See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/28601575

Exact analysis of a nonlinear partial differential equation of gas dynamics

Article  in  Quarterly of Applied Mathematics · May 1997

DOI: 10.1090/qam/1447575 · Source: OAI

CITATIONS READS

0 353

4 authors, including:

B. Mayil Vaganan
Madurai Kamaraj University
37 PUBLICATIONS   151 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Linearisation of nonlinear evolution equations View project

Cole-Hopf Transformations for Nonlinear Dispersive Equations View project

All content following this page was uploaded by B. Mayil Vaganan on 24 January 2017.

The user has requested enhancement of the downloaded file.

QUARTERLY OF APPLIED MATHEMATICS
VOLUME LV, NUMBER 2
JUNE 1997, PAGES 201-229

EXACT ANALYSIS OF A NONLINEAR PARTIAL

DIFFERENTIAL EQUATION OF GAS DYNAMICS

By
P. L. SACHDEV (Indian Institute of Science, Bangalore, India),
S. DOWERAHt,

B. MAYIL VAGANAN (Indian Institute of Science, Bangalore, India),

AND

VARUGHESE PHILIP (Regional Engineering College, Calicut, India)

Abstract. A new second-order nonlinear partial differential equation is derived from

one-dimensional unsteady non-isentropic gas-dynamic equations through the introduc-
tion of three "potential" functions. Appropriate boundary conditions at the shock and
at the piston in terms of the new functions are obtained. The nonlinear partial differen-
tial equation is analysed in great detail. Intermediate integrals and generalized Riemann
invariants are discovered. Using the classical Lie group method, the direct similarity
method due to Clarkson and Kruskal (1989), and equation-splitting etc., large families of
new solutions are found. The direct similarity method is found to yield the most general
results. Solutions with shocks (both finite and strong) are constructed to illustrate the
applicability of the solutions.

1. Introduction. The system of nonlinear partial differential equations describing

plane gas flows have been studied for a long time (see Courant and Friedrich [1]). A
typical physical problem is to determine the trajectory of a shock produced by a given
piston motion. The flow between the piston and the shock is in general non-isentropic.
Some of the earlier analytic studies on this topic used notions of self-similarity to reduce
the problem to that of a system of nonlinear ordinary differential equations subject to
boundary conditions at the shock and at the piston (see Sedov [2], Keller [3], and McVittie
[4]). This approach places severe restrictions on the form of the solutions. Prom all the
previous work it would appear that new exact solutions are difficult to find, at least in
the physical plane; one must seek other forms of the basic equations that would allow the
boundary conditions, both at the piston and at the shock, to be conveniently satisfied.
An early attempt in this direction is due to Martin [5] and Ludford and Martin [6], who
transformed the system of equations for unsteady, non-isentropic adiabatic flows with an

Received August 12, 1993.

