is a section
of variable area connecting two ducts of constant but unequal area, say
4. Flows with Heat Addition: 4.1 The Linearized Equations 37
Chapter 4
(4.1.2)
B t + uBx + Bu x = 0 , (4.1.3)
St+usx=/* (4.1.4)
where /* is the rate of entropy production (due to the heat addition).
38 4. Flows with Heat Addition
(4.1.6)
(4.1.7)
(4.1.8)
Approximate solutions to the characteristic system could be obtained by
finitedifference techniques.
To this point, the discussion has been quite general, and the equations
apply to an arbitrary quasionedimensional flow with heat addition. But,
if it is assumed now that the crosssectional area variations and heat
addition are small and may be considered as small perturbations in an
otherwise uniform base flow, it follows from Eq. (4.1. 7) that B jeis constant
along each particle path and for an initially uniform flow, Bje is constant
throughout the flow. Then, Eqs. (4.1.5)(4.1.6) lead to
~+lf wdP = 0(
(4.1.9)
2 2y P
(4.1.13)
(4.1.14)
(4.1.15)
where
~+ WOP1
R1 = 2 2y Po
S~+ WOP1
 2 2y Po
are the first order generalized Riemann invariants written in terms of U 1
and Pl. For simplicity, let = n n(t) alone; then, exactly as in Chapter 1,
the general solution of Eqs. (4.1.13)(4.1.15) may be written as
81 = r(Po) + JIt (t) dt (4.1.16)
R = F [x (u o + Wo ) t] + 2
Wo

/I*()d
1 t t 2A ( + )
UOWOAI(X)
(4.1.17)
YC v 0 Uo Wo
5 = G[x (u o Wo) t] +  Wo
?
/I*()d
1 t t UOWOAI(X)
2A ( _ ) (4.1.18)
~yc. 0 Uo Wo
r
with F, G and arbitrary functions. For the case of no heat addition,
these solutions are equivalent to Eqs. (1.2.30)  (1.2.32), and, in the limit
of vanishing magnetic field, these solutions reduce to solutions of the
corresponding problem in conventional gas dynamics. An illustration of
this latter fact will be given in the next section.
R=2
1 u + 20pl
w P
y 0
W
=2
0
ye.
f n(t)dt+constant (4.2.1)
on advancing characteristics
x  (u o + Wo) t = constant.
4.2 Extension of Stocker's Work 41
s = ~
2
+ 2yP
WOPl = ~flr(t) dt+constant
o 2yc.
(4.2.2)
on receding characteristics
x  (u o wo) t = constant.
In regions where there is no heating, Eqs. (4.2.1)(4.2.2) are valid with
It(t)= O.
The sound speed perturbation may be determined from
c ( P )(1'1)/21'
Co = Po exp [(s so)/2yc,J
which gives
This is the direct extension to the magnetic case of the problem con
sidered by STOCKER, and its solution is obtained in a similar manner. The
(x, t) plane is divided into regions as shown in Fig. 7, which is taken from
STOCKER'S paper [13J. The region contained between the lines x = 0 and
x ~~ ~l is the heating region and Eqs. (4.2.3)(4.2.4) apply therein. Out
side this region R (5) is constant on an advancing (receding) characteristic.
The lines 0 D and L C are the advancing characteristics
x(u+w)t=o and x(u+W)t=~l'
For simplicity the subscript zero has been dropped from the base flow.
The values of Rand 5 throughout the (x, t) plane may be found by the
use of Eqs. (4.2.3)(4.2.4) . Since R(S) is constant along an advancing
°
(receding) characteristic and R = 0,5 = in regions 1 and 15, the values
t
11 N
/ / 0 c
./ /
.I
.9
.J
Fig. 7. The (x, t) diagram for uniform heat addition to an initially uniform flow. Heat is added over the section
extending from 0 to L. The lines OD and LC are advancing characteristics. The lines OA and LB are reo
ceding characteristics. The lines OM and LN are particle paths
In regions 4, 5, 8, 9
In regions 6, 7
In regions 13, 14
R = ~
2
[t + u;1
+w
X] .
In regions 10, 11, 12
R = A1;1
2(u +w)
In regions 11 and 13
51 YAd;1  x + ut)
Co uw
In region 2
U =  A1
 X]
[t +  
I 2 wu'
~=A[t
P 2w + _X_]
wu '
C1 (Yl)A1 X]
c=
[
4w t+ wu '
M
I
= _ A
I
[2W + (y 
4wc
l)u ] [t + _X_]
wu
.
44 4. Flows with Heat Addition
In region 3
~= 2(wu) '
Cl (y  I) Al~l
C 4w(wu) '
PI YAl~l
P 2w(wu) '
M = _ Al~l [2wc + (y  I)u]
1 4wc(wu)
In region 4
u =~[_X_
1 2 u+w
t] '
PI _
p  2w
~Al [t + u+w
X]
'
Cl Al [
c=4'W (y+l)t+(yl)
X
u+w '
]
M
1
=  ~{[2w + (
4wc y
+ l)u]t+ x[(yI)U2W]}.
u +w ,
In region 5
Al [X t] ,
U1 = T U+W
~_A[_X_
P  2w U+W + t] ,
~= (yI)Alt + Atx[2w+(y+l)u]
C 4w 4uw(u+w) '
At { XU(Y+l)}
M 1 =4wc [2w+(yl)u]t+ U+W •
In region 6
u1 = 0,
PI yAtt
P=W'
M 1

 YAtut

2wc·
In region 7
~2
~[t +wu'
X~l]
In regions 8 and 9
A" [2wx  (u + w) ~,]
U1 = 2(W 2 _ U2) ,
P, yA,,[(u+w)~,2ux]
P 2W(W2U 2)
In region 8
2 = ~ + (y 1) ~,A" _ (y 1).A.l ux
C 2w 4w(wu) 2W(W2U 2) '
M = _ UA,lt + A"x[2w 2 + (y 1)u 2] + (y 1)u]
A,'~1[2w
1 2wc 2WC(W2U 2) 4wc(wu)
In region 9
c: ~W(Wl~~, + ~w~:(:.~~~
M = A"x(W 2 +yu 2) _ A,,~,[2w+(yl)u)]
1 2WC(W 2  U 2) 4wc(wu)
In regions 13 and 14
P,
P
=A
2w
[t+ u~'X]
+w
.
46 4. Flows with Heat Addition
In region 13
C1 (y + l)lrt + ,1.1(~1X) [(y + l)u + 2£0]
C 4£0 4lOU(U + £0) ,
M =).t[2lO(Y+l)U]_).(~_X)[ (y+l)u ].
1 1 4lOC 1 1 4lOC(U + £0)
In region 14
~= (yI),1.1 [t ~lX]
C 4£0 + U + £0 '
M  ,1.1[2lO(yl)u]
1 4lOC
[t+ U+lO
~lX].
In region 15
u 1 = 0,
P1 = 0,
c1 = 0,
Ml=O.
The values of N v Wv Tl in any of the above regions may be obtained from
the relations
£01 [q2 +
y  2] c1
w= (yl)q" c'
M(y2) (q2I)c 1
qNl = M1+ (yl)q 2 c
Tl 2c1
T=c
where N = uJw.
In the limit of vanishing magnetic field, these results reduce precisely to
those obtained by STOCKER.
As STOCKER points out, the unsteady effects are confined to the re
gions: (i) 2, which represents a pressure wave propagating upstream;
(ii) 11, which represents an interface propagating downstream; (iii) 14,
which represents a pressure wave propagating downstream. Consequently,
the flow resulting from the heat addition never becomes steady, though
the flow in any finite section asymptotically becomes steady and, in a
sense, the flow in regions 3, 9 and 10 is the final steady flow resulting from
the heating of the initially uniform flow.
°
In regions 3, 9 and 10, c1 > 0, PI > 0, Ml < and Tl > 0, so that an
addition of heat cannot result in a drop in temperature, and an increase in
total heat added to the fluid passing through the heating section cannot
cause a fall in temperature at the exit from this section. Within the frame
work of the linearized analysis, the final states in regions 3 and 9 depend
4.3 Modified Perturbation Theory 47
only on the product Al ~I' i.e., on the total heat added but not directly on
the length over which the addition takes place.
In the heating region 9, ~: < ° if M> q/yt or, equivalently, if
'
N > Y 2. This is because the gas flowing through the heating section
undergoes a pressure drop and the heat loss from the concomitant
expansion may exceed the gain from the external source. In the unsteady
flow regions 7, 8, 10 and 11, ~: < 0, irrespective of the Mach number.
The solution of this section may be used to discuss the case of a duct
of finite length. The solution is identical to that previously obtained until
the advancing or receding waves encounter an end of the duct where
reflection takes place. Again, the solution in the magnetic case follows
directly from STOCKER'S solution for the nonmagnetic case and will not
be included.
4.3 Modified Perturbation Theory
The linearized analysis of section 4.2 led to a solution which was sin
gular for truesonic speed, i.e., for u = w, in the neighborhood of which
the disturbance 5 tends to accumulate and does not propagate through
the channel. Exactly as in Chapter 2, this difficulty is due to the lineari
zation, which approximates the negative characteristics dx/dt = u  W
by the rectilinear family x  (u  w) t = constant, and again a more
detailed examination of the receding characteristics must be made.
From Eqs. (4.2.1)(4.2.2), it is seen that the values of U I and PI
depend on the time taken by a small disturbance to pass through the
heating region. This is true because the function If was assumed constant.
Although the time will always be small for the advancing characteristics,
it becomes large for the receding characteristics when the base flow is
near truesonic. Thus, it may be assumed that the advancing characte
ristics and particle paths are each rectilinear and parallel, viz.,
x  (u + w)t = constant,
x  ut = constant,
The explicit solution for the fifteen regions in the (x, t) plane will not
be given; rather, only the regions 7, 13 and 14 will be considered.
These are the cases considered by STOCKER.
In region 14, it is known that
R= ~
2
[t + u+w'
~lX] S = 0, SI = O.
Substituting Eq. (4.3.3) into Eq. (4.3.2) and carrying out the integration,
there results
x = (u  w) t + Al [q2;q~ + 2] {u ~2W + t~l;WT))} + h (1]) + 0 (Ayt) .
The function of integration h(1]) is not completely arbitrary since by the
definition of 1], it follows that
h(1]) = 1] + Al hI (1]) + o(AY)
where hd1]) is arbitrary. It may be determined by requiring 1] to be
continuous. In region 15, the receding characteristics are rectilinear and
1] = x (u w)t. When 1] is required to be continuous across LC, i.e.,
x = (u + w) t + ~1' it is found that
[q2  Y + 2] (T)  ~l)2
hI (1]) = 16q 2 w (u + w) .
Thus
ye. uw
Thus,
i.e.,
(11~1)2 {q2_Y+2 (q2+ Y2)}
~(1]) = 4w 4q2(U + w)  (y1)q 2u .
In region 7, the modification of the characteristics leads also to different
values of S. It is known that
R Alt $1 _ Alt
=2' 2yc. 2£0.
It is known that 5  Altj2 is constant along 1] = constant. Thus, let
5 A1 [tr(11)]
 2
where .(1]) is value of t where the characteristic meets the line x = ~l.
Thus
5 = ~ [t + 11~1]
2 uw'
so that
Al (11~1)
u1 =  2(uw) ,
WI 71 Al t [q2 + 71  2] [q2 +y 2] Al (11  ~1)
m= 2w(y1)q2 + 4w(UW)q2
Thus
(~) = u  OJ ~(11~1) [3q' + 71 2 ] _ [q2 + y2]YAlt
ot 1J 4q2(UW) 2(y1)q2
and
x= (uOJ)t Al(l1~I) (3q2 + y2)t_
4 qZ(uw)
(q2 + 71 
2) yAlt2 2
 4(y1)q2 + 1] + Al n l (1]) + O(Al) .
The arbitrary function nl is determined by choosing 1] to be continuous
across the line x = ~l. This gives
(~1 11)1 {q2(3 271) + 71  2
n l (1]) = ml(1]) + 4q2(UW)2 71 1 +
ul[(y  3)q2 +2+ 71  71 2] uw(qZ + 71  2) + W2(q2 + 71 2)}
+ (y1)u(u+w) •
Springer Tracts, Vol. 1: Gundersen 4
so S. Simple Wave Flows
Chapter 5
Simple Wave Flows
5.1 Introductory Comments
In isentropic flows, simple waves are defined by the condition that the
fluid velocity, the local speed of sound and the magnetic field are constant
along any curve of one family of characteristics. As a consequence, this
family of characteristics is rectilinear. For nonisentropic flows, the same
condition cannot be assumed for the entropy. This is due to the fact that
the entropy is constant along a streamline. In addition, if the entropy is
constant along a family of characteristics, which is distinct from the
streamlines, then the flow is isentropic. It might be assumed that a
possible solution would be to assume that the entropy is not constant
along a simple wave, but that the fluid velocity, local sound speed and
magnetic field are constant along a simple wave. But from the charac
teristic form of the basic equations, viz., Eqs. (1.2.4)(1.2.10), it is seen
that such an assumption is not possible, so that simple waves (in the
above sense) do not exist in nonisentropic flow. However, nonisentropic
perturbations of simple waves have been shown to exist [30].
For conventional onedimensional gas dynamic flows, the nonisen
tropic perturbation of a centered simple wave and the isentropic pertur
bation of an arbitrary (noncentered) simple wave were determined by
GERMAIN and GUNDERSEN [2], and a further discussion given by GUN
DERSEN [5], [6]. The nonisentropic perturbation of an arbitrary simple
wave was determined by GUNDERSEN [38], and the results of these papers
were used to discuss centered and arbitrary simple wave flow in ducts with
small nonuniform crosssectional area perturbations, viz., linear and
quadratic area distributions [39].
ROSCISZEWSKI [41] has used the GermainGundersen solution for the
isentropic perturbation of a centered simple wave to consider centered
simple wave flow in a duct with a small linear area variation; however,
there is no reference to that effect, nor, for that matter, is there any
reference to the earlier and more extensive discussion of effectively the
same problem by GUNDERSEN [39]. FRIEDMAN [42] has used a minor
5.2 The Monatomic Fluid 51
where
a3 = : (flo + ;t) .
From this result and the definition of the characteristic parameters, it
follows that
kc = a~/3 + a~/3
7=
"9 [k"2 ((X + flo) ]2/3  1, (5.2.3)
In order to transform Eq. (5.2.9) into a form for which results are known,
a new dependent variable l(lX, fJ) is introduced through the substitution
t*(
IX,
fJ) = (o:: +q,2/8fJ)1/81{0::,
_ 1
fJ)
R* = ~
00
y = 0, ~~ = 1 for ~ = f.
54 5. Simple Wave Flows
Then, changing back to the variables (ac, fJ), R* satisfies the relation
1
R* = "2 [sgn (ac  ac') + sgn (fJ  fJ')] R (ac, fJ; ac', fJ')
where R(ac, fJ; ac', fJ') is the Riemann function for Eq. (5.2.11).
From this result, the Riemann function of Eq. (5.2.10) may be ob
tained from the Fourier superposition of the solution of
(5.2.13)
Since the smaller of the exponents relative to the singular points (0, 1)
is ( ! ' 0) , the substitution Y2 (e) = C t Ya (e) is made which gives
(5.2.14)
(5.2.15)
x R+ [(I'2)C~W~] S}  ~2 '
2w~ IX.,
(5.3.1)
X
R [3W~ +2w~
(I'2)C~] S} {J  _ ~
.,  2'
(5.3.2)
For the case of no applied field, Eqs. (5.3.1) (5.3.2) reduce exactly to the
basic perturbation equations derived in [2J.