1991 Mathematics Subject Classification. Primary 76N15, 76N20.
tSince deceased.

©1997 Brown University

201
202 P. L. SACHDEV, S. DOWERAH. B. M. VAGANAN, and V. PHILIP

arbitrary equation of state into a Monge-Ampere equation. In the latter study, Ludford
and Martin [6] found intermediate integrals both for isentropic and non-isentropic flows
and interpreted them as simple waves and generalized simple waves. A good qualita-
tive study of the mathematical solutions was given; however, no physical solutions, in
particular none with shocks, were found.
In another formulation, Steketee [7], [8] started with equations in Lagrangian coor-
dinates. He introduced functions that are similar to potentials and stream functions,
as well as their Legendre transformations. These arise naturally from the divergence
form of the Lagrangian equations. The forms obtained by Steketee [8] included those of
Ludford and Martin [6] and Smith [9] as special cases. He obtained generalized Riemann
invariants for the Ludford-Martin-Stanyukovich gas and illustrated one of the solutions
as a generalized simple wave. The second-order nonlinear partial differential equation
(see Steketee [8]), equivalent to the first-order system, has a rather simple structure so
that there exist solutions that are homogeneous in the Lagrangian coordinate and time
and have nice properties.
Ardavan-Rhad [10], in his study of the decay of a plane shock due to interaction with
a centered simple wave, derived from the basic equations of gas dynamics a second-order
nonlinear partial differential equation with particle velocity as dependent variable and
sound speed and entropy as independent variables. However, the intermediate integral for
this partial differential equation found by Ardavan-Rhad [10] involved only one arbitrary
function and so could not yield the exact solution of the interaction problem. The partial
differential equation derived by Ardavan-Rhad [10] is quite complicated and has yet not
been treated in any generality.
Ustinov [11] introduced a very interesting system of coordinates such that the shock
trajectory was given by a constant value of one of the coordinates while the piston path
and the particle trajectories were marked by constant values of the other coordinate. He
found a special class of solutions which described gas flows, headed by a shock into a
non-uniform undisturbed medium. These flows are produced by a certain piston motion.
Some flows without shocks were also found by allowing the shock strength to tend to
zero. This work was generalized by Sachdev and Reddy [12] who, using the method
of infinitesimal transformations, found several classes of solutions describing flows with
shocks, which either decay or grow in a finite or infinite time, depending on the density
distribution in the ambient medium; the shock strength, as measured by the pressure
ratio, however, remained constant. Sachdev and Gupta [13] dispensed with the similarity
assumption in the work of Sachdev and Reddy [12] and significantly enlarged the scope
of exact solutions. By adopting series solutions, the solutions obtained by Sachdev and
Reddy [12] were embedded in a larger family of solutions.
Ustinov [14], [15], [16] introduced yet another coordinate system, which arises from a
different set of conservation laws of the one-dimensional gas-dynamic equations. It turns
out that an equivalent total differential form gives rise to a second-order nonlinear partial
differential equation (see Eq. (2.20)), which enjoys considerable symmetries and admits
much analysis. While the dependent variable is one of the "potential" functions, it does
not have a simple physical interpretation; however, the physical variables such as velocity
and sound speed can be derived from it in a simple way. One of the independent variables,
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 203

equated to a constant, gives the streamlines while the other is again hard to interpret; for
strong shocks this second independent variable is zero along the shock trajectory. Thus,
the solutions of (2.20) express relations between three "potential" functions.
The purpose of the present work is to demonstrate that (2.20) admits a considerable
analysis. First, it is possible to find some intermediate integrals in a systematic manner;
these integrals, in turn, lead to explicit solutions. There are also exact solutions, which
are quadratics in one of the independent variables with the coefficients depending on the
other independent variable. This form of the solutions arises naturally from consideration
of the problem of a short intense impact of a piston moving under gas pressure. The
invariance of (2.20) under the group of infinitesimal transformations reveals several classes
of similarity solutions. Many of the solutions can be interpreted as flows produced by
piston motions and headed by strong shocks; there are also solutions that describe shocks
of finite strength. We apply the direct similarity method (see Clarkson and Kruskal [17])
to (2.20) to identify the most general similarity forms of solutions of (2.20). These
similarity forms involve five functions of which four are governed by a system of coupled
nonlinear ordinary differential equations while the fifth, the similarity function, is given
by a second-order nonlinear ordinary differential equation. It does not seem possible to
solve this system generally. However, some special solutions are obtained and shown to
contain those obtained by the other methods including the group-theoretic method [18].
The scheme of the present paper is as follows. The governing equations, boundary
conditions, and their transformations are given in Sec. 2. Intermediate integrals and
explicit solutions from them are found in Sec. 3. In Sec. 4 we obtain similarity solutions
using the group-theoretic method; in particular, we construct solutions that describe flows
with strong shocks or shocks of finite strength. In Sec. 5 more general similarity solutions
are obtained using the direct similarity method. In Sec. 6 we directly seek solutions that
are quadratic in one of the independent variables with coefficients depending on the
other. Section 7 gives solutions obtained by splitting (2.20). The results and conclusions
of the present study are contained in Sec. 8.

2. The governing equations, boundary conditions, and their transforma-

tions. The equations governing plane, adiabatic flows of an ideal gas are
dp d(pu)
= 0, (2.1)
dt dr
du du 1 dp , _
di+Ufr+pfr=°' ( ^
d(pp~y) d(pp~7) ,

where p, u, and p are density, velocity, and pressure, respectively; r is the spatial co-
ordinate and t is time. The constant 7 is the ratio of specific heats. It is possible to
manipulate Eqs. (2.1)-(2.3) so that, apart from (2.1), we have the conservation laws
d d
•^(H + ^(P + ^2) =°> (2-4)
9 ( 1 _ , 1 , 9 ( T x 1 „.3
dt\1-lp+ 2PU )+d-r[—lPU+2PU )=°- (2'5)
204 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

Equations (2.1) and (2.4) are equivalent to the differential relations

dr = pdr —pudt = mdy and d£ = pudr —(p + pu2) dt = —ma^ dx, (2.6)

where m and ao are positive constants, having dimensions ML~2 and LT~2, respectively.
Using (2.6) and the equation of state

/w " (T^ry <27)

we replace Eq. (2.5) by the total differential

dr) = ud^ + (fv — |u2) dr — ma^dz, v = p^~1^1. (2.8)

It is evident from (2.6) and (2.8) that x,y, and 2 are dimensionless variables. Now
we proceed to obtain a partial differential equation governing the function z = z(x,y).
Equation (2.8) implies the relations

dz 2 dz 1 p 1 2 ,
u = -ao^- and a0— = - -u , (2.9)
ox ay 7 - 1p 2

which, after eliminating u, give

dz 1 f dz\2
F-=a2{ 7-1) (2.10)
dy + 2 )

Differentiating (2.10) with respect to x and y, respectively, we obtain

1 dp _ p dp _ 2 (d2z | dz d2z \
p dy p2 dy a° ^ \ dy2 dx dxdy J

Differentiation of the equation of state (2.7) with respect to x and y, respectively, yields

I dp 1 dp
p dx 7 p dx' (2.13)
1 dp I dp m df
(2.14)
p dy 7p dy f dr'

On inserting (2.13) into (2.11) and (2.14) into (2.12), we get

I dp 2 / d2z dzd2z\ . \
p dx ^a° \ dxdy dx dx2 ) '
1 dp = 2 (d2z dz d2z \ my 1 df p
p dy ^a° \ dy2 dx dxdy ) 7 —1 f dr p
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 205

Solving Eqs. (2.6) for dt, we have

man , mu ,
dt = dx H dy, (2-17)
P P

so that
dp dp du
a0-=u- p—. (2.18)
dy ox ox
On using the first of (2.9), (2.18) becomes

dp dz dp d2z n
dy + frcfrx ~ d^P~
Substituting (2.10), (2.15), and (2.16) into (2.19), we arrive at a second-order nonlinear
hyperbolic partial differential equation

d2z dz d2z 7+1 ( dz\" 7 —1 dz d2z

2-
dy2 dx dxdy 27 \dx J 7 dy dx2
/ x 21 (2-20)
dz 1 dz\2
dy + 2\dx) '

where (m//)(d//dr) = <f>(y).

The Rankine-Hugoniot relations for a plane shock are

2U 2^p0
u = (2.21)
|Ti V7 T
= 2poU _ ^Lp0j (2.22)
7+ 1 (7+1)

(2.23)

where U is the shock velocity. The quantities with zero subscript pertain to ambient
conditions ahead of the shock. The shock locus is given by

Tt - u■ ^
The shock conditions (2.21)-(2.23) may be transformed to the x, y, z variables as follows:
We assume that the shock locus in the (x, j/)-plane is given by y = yo{x). On inserting
(2.21)-(2.24), the differential relations (2.6) and (2.8) yield

Upodt = mdy0, (2.25)

Po dt = mao dx, (2.26)

dt = maQ dz. (2.27)

7-1
206 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

Dividing (2.25) by (2.26), we obtain

dJ»=lE, (2.28)
ax ao

where we have used a§ = 7Po//°o- Eliminating dt from (2.25) and (2.27), using dp =
1Po/Po, and integrating the resulting equation subject to the condition z = 0 at y — 0,
we get