In the present section, it is assumed that the wave is centered at the
origin and characterized by {J = {Jo, a constant. For convenience, the sub
script zero will be dropped on the base flow. On each characteristic
dxjdt= U + w, (u, c, B) are constant, and these characteristics are straight
lines in the (x, t) plane so that the wave may be represented as:
x = (u + w)t (5.3.3)
{J = {Jo' (5.3.4)
From the definition of the characteristic parameters of the base flow, it
follows that
u = IX  {Jo (5.3.5)
w = (y  1) (IX + {Jo)j2 . (5.3.6)
In Eq. (5.3.1), an expression for IX., is needed, and this is obtained by
differentiating Eq. (5.3.3) with respect to x which gives the result
1
"T=U"+w.,. (5.3.7)
. [ w 2 + (I'  2) c2 ] •
However, smce w.,= (I' 1)w 2 W." Eq. (5.3.7) may be wntten as
+ [W2 (I'+_(I'2)C
~_ 2]
t  u., 1) w2 w., . (5.3.8)
The derivatives u., and w., may be expressed in terms of IX., through
Eqs. (5.3.5)(5.3.6) so that
IX., = 2w2j[3w2 + (y 2) c2Jt . (5.3.9)
Thus, for the wave centered at the origin, Eqs. (5.3.1)(5.3.2) reduce to
(5.3.10)
(5.3.11)
Eqs. (5.3.10) and (5.3.11) will be solved by considering the homogeneous
system with To = 0, and the physical significance of this reduced system
56 5. Simple Wave Flows
is that to the order of terms retained, the perturbed flow is still isentropic,
and then the solution of the complete system may be obtained by adding
a particular solution of the complete system. The function S may be
obtained from Eq. (5.3.11), and then R may be determined from Eq.
(5.3.10).
The characteristics of the homogeneous equation associated with
Eq. (5.3.11) are
dS = 0 and dxJ(u w) = dt. (5.3.12)
From Eq. (5.3.3), the second of Eqs. (5.3.12) may be written as
dx d
de = u w= u+w + t de (u + w) . (5.3.13)
From the definitions of the characteristic parameters and w, it follows that
From the first two ratios of Eq. (5.3.18), one first integral is
xlt = constant (5.3.19)
i.e., the rectilinear characteristics. From the first and last ratios of
Eq. (5.3.18)
d(dRt )
t
= [t.++ (~~)C'2]
00 y c
dd
t
[F((loo~.)]
((100)2
• (5.3.20)
To integrate Eq. (5.3.20), the first integral, Eq. (5.3.19) is utilized, i.e.,
on xlt = constant, w, e, c are constants, so that the solution of Eq. (5.3.20)
may be written as:
oo' + (2  Y)C 2 ] F[(loot2]
Rt  [ 300' + (y 2)c' ((loo)"~ = constant. (5.3.21)
From the two first integrals, Eqs. (5.3.19) and (5.3.21), the solution of
the homogeneous equation associated with Eq. (5.3.10) may be written as
Rt = [ oo' + (2  y)c' ] F[(loot2j G (.!....) (5.3.22)
300 2 + (y2)c 2 ((loo)t + t
From the first two ratios of Eq. (5.3.25), one first integral is
tt c (Y+1l/4(,,ll [1 + kC'l*]1/8= a (5.3.26)
where a is constant. From the first and last ratios of Eq. (5.3.25)
dS cD' [C<"+1)/2 <,,1l(1 + kC'I*)t tt]
(5.3.27)
Tt  (1 + kC'I*)tt
Using the first integral, Eq. (5.3.26), and the definition of y, Eq. (5.3.27)
may be written as:
2" (1.)
~ = ~c ,,1 t5/2 d D a2jtS (5.3.28)
dt 6 a dt·
From the two first integrals, Eqs. (5.3.26) and (5.3.29), the particular
solution for S may be written as:
S _ 2c 2 D(y) _ 2
 roy royt5/2C2/(" 1) X
(5.3.30)
X {f Q[a2/tt] d[c2"/("lltO/2]}a = 2C~(Y) + Ql(X, t)
where h is a constant. From the first and last ratios, the following differ
ential equation is obtained:
d[Rt] [00 2 + (2 y )C 2 ] [2C 2
+ + ill (x, t) +
]
;u = 300" (y  2) c" ooy il (y)
(5.3.32)
1
2c"t2 dQ
+ ooccy+lll,(Yll[l + kC'1*]t fit·
To integrate Eq. (5.3.32), the first integral, Eq. (5.3.31) may be utilized,
i.e., u and c may be treated as constants. This gives:
(5.3.33)
Rt = 2c 2 tQ(y)
ooy
+
ooy
1
The parameter r defined in Eq. (5.4.1) should not be confused with the
use of this symbol as a shock strength in Chapters 2 and 3. This symbol
is used here to agree with the notation of [2J, [5J, [6J, [38J, [39J and [40J
and to facilitate comparison with these papers.
The governing perturbation equations are Eqs. (5.3.1)(5.3.2), and
because of Eq. (5.4.2), Eq. (5.3.2) reduces to
5t + (u w)5", =  To/2. (5.4.3)
In order to solve Eq. (5.4.3), it is convenient to change from the indepen
dent variables (x, t) to (z, r). With the definition that 5 (x, t) = 51 (z, r), it
follows that:
J 5", = 51.  tb 51 T ,
J5 t = [xb+ r(u' + w')J51T  (u + w)51Z '
J = xb + r(u' + w')  (u + w)t~
where primes denote differentiation with respect to the argument.
From the definitions of the characteristic parameters and w, it
follows that:
u +w= 2w(z)/(y1) 2fJo + c[l + kc'1*J¥
so that
2t
, + ' _ [3 w (yI)wc
W 
+ (y  2) 2 C2 ]
c.
,
(5.4.4)
J   2 t'
w
+I(Y2)C
+ [3W (y
2 2
] ,
rc . (5.4.5)
0  ) wc
Under the prescribed transformation of independent variables, Eq.
(5.4.3) becomes
_ 2w5 [3w 2 + (y  2)c 2Jc'T 5 __ fTo
1.+ (yI)wc 1T 2' (5.4.6)
Since the particle paths in the simple wave are given by ewr = constant
or, more conveniently by
y = C(Y+1)/2(y1) [1 + kc'1*Jh¥ = constant (5.4.7)
and
l' C2 F' ('Po)
0= ewr y(y _ I)C.WT '
5.4 Perturbation of an Arbitrary Simple Wave 61
it is convenient to write
To c2 ,Q'(y)
2 ror
(5.4.8)
From the first two ratios of Eq. (5.4.9), the following first integral is
obtained:
C(Y+l)/4(Yl)·d" [1 + kC f1 *Jl/8 = a (5.4.10)
a constant or
eW7:2 = constant. (5.4.11)
2dS l = [ 3ro' +
(y2)c 2
(yl)ro'(I+kcf1*)'i
1] Q'(y) dc 2c',Q'(y) dto.
ror
(5.4.12)
(5.4.14)
2a Sl =
6 1
c 2 }'/(yl) 7: 5 / 2 Q(a 2 j7:2)
 f +
C'Y!<yI)
r [3ro'
r5/2 2(y  1) ro' ct~
(y  2) c'] c' dQ + constant.