Z(X>y)\y=y0(x) = (2.29)

Equations (2.9), (2.21), and (2.28) lead to

Thus, along the shock curve y = yo{x) in the (x,y)-plane, the conditions (2.28)-(2.30)
must hold.
Now we consider the motion of a piston produced by a gas pressure, with the velocity
of the piston equal to the gas velocity V(t) there. Therefore, we have

m^ = -p, V(0) = u0, (2.31)

dt

where uq is the initial velocity of the gas. Integrating (2.31) and using (2.6) at y — 0, we
obtain
z( x,0) = -x2 —a:. (2.32)
2 ao
The strong shock condition is

*(0,i/) = 0, y > 0. (2.33)

Thus, the boundary value problem for the second-order nonlinear hyperbolic partial dif-
ferential equation (2.20) has been posed. The function z(x, y) must satisfy the conditions
(2.29)-(2.30) at the shock curve y = yo{x), and (2.32) at the piston y = 0. The third
condition would make it possible to determine the shock locus y — yo(x), so that the
problem is well-posed.
Since the sound speed a2 = 7p/p, Eq. (2.10) gives

a2 = 7(7 - l)oo(zs/ + \zl)- (2-34)

Using the relation a2 = 7p/p in (2.7), the density is found in the form

' = (2'35)
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 207

3. Intermediate integrals. We assume that there exists an intermediate integral

of Eq. (2.20) in the form
q = I(x,y,z,p), (3.1)
with the notation p — zx and q — zy. Differentiating (3.1) with respect to x and y to
obtain, respectively, d2z/dxdy and d2z/dy2 and using them in (2.20), we get

Iy + qlz + (Ix + plz){2p + Ip) - <t>(y)(q + ^p2^

+ Ip + 2plp -nlq+^p2)+p2
£=0, (3.2)
Therefore, the necessary conditions for (3.1) to be an intermediate integral of (2.20) are

Iy + Hz + (Ix + plz)i?p + IP) - <i>(y)(q

+ \v2) = 0, (3.3)
I2 + 2plp - n(q + ip2) + p2 = 0. (3.4)

Solving (3.4) for Ip and integrating with respect to p, we obtain

q= i F(x, y, z) ± ~^p
\p2, (3-5)

where F(x, y, z) is a function of integration. Replacing p by the first of (2.9) and inserting
the definition c = 00^/7(7 — l)\/<ir + (l/2)p2, Eq. (3.5) gives

u ± —F(x, y, z). (3.6)

7 - 1 y/Tl

For the isentropic case the Riemann invariants are given by (3.6) with F a constant.
Therefore (3.6) elegantly generalizes the definition of Riemann invariants to the non-
isentropic case. To obtain F(x,y,z), we substitute (3.5) in (3.3) and equate the coeffi-
cients of different powers of p to zero; we have
dF _ dF 1

/n \ dF Jn ,, .

and
Fz = 0.
Integrating the second of (3.7) with respect to x, we get

F(x,y) = ±Y^{y) + ^K{y), n0 = (3-8)

where K(y) is a function of integration. On inserting (3.8) into the first of (3.7) and
equating the coefficients of different powers of x to zero, we have

#+/o02 = O and ^+Zo0A' = O, l0 = ^ ■ (3.9)

ay ay n+ 2
208 P. L. SACHDEV, S. DOWERAH. B. M. VAGANAN, and V. PHILIP

A solution of the system (3.9) is

4>{y)= (Qo + loy)~1 and K(y) = 0o{ao + l0y)~\ (3.10)

where c*o and 0o are arbitrary constants. On inserting (3.8) and (3.10), Eqs. (3.5) and
(3.6) give the following intermediate integrals and the corresponding Riemann invariants
for the non-isentropic case:

1 0o ± n0x r
— ± Vnp
1= 7
&o + toy 2V, (3.11)
2c a0 0o + n0x
u± =—= —. (3-12)
7-1 i/n q0 + l0y

Now we rewrite the first integral (3.11), with plus sign, in the form

dz ( dz \ 2 X dz X2
0dY+ai\dXJ ~^YdX~ 4Y* = ' ( ' ^

where X = 0o + nox, Y = a0 + loV, "l = [n{2 - n)/(4(n + 2)2)], and 0\ = [n/(2(n + 2))].
Some solutions of the first-order partial differential equation (3.13) are obtained below.
1. We apply Charpit's method to Eq. (3.13). Writing the characteristic equations for
(3.13), we have

= 0iY~1X - 2otip, ~ = -l0, ^ = 0iY~lXp-2aip2 - l0q,

ds ds ds (314)

fds = -LY-2X-01Y-1p,
2 §■
as = 21-Y~3X2+01Y~2Xp,

where s is a parameter measured along the characteristic. A solution of the system (3.14)
is
v( x = 7+1
X<s> 7-1 c

Y(s) = S,

z(s) (7 - I)2 , 0
__cocl1ogs__cls+__c„s 7 - 1 2c , 7 + 1 2c-l I
+C4,
/ \ c—1 . (3-15)
p{s) = coS +cx,

«•) =

where Co, c\, C3, and C4 are arbitrary constants. An initial value problem may be posed
for (3.13) and solved with the help of the solution (3.15). Eliminating s from X, Y, and
p, we obtain

2frr&+*w-x-°'-0- <316»
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 209

Integrating the relation dz = pdx + qdy with p and q from (3.13) and (3.16), we obtain
the following solution of (2.20):

*(*,»)■=la»- 7 rx'+f I
2(37-1) 37-1 ^ V7 7
- 1
37 — 1 2 72 — 1 2
+ Cl^° ~ 2(7 — 1) ~~ 47(37- l)Cl

(3.17)
The sound speed is easily obtained by using (3.17) in (2.34):
-2
a2 = 7(7 - 1)c a0 - lo 7 ) • (3.18)
4(37-1) (37-1)'
The density is obtained from (2.35), (3.10), and (3.17):
g2/(7—i) / 7 \ (37-i)/(7-i)
P = 7l/h-l)(7- 1)7/(7-1) (^Q°- (37_l)yJ • (3'19)
2. Another distinct solution of (3.13) may be obtained by writing z(x,y) = Ymig(£),
£ = XY™2. Therefore, it follows from (3.13) that, if mi = 1, m,2 = —1, then g(£) satisfies
the ordinary differential equation

/2 2(n + 2) . 4(n + 2)2 / 1 1 9\ ,

9« + ^tf + _L_i^9+_{^=0, ,3.20)
A solution of (3.20) is
4(n + 2) l1/2
5(0 = c4 ±
71(71 — 2)

where C\ is a constant. Thus, the solution of (2.20) in the present case is

-1
7(7-1) 7 __2

-1
2C47 , „ a/7(7 - 1)
+ VFTIW^I) +(3^T)f
(3.22)
+ c' I °° ~ (3^1)») + 2^7)"° (°°" (3t~—T)"
± c4/30<
147(37- 1)
72 — 1
In this case, the density is again given by (3.19) while the sound speed is found to be

a2 = 7(7 - 1)a20
1 U+
7+ 1 v (37-l)V v v 37-!
2 (3-23)
±J 72-7
372 + 27 - 1
210 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

It is interesting to note that the solution (3.22) generalizes the corresponding isentropic
solution found by Ustinov [16]:

/ x 7 [(7 + \)x + 2]2 2 2 2

Z{X'V)~ (7+l)2 (7 + i)2/ + 27 (7 + 1)2- (3'24)
This solution describes a rarefaction wave between a certain moving piston and a leading
front.