62 5. Simple Wave Flows
[ 2(Yl)CW2t~] 2
X r: [3w2 + (y2)c 2]c'  WYT6/ 2 C2!(y1l X
, (5.4.16)
X {!.Q(a2jrl)d[r:S/2C2l'/(l'I) {r: 2(yl) cw2to }]} ==
[3w· + (I'  2) c2]C' a
2c l .Q(y) [ 2(Yl)CW"t~]
==2(e w )hF'[e W r:2]+ WyT r:[3w2+(y_2)c2]c' +4>(z,r:)
Substitution of Eq. (5.4.17) into Eq. (5.3.1) gives the equation for R, viz.,
(y2)C 2 _W 2 ] }
Rt+(u+w)R.,+ {R+ [ 3W2+(y2)c" 5 X
(5.4.18)
1
X { 2(y 1) w"cto }
T [3w2+ (y2)c 2]c'
In order to integrate Eq. (5.4.21), the first integral Eq. (5.4.20) may be
used, i.e., z may be treated as a constant. This gives
vR = [ w' + (2  y) c' J F [e WT2 ] + [ w' + (2  y) c2 ] ~! X
3w 2 + (y 2)c 2 (ew)t 3w 2 + (y  2)c 2 W
Q(y) [
X y(tto) tt o
2(yl)cw 2to J
[3w'+ (y2)c2]c'
c2
dt+o;
Q'(y) ! tto X
X [tto
2(yl)cw2t~
[3w 2 + (y2)c 2]c'
J
dt+
[W 2 +(2 y )C 2
3w2 + (y2)c 2
J! X
(5.4.22)
+t
f[ w' + (2  y)c' ] 2C 2
3w 2 + (y2)c 2 W
! Q(y)
y(tto) vdt+
(5.4.23)
+ [t~.: (~y=y~)C;2 J ! (/J (h, t  to) dt 
Pt + uP., = o. (5.5.7)
Since P, e and s are related by some equation of state, e.g., Eq. (5.5.5),
Eqs. (5.5.4) and (5.5.7) give
et+uex=O (5.5.8)
so that Eq. (5.5.1) reduces to
eux = 0 (5.5.9)
and u = 1t(t) alone, say
u =  f' (t) . (5.5.10)
From Eqs. (5.5.3) and (5.5.9), B satisfies the equation
Bt + uBx = O. (5.5.11)
Consequently, B, P, e, sand lJI satisfy the same partial differential equa
tion. Further, no explicit assumption of an equation of state has been
made in the derivation to this point, and this class of flows may be gene
rated from the single assumption B = B (lJI).
When Eq. (5.5.10) is substituted into Eqs. (5.5.4), (5.5.6), (5.5.7),
(5.5.8) and (5.5.11), it is seen that s, lJI, P, e and B are each functions of
y = x + I(t). Eq. (5.5.2) may be written as
e I"()t +dP
 +B'" 
dy
dB
dy
=0 (5.5.12)
and since Band P are constant along a trajectory, Eq. (5.5.12) reduces to
I" (t) = 2A
where A is constant, so that
I (t) = A t2 + Al t + A 2 •
With no loss of generality A2 may be taken as zero, and, if u (0) = U o,
u(t) = 2At + U o
and the trajectories are given by
y = x + I (t) = x + A t 2  uot = constant, (5.5.13)
i.e., a family of coaxial parabolas.
From the definition of lJI, dlJl = edy so that from Eq. (5.5.2)
1 d [ B2] (5.5.14)
e= 2A dY P+ 2",
Springer Tracts, Vol. 1: Gundersen 5
66 5. Simple Wave Flows
1
with the solutions
t= ± 1 d
1;[p +:i]
[B8]
}t
dy. (5.5.17)
bB dy P+ 2p + 2A 'Y P
If the equation of state is prescribed, e.g., Eq. (5.5.5), and also one initial
condition, e.g., s = s (P) or P = P (x) at t = 0, the solution may be
determined uniquely.
Example 1. Isentropic flow with B = Constant. If. Eq. (5.5.5) is
written as P = K(!'" with a constant K, it follows from Eq. (5.5.14) that
1
2Ay=f[!]Y dP
so that
)'
x + At 2 = ± j [(p.~ ++ E~)] It +
[2A
A, Po
2AYPo]
2p.
t constant.
2 2A Y,1.;Y
c = ,1.3 +,1.4 '
w2  2Ay(,1.3Y + 2,1.4)
 ,1.3 +,1.4
Chapter 6
[46] of the formation and decay of shock waves to the magnetic case
with little difficulty. Since the generalized Riemann invariants and an
exact solution for simple wave flow are known explicitly for a monatomic
fluid, the discussion will be limited to a monatomic fluid in order to sim
plify the expository analysis of this chapter.
Because the differences between a shock transition and a simple wave
transition with the same strength and same initial state involve only
terms of third and higher order in the strength, the Friedrichs theory,
which may be used for an approximate description of the propagation of
nonuniform shock waves which are not too strong, replaces the shock
transition relations by those through a corresponding simple compression
wave. Since simple waves exist only in isentropic flows, the replacement
is equivalent to the assumptions that the entropy and an appropriate
generalized Riemann invariant do not change across the shock. After the
simple wave, which replaces the actual disturbance, the fluid must be
again at rest. The method is most easily applied when the flow in front
of the nonuniform shock wave is initially at rest, and, for simplicity, the
present discussion will be limited to that case.
The critical evaluations of the Friedrichs theory which have been
made will not be discussed, though, in this connection, see the paper by
LIGHTHILL [47], who showed that although the Friedrichs theory
neglected third order terms, some of the neglected terms can integrate
to give a second order effect, particularly for very weak shock waves
which take a very long time to decay. This occurs, for example, when
discussing the tail shock of an Nwave, behind which the Friedrichs
theory assumes undisturbed fluid while LIGHTHILL shows the region to be
a compression wave.
r
2+ k (6.2.1)
 ~ + (1 +:C)8/2 =  ; + ! (~ = {J • (6.2.2)
where the subscript zero denotes quantities in the constant state in front
of the simple wave. From Eq. (6.2.3)
C= 
1 1 [
""II +""11 (1 + kCO)3/2 + 2k (u  u o)
]2/3 (6.2.4)
70 6. Formation and Decay of Shock Waves
and since
w = c(l + kC)1/2,
:0 = (:.r,
P (
p;=c;;'
)5 C
pansion is
I 1
U = u o + Wo + 2" (u +W  u o  wo) + 2co(8 + 9m~)2 X
(6.2.7)
69mt + 136m~ + 64
X[
4(1 + m~)"2
1
Inserting Eq. (6.3.3), which is correct through the second order, into
Eq. (6.3.2), the solution may be determined uniquely when the initial
condition t = 0 when ~ = 0 is appended. It is convenient to express the
solution in terms of the function O"t (~) defined by
D Wo = WOO"lW .
The explicit solution is
(6.3.4)
where
D _ 69mt + 136m~ + 64 4mW + m1)3!2
 (8 + 9mn'
Inserting Eq. (6.3.4) into the first of Eqs. (6.3.1) gives
Da,(;) ]2 f
o
x = ~ + 8[1 + O"l(~)J [4a,(;) a,(y) dy
[4Da,(y)]8 (6.3.5)
<
Eqs. (6.3.4)(6.3.5) give a parametric representition of the shock path.
6.4 Decaying Shock Wave 73
~
the following asymptotic expansion .:c
Fig. 9. The lines OA and A B represent the path
is obtained from Eq. (6.3.4) of a piston. When the piston is arrested at point
A I a simple rarefaction wave is generated. The
(It)ot)~ = 2A ~ [~ 
01
_E_]
4
simple wave catches up with and causes the initially
l1nifonn shock wave to decay. d (t) is the width of
the wave zone
Fig. 10. The lines 0 A. A Band Be represent the path of a piston. The decaying N wave may be discussed
by the use of the Friedrichs theory
(6.4.4)
for u and £0 then follow directly from the explicit representations of these
functions in terms of D (;) .
From Eq. (6.4.5), it follows that
a,d (t) a 5 d 2 (t)
(11 = a,wot + 1 + (a,wot + 1)3
which is equivalent to Eq. (6.4.1). Exactly as in the nonmagnetic case,
it may be shown that the shock strength decreases like rt.
The problem of this section may be generated as illustrated in Fig. 9,
and the constants a", as and A may be determined quite easily [46].