4. Classical Lie group method. We now seek the following group of infinitesimal
transformations (see Bluman and Cole [18]) which takes the (x, y, z) space into itself and
under which (2.20) is invariant:

x*=x + eX, y* = y + eY, z*=z + eZ, (4.1)

where the generators X, Y, and Z are functions of x, y, and z. Invariance of Eq. (2.20)
under (4.1) gives

^yy ^yy)^y XyyZx "I- (Zzz 1YyZ*)Zy *2XyZzxZy YzzZy XzzZyZx

+ (Zz - 2YZ y) yZy 2 ZxJ ZyZxx ~ ZXZXX— 2zxz. 2 Xy Zxy

~~2>4>{y)Yzzy^Zy -(- —za/J 3 — YzZyZxx 3 Yzzyzxzxx 6YzzxZyZ:

&(y) X ZZx \ Zy
y "T~~ZZX
c\^x )J X zZXZyZXX~\
^z^x^y%xx — X zZxZXX ~\~2XzZxZXy
XzZxZxx ~\~2X ZZxZXy 2XzZyZ:
z^y-^xy
\ 2 J 7 27

+ 2 zx Zxy (Zyz XXy)%x H~ (Zxz YXy)Zy XyZZx (Zzz VyZ Xxz)z

YXZ%y X ZZZyZx YZZZXZy XyZXX ~l~{Zz Xx Yy)zxy

0(2/)Yx ^Zy -J- ^ Yx^yZxx ^x^x%xx 2YxZxZXy

2 J%zZxZXy (j)(y)YzZx (Zy -f- ~ZX J YzZxZyZxx ~f~ ^ Yzzxzx

~(~<2tYzZxZXy XZZyZxx 2YzZyZXy

2zXy[Zx -H (Zz Xx^zx Xzzx YxZy Yzzxzy\ H~ ^ zx[Zxx -b (2Zxz Xxx)zx

Z7

YXx%y ~i~{Zzz 1Xxz)zx 2YxzZxZy XZz%x Yzz%xZy {Zz 2Xx)zxx 2YxZXy

3 X Zzxzxx YzZyZxx 2 Yzzxzxy\

7 + 1_ n.\ „
+ Zx%xx vyyjZx \ZX + (Zz Xx)zx Xzzx YxZy Yz zx Zy]
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 211

Zy[Zxx ~l~{2Zxz Xxx)zx YxxZy ~i~(Zzz 2Xxz)zx 2YxzZxZy

7
Xzzzx
zz Yzzzxzy + {Zz 2Xx)zxx 2Yx^xy 3Xzzxzxx Yzzyzxx 2 Yzzxzxy\

7 - 1 ZXX "+~
%xx 0(?/) \^y (Z z Yy^Zy XyZX X z ZXZy Yz Zy]
7
1
- (t>'(y) Yyz + = 0. (4.2)
2

Equating the coefficients of zxzxx, zx, and zxx in (4.2) to zero, we get

Yz = Zzz = Zy = 0. (4.3)

The coefficients of zxzxx and zyzxy in (4.2) when equated to zero give

2XZ -Yx = 0 and 2XZ + -Yx = 0, (4.4)

7
implying that
Xz=Yx = 0. (4.5)
Again equating the coefficients of 4zx, zxzxy, zx, zxy, zxzy, zx, and zy in (4.2) to zero and
making use of (4.3) and (4.5), we get the determining equations

\<t>{y)(2Xx- 2Yy- Zz) - 2Xxy + ^~ZXX - I(t>'(y)Y = 0, (4.6)

2Xx -Yy-Zz= 0, (4.7)
2 Zxz — Xxx = 0, (4-8)
Xy-Zx = 0, (4.9)
2Zxz + (7 - l)Xxx = 0, (4.10)
Xyy - 4>{y){Xy
- Zx) = 0, (4.11)
Yvv + cp(y)Yy + <p'(y)Y+^±Zxx = 0. (4.12)
7
Using (4.9) in (4.11) we get
Xyy = 0. (4.13)
Equations (4.8) and (4.10) imply that

ZX2 = Xxx = 0. (4.14)