Another problem, which may be treated in an entirely similar manner,
is the decaying Nwave, see Fig. 10. In front of the head shock, the flow
is known exactly, viz., a uniform state of rest, so that the flow behind the
shock may be determined with an error of third or higher order in the
shock strength, and the second order theory is valid for all time. However,
the treatment of the tail shock presents difficulties since the flow which
precedes it is only known approximately. See LIGHTHILL [47].
or
r AS kr A )1/3]
X = 2 + [T3 rA + Po  Ii k Po + 2
1 (
(t  A) . (6.5.2)
Although the envelope ofEq. (6.5.2) is easily determined, say [xE (A),tE (A)],
the point at which the shock formation begins is determined by finding the
minimum of tE (A). Since this is given at best by a complicated implicit
representation, it seems more fruitful to give an approximate represen
tation for Eq. (6.5.2), valid for small applied fields. Since
~
k
[kp 0 +
kr A ]1/3 = (1
2
+ kco)" +
k
1
r AC~ _
6w~
__1_
36
kr2A2cg
wg+
o(k2"s 13)11.,
x= 2
d. + [3"2 r A + Wo 
I
6 (1
d.
+ mi) ] (t  A) . (6.5.3)
The envelope of Eq. (6.5.3) is given by
E_ 2(5 + 6mi) A. + 6(1 + mnS/Bco (6.5.4)
t  (8 + 9mi) (8 + 9mnr '
E _ ~ + {(2 + 3mi) A. + 6(1 + mi)S/co}
x  2 (8 + 9mi) (8 + 9mi)r X
grA +
(6.5.5)
X wo 6(1 ~mi)}'
The initial point of shock formation is given by
E 6(1 + mi)S/co
tmln= tv = (8 + 9m¥)r '
Xv = wotv'
Since the shock begins at (xv, tv) and lies within the region x > wot, the
region in front of the shock is a constant state so that the methods of this
chapter are applicable.
By shifting the origin to the point (xv, tv), the simple wave may be
represented by
(8 + 9mnrA. ] (5 + 6mi) I' A.'
x x.= (t tv) [
6(1 + mi) + Wo  6(1 + mi) (6.5.6)
(8 + 9mi)rA. .
Since Q = wo(1 + 0'1) = 6(1 + mi) + wo, It follows that
(8 + 9m¥) rA.
0'1 = 6(1 + mi)wo
(6.5.7)
so that
l:. = _ 6 (5 + 6mi) (1 + mi) wg~
S" 1'(8 + 9mi)2
In order to determine the motion of the shock wave, these results are
substituted into Eq. (6.3.4). This gives
W
oVal
(t t) = 96 [4 Dal ]2 (5 + 6mi) (1 + m¥)wg
(8 + 9m¥)2r
f a,
aida,
(4Dal)8
o (6.5.8)
= 12(5 + 6mi) (1 + m¥)wg
(8 + 9mi)2r
{~
3 0'1 +
(~_~)
8 3
2
0'1
o(err)}
+ l'
By inverting Eq. (6.5.8), the following expansion for 0'1 is obtained
(8+9mi)2(ttv)r [ (8+9m~)8(tt.)r]
0'1 = 8w o(5 +
6mi) (1 mi) 1 128wo(5 +6m¥) (1 mi) • + + (6.5.9)
Thus, from Eqs. (6.5.6), (6.5.7) and (6.5.9),
1'(8 + 9mi)2(tt.)2
x x" = wo(t t,,) + 32(5 + 6m¥) (1 + m¥) + .... (6.5.10)
78 7. The Effects Due to an Oblique Applied Field
Chapter 7
(7.1.2)
(7.1.3)
2
B
at • Y (y
1 ) c. = 0, (7.1.9)
80 7. The Effects Due to an Oblique Applied Field
COsc
y(yI)c. = 0, (7.1.11)
s~ = o. (7.1.12)
For the case of a transverse field only, Bl = 0, as = 0 and at = W
= [b~ + c2 Jt, and the above characteristic system reduces to that derived
in Chapter I, viz., Eqs. (1.2.6)(1.2.12).
In the transverse field case, the particle paths were characteristics of
multiplicity two. Three independent characteristic directions led to
four independent characteristic forms of the governing equations, and a
reduction in the number of equations was possible. A similar reduction is
possible in the present case for simple wave flows.
Hydromagnetic simple wave flow was considered by FRIEDRICHS [IJ.
who showed that the problem could be reduced to finding the solution of
a single linear firstorder differential equation, which could be solved
explicitly for the case of a polytropic gas, and by BAZER [54J, SHERCLIFF
[55J and others. In principle, simple wave flow would be characterized by
integrating one of Eqs. (7.1.8)(7.1.11) and setting the result equal to a
constant thus determining a Riemann invariant. In actuality, a better
approach is to look for solutions constant along a single family of rectilinear
characteristics, i.e., to determine the motion of a "phase" which moves
with constant velocity. For example, consider a forwardfacing wave cha
racterized by
x = [Ut(b) + a(b)Jt + b (7.1.13)
where b is a parameter. From Eq. (7.1.13), it follows that
a 1 a
ax 1 + [ui(6) +a'(6)]t ac5'
a [ud6) + a(6)] a
at 1 + [ui(6) + a'(6)]t ac5.
Substituting these results into Eqs. (7.1.1)(7.1.5) gives
(1'1)
 ac,,+ 2 cU}d= 0,
2cc" b~ C'S6
aU16+1
1'
+B• B 2,, y ( y  I )=0
c.'
of Eqs. (7.1.14) may be put into a more convenient form, e.g., for a for
wardfacing fast wave
(7.1.15)
where b has been replaced by 0: to agree with Eq. (7.1.9). When the first
of Eqs. (7.1.15) is integrated and set equal to a constant, a generalized
Riemann invariant is determined which is completely analogous to
Eq. (1.2.9) to which it reduces when Bl 7 O. The system of Eqs. (7.1.15)
is completely equivalent to Eq. (7.1.9).
The system corresponding to Eqs. (7.1.15) for a backwardfacing fast
wave is
(7.1.16)
where the symbol a has been written without subscript since the same
relations hold for a = a, or a = as. It follows immediately that these
differential relations must hold throughout the flow, so that the follo
wing integrals are obtained:
10gB 2 + y
1
2 f a 2 de
(b I2  a 2) e = constant, (7.1.17)
Uz + n:. f
1 B.b~
(a 2 _ b~) dU l = constant. (7.1.18)
Springer Tracts, Vol. 1: Gundersen 6
82 7. The Effects Due to an Oblique Applied Field
St + Uts.. = 0 (7.1.21)
where an arbitrary, timeindependent area variation has been added, and
it should be noted that the system of Eqs. (7.1.19)(7.1.21) is valid
for isentropic flow only. The entropy has been included since these
equations will be linearized in the neighborhood of a known isentropic
state. In order to make the generalized Riemann invariants linear
relations between the dependent variables, the variable w is introduced by
the substitution w = f:
de. Then, appropriate combinations of the
resultant system lead to the following characteristic form of the equations
U1
[2+ W] t+ (Ut + a) [U
,,1
1 ,,1
2+ W] ..
(7.1.22)
St + U1S.. = O. (7.1.24)
Then, linearizing Eqs. (7.1.22)(7.1.24) in the neighborhood of a known
(isentropic) constant area flow, i.e., a uniform or simple wave flow,
denoted by the subscript zero, the following system of linear equations is
obtained for the terms of first order, denoted by a bar:
Rt + (Uto + ao) R .. + (u 1 + a)oc..
(7.1.25)
(7.1.26)
ulOaoA(x)
2Ao 21'(1'I)c. '
St + Utos.. = 0 (7.1.27)
where the firstorder generalized Riemann invariants have been denoted
7.1 The Characteristic Form of the Governing Equations 83
by Rand 5, i.e.,
U iii
R= u;. +~ S=.!..+
2 yl' 2 yl
and it has been noted that
~+~= 0: ~+~=fJ
2 yl ' 2 yl
are the characteristic parameters of the base flow. Throughout, the symbol
a will appear without the subscript s or f since the analysis is valid for
slow or fast waves, and the appropriate characteristic parameters will be
denoted by (0:, fJ). Further, the subscript zero will be dropped from base
flow quantities in order to simplify the notation.
Since Eq. (7.1.7) for the wave speeds may be written in the equivalent
form
b~
a2b~
(7.1.28)
it follows from Eq. (7.1.28) that
ii = [ (a'  c2 )' + (y  1) b~C'] ac
b~c' + (a 2  C2)2 (y 1)c
Consequently, the firstorder generalized Riemann invariants may be
written in terms of u and ii, namely,
u;. [ b~C2 + (a 2  C2)2 ] _
R=""2+ (a2c2)2+(y_l)b~c2 a,
u;. [ b~C2 + (a 2  C2)2 ] _
5 = ""2 + (a2 _ C2)2 + (y _ 1) b~C2 a.
Thus
_ __{3(a'C 2)2 + (y + 1) b~C2}
u1 +a 2 [b~C2 + (a2 _ c2)'J X
(y  3) b~c'  (a 2 _ C')2}
X R +{ 2[b~C2 + (a'_c 2)2] 5,
_ _ { (a'  c 2)'  (y  3) b~C2}
U1  a = 2[b~c' + (a' _ c2)'J X
(7.1.29)
For an initially uniform flow, ex and f3 are constants, and the general
solution of Eqs. (7.1.29)(7.1.30) is
R=F[x( a)t] E[xu1t]_ u1al'(x)
Ut + + a 2A (u 1 + a) ,
S=G[x( a)t] E[xu1t] _ u 1al'(x)
Ut + a 2A (u1  a) ,
T o =2E'[xu1 t]
in terms of three arbitrary functions of one argument. This solution is
directly analogous to the solution given by Eqs. (1.2.30)(1.2.32), which
it includes as a special case, and may be used to extend the analysis of
Chapters 2, 3 and 4 to the oblique field case. Although the extensions
will be given in a future publication, the next section is devoted to a
relatively simple formulation of the transition relations across oblique
shock waves. The final two sections of the chapter are concerned with
simple wave perturbations.
(7.2.8)
86 7. The Effects Due to an Oblique Applied Field
and
c~ T
(7.2.9)
4 a
From the continuity of the normal component of the magnetic field, i.e.,
Eq. (7.2.1) and Eqs. (7.2.7) and (7.2.8), the following relations are derived
B.. t/>.
Bn = t/>1·'
b~2 1
bh a'
m~2 t/>~
m~l = Tt/>~ ,
b~2 = t/>~ (7.2.10)
bh at/>~'
m~2
m~l :r '
nh = ;;;,
n~2 t/>~
n~l = aTt/>~ .
Since u 1 = v in the coordinate system utilized, Eq. (7.2.4) may be written as
(11 tPl [vi  bi IJ = (12 tP2 [v~  bi 2J
or, solving for tP2
(7.2.11)
which expresses tP2 in terms of (1, tPl' m 21 = tPl mn and the known flow in
front of the shock.
Since Eqs. (7.2.1), (7.2.2),7.2.4) and (7.2.5) have been utilized, only
Eqs. (7.2.3) and (7.2.6) remain to be considered. These may be rewritten as
P1 + b~l P b~2
+ (112
(11 VI
2
= 2 + (12 V2 + (12 2 ,
2
vi(1 t/>i) + yP + b2 1
2 (yl)e1 21
= vWt/>~) + yp. + b2
2 (yI) eo 22
or, if all terms with subscript two are expressed in terms of those with
subscript one
.1.2.1.2 1 "m21
[a2 (I'I'I)+'I'2 I J [ +  2 
2 (1 :1:2
.p~ )] 
'1'.
2
"0'(0'1)
.[ 1 2 (1
y=I+m21 . 1 . '1'2
~
2 )]
7:= (7.2.14)
a'(Icf>~) + cf>~I _~ (0'1)
,,1
which expresses 7: in terms of ,pI and 'ln 21 • Eq. (7.2.14) gives the Hugoniot
function for the case of an oblique magnetic field.
Because of Eqs. (7.2.11) and (7.2.13), Eqs. (7.2.7)(7.2.10) express
the flow quantities behind the shock in terms of those in front and the
three parameters a, ,pI and ~l. In the limit of vanishing Bv these results
reduce to those presented in section 2.2 for the case of a normal shock.
Some care must be exercised in carrying out this limit, e.g., it must be
noted that
lim 1:!. = a .
B,+O cf>l
fJ = fJo· (7.3.2)
However, the derivatives U 1 ", and w'" may be expressed in terms of IX",
through Eqs. (7.3.3)(7.3.4). This yields the result
(7.3.5)
(7.3.6)
(7.3.7)
where attention will be restricted to simple wave flow in a uniform duct.
Eqs. (7.3.6)(7.3.7) will be solved by first considering the associated
homogeneous system with To = 0, and then the solution of the complete
system may be obtained by adding a particular solution of the complete
system.
The characteristics of the homogeneous equation associated with
Eq. (7.3.7) are dS = 0 and
(7.3.8)
From the first two ratios of Eq. (7.3.14), one first integral is
xlt = constant, (7.3.15)
i.e., the rectilinear characteristics. From the first and last ratios of
Eq. (7.3.14), the following differential equation is obtained
d (tR) = [ (3  y) b~C2 + (a'  C2 )2 ] ~ [F [eat']]
dt 3(a 2 c 2 )' ..L, (y + I) b22 c 2 dt
(7.3.16)

(e a ).L2
In order to integrate Eq. (7.3.16), the previous first integral, Eq. (7.3.15),
may be utilized, i.e., on xlt = constant, the flow parameters are constant
so that the solution of Eq. (7.3.16) may be written as
(3_Y)b~c2+(a2_C2)2] F[eat2]
Rt [ 3(a2c2)2+(y+l)b~c' ()t =constant. (7.3.17)
ea
From the first integrals, Eqs. (7.3.15) and (7.3.17), the solution of the
homogeneous equation associated with Eq. (7.3.6) may be written as
(3  y) bilc' + (a 2 c')' ] F[eat'] (X)
Rt= [
3(a'c2)2 + (y + l)b~c'
2
.L
(e a )2
+G t (7.3.18)
it is convenient to write
To = 2c 2 .Q'(y)
at
90 7. The Effects Due to an Oblique Applied Field
From the two first ratios of Eq. (7.3.20), one first integral is
(7.3.21)
where .It is a constant. From the first and last ratios of Eq. (7.3.20)
dS tt
c2 Q' [at C1/(Yl)]
dt at (7.3.22)
Using the first integral, Eq. (7.3.21), and the definition of y, Eq. (7.3.22)
may be written as
(7.3.23)
From the two first integrals, Eqs. (7.3.21) and (7.3.24), the particular
solution for S may be written as
S = 2c 2 Q(y)
ay
(7.3.25)
_ 2c 2 Q(y) Q ( )
= ay + 1 X, t
where the notation { h denotes that the quadrature is to be carried out
with constant .It which must then be replaced by its equivalent from
Eq. (7.3.21) to give the particular solution for S, and the quadrature is
abbreviated by Q 1 (x, t).
The particular solution for R may be obtained from the equation
(y_3)b~c2_(a2_c2)2 }] c'Q'(y)
tR t + xR.,+ R
[
+ { 3(a' _ c')' + (y + 1) b~C2 S = a
7.3 Perturbation of a Centered Simple Wave Flow 91
To integrate Eq. (7.3.27), the first integral, Eq. (7.3.26) may be utilized,
i.e., the flow parameters are constant on the rectilinear characteristics.
This gives
(3  1') b~c' + (a 2  C')2 ]
Rt = [ 3 (a' _ C')2 + (y + 1) b~c' Q 1 (ht, t) dt +
f
+[
(3  y) b~C2 + (a 2  C')']
3(a'  c2 )' + (y 1) b~c' +
2c 2 f Q(y) dt
+ (7.3.28)
f
a 3/2 cl/(1'I) t§
2c' dQ(y)
+ a 31 'cl/(1' 1) t2 
1
d t  dt + constant.
Performing an integration by parts in the third term on the righthand
side of Eq. (7.3.28) and using this result and the first integral, Eq. (7.3.26),
the final particular solution for R may be written as
1
2c 2 tQ(y) c 2 t2 (531') b~c' (a'c')']
Rt= ay
+ [ ay c')' + (y + 1) b~c'
3 (a' 
X
X {f Q(y) dt}
.l
t" h = zit
...L [
I
(3  y) b~c'
3 (a' _ c2 )'
+ (a 2  c2 )' ]
+ (I' + 1) b2c2
2
X (7.3.29)
X {f Q1(ht, t) dtL=zJt
where the notation { h = zit denotes that the quadrature is to be carried
out with xlt constant; then, replacing h by xlt, the solution, Eq. (7.3.29),
is obtained.
The general solution for the nonisentropic perturbation of the
centered simple wave is given by Eq. (7.3.12) plus Eq. (7.3.25) and
Eq. (7.3.18) plus Eq. (7.3.29).
92 7. The Effects Due to an Oblique Applied Field
(7.4.3)
so that
, , [3(aSC2)2+("+1)b~CS] ac'
Ul +a = b~C2 +
(a s _c 2)S (,,l)c (7.4.4)
Since the curves Po = constant ar~ given by eaT = constant or, more
conveniently, by
1
= constant,
~11
it is convenient to write
(7.4.8)
From the first two ratios of Eq. (7.4.9), the following first integral is
obtained
1
C 2 (1'1) at't't = A, (7.4.10)
where A, is a constant or
(la't'2 = constant. (7.4.11)
Eq. (7.4.11) gives a convenient representation of the curvilinear cross
characteristics of the simple wave. From the first and third ratios of
Eq. (7.4.9), the following differential equation is obtained
cD' (y) [3 (a'  c2)O + (I' + 1) blC'] dc
2dS 1 = (yl)a b~c2+(a'c')'
(7.4.12)
2c'D'(y) dto
aT
In order to integrate Eq. (7.4.12), the previous first integral may be used,
and this gives a relation between z and 't', namely,
dz 2(yl)c[b~c2 + (a 2  c 2)']
"""dT =  T [3 (a 2  C')2 + (I' + 1) b~C2J c' (7.4.13)
A6 dSl =
2 't'
21' [']
5/2 1'1 dD A2 /T2
C dT
dt
0
+ 't'6/2 C 1'1
21'
dQ
• (7.4.14)
X ~~ + 4(y; 1) {f Q[A,2/Tt] d X
From the two first integrals, Eqs. (7.4.11) and (7.4.15), the solution of
Eq. (7.4.3) may be written as
1 2c 2 .Q(y) [ t~
SI=2(e a )'TF'[e aT2 ]+ ayr T2(yl)ce'X
{ b~C2 + (a 2  C2) 2 } ] 2
X 3(a2_c2)2 + (y + 1) b~C2  _;O2~5 X
r' ay
{f
c yl
Substitution of Eq. (7.4.17) into Eq. (7.1.29) gives the equation for R, viz.,
R t + (u1 + a)R", +
(7.4.18)
(y3)b~c2(a2c2)2 }] 1 To
+ [ R+ 5 { 3(a2_c2)2 + (y + I)b~c2 ;=2
2 (y  1) ct~ [b~C2 + (a 2  C2)2]
where l' =T  c' [3 (a2 _ C2)2 + (y + 1) b~C2]
7.4 Perturbation of an Arbitrary Simple \Vave 95
I [ (a 2  ( 2)2  (y  3) b~e2 ]
T 3(a2 _ C2)2 + (y + 1) b~e2 lP(h, t to) + (7.4.21)
+[ (a2_C2)2_(Y_3)b~e2] 2e 2vQ(y) + e 2vQ'(y)
3(a 2 _e 2)2 (y 1) b~e2 +
ay(tto) + a(tto)
In order to integrate Eq. (7.4.21), the first integral Eq. (7.4.20) may be
used, i.e., z may be treated as a constant. This gives
_ [ (a2C2)2_(Y_3)b~e2] F[ear2J
vR 3(a 2  c 2)2+(y+l)b2e 2
2
~
(e a ) 2
+
(a2_c2)2_ (y3) b~c2 ] ~f Q(y)vdt
+[ 3(a2c2)2+(y+l)b~c2 +
f lP(h, t to) dt +
a y(tto)
(7.4.22)
(a2  C2)2  (y  3) b~C2 ]
+ [
3(a2  c2)2 + (y + l)b~c2
e2
+;;
f Q'(y)vdt
tto +constant.
After an integration by parts is performed in the last integral in Eq.
(7.4.22), the following solution is obtained from Eqs. (7.4.20) and (7.4.22)
(a2  C2)2  (y  3) b~C2 ] F [ear2J
[
vR=G[zJ+ 3(a2_e2)2+(y+l)b~c2 (ea)t +
2e 2v Q (y) {2e 2 [ (a 2  ( 2)2  (y  3) b~e2 ]
+ ay + a 3(a 2 _e 2)2 + (y + l)b~c2 X
X
f Q(y)vdt
+ 3(a2c2)2+(y+l)b~e2
y(tto)
[(a2_e2)2_(Y_3)b~C2]
X
References
[1] FRIEDRICHS, K. 0.: Nonlinear wave motion in magnetohydrodynamics. Los
Alamos Report 2105, 1957. See also a later version of this report by
K. O. FRIEDRICHS and H. KRANZER: Nonlinear wave motion. A. E. C.
Research and Development Report, NYO6486, 1958.
[2] GERMAIN, P., and R. GUNDERSEN: Sur les ecoulements unidimensionnels d'un
fluide parfait a entropie faiblement variable. C. R Acad. Sci. (Paris) 241,
925 (1955).
[3] LUNDQUIST, S.: Studies in magnetohydrodynamics. Arkiv flir Fysik 5, 297
(1952).
[4] GUNDERSEN, R.: The propagation of nonuniform magnetohydrodynamic
shocks, with special reference to cylindrical and spherical shock waves.
MRCTS310, 1962. Published in Arch. Rat. Mech. Anal. 11, 1 (1962).
[5]  The flow of a compressible fluid with weak entropy changes. Thesis,
Brown Univ., Mar., 1956.
[6]  The flow of a compressible fluid with weak entropy changes. J. Fluid Mech.
3,553 (1958).
[7]  Shock propagation in timedependent ducts. J. Aerospace Sci. 28, 748
(1961).
[8] CHESTER, W.: The quasicylindrical shock tube. Phil. Mag. (7) 45, 1293 (1954).
[9] CHISNELL, R: The motion of a shock wave in a channel, with applications to
cylindrical and spherical shock waves. J. Fluid Mech. 2, 286 (1957).
[10] WHITHAM, G.: On the propagation of shock waves through regions of nonuni
form area or flow. J. Fluid Mech. 4, 337 (1958).
[11] ROSCISZEWSKI, J.: Calculations of the motion of nonuniform shock waves.
J. Fluid Mech. 8, 337 (1960).
[12] GUNDERSEN, R: A note on shock flow in a channel. J. Fluid Mech. 4, 501 (1958).
[13] STOCKER, P.: The transients arising from the addition of heat to a gas flow.
Proc. Camb. Phil. Soc. 48, 482 (1952).
[14] FRIEDMAN, M.: An improved perturbation theory for shock waves pro
pagating through nonuniform regions. J. Fluid Mech. 8, 193 (1960).
[15] GUNDERSEN, R: Secondary shocks in magnetohydrodynamic channel flow.
Int. J. Eng. Sci. 1,241 (1963).
[16]  The propagation of a plane shock wave into a moving fluid. J. Aerospace
Sci. 28, 584 (1961).
[17]  A perturbation analysis of shock flow in a nozzle. J. Aerospace Sci. 26,
763 (1959).
[18] MIRELS, H.: Source distribution for unsteady onedimensional flows with small
mass, momentum, and heat addition and small area variation. NASA
Memo 5659E, 1959.
[19] GUNDERSEN, R: The pistondriven shock. J. Aerospace Sci. 27, 467 (1960).
[20]  Magnetohydrodynamic shock propagation in nonuniform ducts. MRC
TS287, 1961; Z.A.M.P., 14, 124 (1963).
[21]  Cylindrical and spherical shock waves in monatomic conducting fluids.
MRCTS299, 1962. Appl. Sci. Res., 10 B, 119 (1963).
[22]  The propagation of a nonuniform magnetohydrodynamic shock wave into
a moving monatomic fluid. J. Aerospace Sci. 29, 1421 (1962).
[23] DE HOFFMAN, F., and E. TELLER: Magnetohydrodynamic shocks. Phys. Rev.
80,692 (1950).
[24] LUST, R.: Magnetohydrodynamische StoBwellen in einem Plasma unend
licher Leitfahigkeit. Z. Naturforsch. 8a, 277 (1953).
References 97
7*
Appendix B
The Characteristic Form of the Basic Equations
An outline of the CourantFriedrichs method [43J for obtaining the
characteristic form of the basic equations (1.2.2)(1.2.5) for the case of
no area variations will be given.
If Eqs. (1.2.2)(1.2.5) are multiplied by~, A2 , Aa and A"" respectively,
the resulting equations added, and it is required then that the derivatives
of u, c, Band s enter only as directional derivatives with respect to e,
there results
2 AIC t + [2AIU + 2A.C] (B I)
1'1 1'1 1'1 C.,=CE,
A2 H t + [~c + A2U + AaBJ u., = UE, (B2)
(B4)
From (B 1)(B4)
XE AIU + A.c AIC + A.u + A.B
IE = ,1.1 = A.
(B5)
A.b"IB + A.u A4uA2c"ly(y I)c.
A. ,1.4
The Eqs. (B 5) lead to the following three independent relations for non
vanishing Aj (j = I, 2, 3, 4)
cA~ = cAr + AlAa B (B6)
Alb"
CAa=~' (B7)
A  AIC (BS)
"' y(yI)c.·
If A2 = 0, then Eqs. (B5) lead to two relations when A", is arbitrary but
Al and Aa do not vanish. Thus
A2 = 0 , Al C + Aa B = 0 . (B 9)
The cases when Al and/or Aa and/or A", vanish lead to no new equations.
If the variable w 2 = b2 + c2 is introduced, then after eliminating ;: in
Appendix B 101
(B 11)
Consequently, there are three basic types of relations for Aj which follow
from Eq. (B 5): (1) the Eqs. (B 7), (B 8) and (B lO); (2) the Eqs. (B 7),
(B 8) and (B 11); (3) the Eqs. (B 9) with }'4 arbitrary. From Eq. (B 5), it is
seen that if the curves of the above three families are parameterized by
{3 = variable, a; = variable, ~ = variable, respectively, then
x/3 = (u + w) t/3' (B 12)
x"=(uw)t,,, (B 13)
x~ = ute. (B 14)
Thus Eqs. (B 12) and (B 13) are the C+ and C_ characteristics and (B 14)
are the streamlines.
Now, returning to Eqs. (1.2.2)(1.2.5) and multiplying them respect
ively by AI' A2 , A3 and A4 , adding them and using the relations (B 7), (B 8),
(B lO) and (B 12) gives
2 2 B/3 2cC/3 _~ _ _
(wC)B+ WU /3+(y_l) y(yl)c.O. (BI5)
Similarly, from Eqs. (B7), (B8), (B 11) and (B 13), there results
2 2 B" ~_ c's" _
(wC)B wu"+y_l y(yl)C.O. (BI6)
Finally, by use of Eqs. (B9) and (B 14)
.!:!L _ 2ce
(B 17)
B (yl)c = O.
The Eqs. (B 15)(B 17) and
Se = 0 (B 18)
are the characteristic forms of Eqs. (1.2.2)(1.2.5).
Appendix C
Sigma K«(f, .0, .00) Y«(f, .0, .00) Z«(f, .0, .00)
1.10 .4 8 33 28 .100901 .5 84 22 9
1.20 .47 14 84  .206667 .6781 51
1.3 0 .4 6269 6  .321260 .7 83955
1.40 .455940 .448485 .90 44 2 5
1.5 0 .45 0 593  .59 2 593 1.043 186
1.60 .4462 51  .75 882 4 1. 20 50 75
1.70 .44 26 47 .9540 74 1.39 67 21
1.80 ·439593 1. 18 7 879 1. 62 747 2
1.90 .43 69 60  1.474016 1.9 1097 6
2.00 .434 6 53 1. 833333 2. 26 79 86
2.10 .43 2601  2.299 0 99 2.731700
2.20 .43 0 753  2.928000 335 8753
2.3 0 .4 2906 7  3. 82 528 7 4. 2 54354
2.40 .4 27514  5. 211111 5. 6 3 862 5
2.5 0 .4 2606 9  7. 6 36 364 8.062433
2.60 .4 247 14 12.97777 8 13.402 49 1
2.70 .4 2 3433  34·453333 34. 876 766
2.80 .4 2221 5 71.100000 7 0 .6 77786
2.90 4 210 50 18.884848  18463799
3.00 .41 99 29 11.333333 10.913404
3. 10 .4188 4 8 8.3 0 5618 7. 886770
3. 20 .41 7 800 6.673333  6.255534
3.3 0 .4167 80 5. 6 52 288  5. 2 355 0 7
3.40 .41 57 86 4.953 1 91  4.53740 5
3.5 0 .414 81 5 4·444444  4. 02 96 30
3. 60 .4 1 3 86 3 4. 0 5757 6  3. 6437 1 3
3.70 .412 9 29 3·753443  3.340 514
3. 80 .412011 3.5 0804 6  3. 0 960 35
Appendix C 103
Sigma .0,1.00)
K{a, .0,1.00)
I Y{a, .0,1.00) Z{a,
Sigma K(a, 2.0, .00) Y(a, 2.0, .00) Z(a, 2.0, .00
Sigma K(a, 2.0, .00) Y(a, 2.0, .00) Z(a, 2.0, .00)
Sigma K(a, 100.0, .00) Y(a, 100.0, .00) Z(a, 100.0, .00)
Sigma K(O', 100.0, .00) Y(O', 100.0, .00) Z(O', 100.0, .00)
1[(11
(]
1.2 6.65000 (0) 6.66275 (0) 1.2 6.70180 (0) 6.76868 (0)
1.4 1.36500 (1) 1.37038 (1) 1.4 1.38736 (1) 1.41736 (1)
1.6 2.06500 (1) 2.07765 (1) 1.6 2.11842 (1) 2.18986 (1)
1.8 2.76500 (1) 2.78834 (1) 1.8 2.86366 (1) 2.98974 (1)
2.0 346500 (1) 3.50263 (1) 2.0 3.62160 (1) 3.80758 (1)
2.2 4.16500 (1) 4.22054 (1) 2.2 4.38964 (1) 4.63410 (1)
2·4 4.86500 (1) 4.94203 (1) 2·4 p6488 (1) 546244 (1)
2.6 5.56500 (1) 5.66694 (1) 2.6 5.94468 (1) 6.28849 (1)
2.8 6.26500 (1) 6.39504 (1) 2.8 6.72696 (1) 7.11011 (1)
3. 0 6.96500 (1) 7.12610 (1) 3. 0 7.51020 (1) 7.92647 (1)
K.(a, m1)
and the functions
Fe (a, m1 ) ,
F.(a, m1) •