Equations (4.5), (4.13), and (4.14) lead to

X = k\xy + a\x + <i2y+ as, (4-15)

where fci, aj, 02, and 03 are arbitrary constants. In view of (4.3) and (4.14), (4.9)
integrates with respect to x to give

k\
Z = + a2% + (14Z + 05, (4-16)
212 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

where a4 and 05 are arbitrary constants. Substituting (4.15) and (4.16) in (4.7) and
integrating with respect to y we obtain

y = klV2 + (2ax -a4)y + a6, (4.17)

where a6 is a constant. Using (4.7), Eq. (4.6) can be rewritten as

\4>(y)yy+ - ~zxx + ~4>'(y)y = o. (4.18)

Using (4.15)-(4.17) in (4.18) and integrating with respect to y we get

= kiy2v;y>
+ (2a 1 - -a4)y + o6 , (4.i9)
where 0,7 is a constant. It is easily checked that Eq. (4.12) is automatically satisfied when
(4.15)-(4.17) and (4.19) are used in it.
On inserting (4.15)-(4.17), the invariant surface condition for z, namely, dx/X =
dy/Y = dz/Z becomes

dx dy dz
k\xy + a\X + a2y + a3 kiy2 + (2a\ — a4)y + ae (k\/2)x2 + a2X + a4z + a^
(4.20)
We now consider a few special cases of (4.20).
1. k\ = a4 = a2 = a$ = 0.
Integration of Eqs. (4.20) gives a similarity solution of (2.20) in the form

z(x,y) = F(ri), rj = °^X+p=. (4.21)

s/2axy + a6

Putting (4.21) in (2.20) we get the following ordinary differential equation for the simi-
larity function F(t}):

2 r//
^ 77'/' 7 + 1
Tj r —- a1r1F'F"+ a\F'2F"~ (2ax + y) F'2+ ^3 + ^ r]F' = 0. (4.22)

A local balance argument shows that we may seek a solution of (4.22) in the form

F(t,) = C2r,2 + ClV + C0, (4.23)

where the constants C2, C1, and Co are to be determined. Substituting (4.23) in (4.22)
and equating the coefficients of different powers of 77to zero, we get

7+1 2- o a7
c\ 7
a1C2 — 2«i ——
2
= 0, (4.24)

,7+1 2^2 7+1

aiC2-' 2 a\ -|- 4cti -f- 07 C2 + ^ + 4 = 0, (4.25)
7 7 a\

4l±la2C2-2 7+1
a\ -f- 4a\ -f- (J7 C2 + — + 3 = 0. (4.26)
7 7 ai
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 213

Equation (4.25) is satisfied if C2 = l/a\. Equation (4.26) then gives 0,7 = [3 —(2fa)\a\.
Equation (4.24) now yields C\ = 0. The constant Co is arbitrary. Therefore, (4.23)
becomes
F(V) = -r,2+C0. (4.27)
al

Substituting (4.27) in (4.21), we obtain the following solution of (2.20):

z(x, y) = -( -%f==) 2 + Co- (4.28)

al \ y/2ai y + 06/

To fit a strong shock to the solution (4.28), we assume 03 = 0 in (4.21). The strong
shock condition (2.33) and Eq. (4.21) imply F(0) = 0. It followsfrom (4.27) and F(0) = 0
that Co = 0. From Eq. (4.22) and F(0) = 0 we have F'(0) = 0. Thus the solution of
(2.20) satisfying the strong shock condition (2.33) is

z(x, y) =F(ri)= —. (4.29)

2a\y + a,6

2. fci = 02 = 0.
In this case, it is easy to verify that the solution of (2.20) has the form

*(*,») = ,(*,„) = |(2ai _ ' <4'30)

Let a4 = ai, a3 = a2 = 05 = 06 = 0. Then

z(x, y) = aixF(r]) and r) = x/y. (4-31)

Equation (2.20) now transforms to

air/2 - 7 + 1 a\r]2F' - 2a2rjF + 7 1 a\F2 + 7 + 1 a\r]2F'2 + af 7 + 1 rjFF'

7 27 27 7
+ 7 a\r] 1F2F' — 01 ^2ai
( 2a + y ^ VF'2 - 01 (4oi + a7)FF'
7
+ 2^^a\FF'2 + 1^1alr/F13 + (2oi + ar)r]F' - ^l^F2 = 0. (4.32)
77 1

The dominant balance argument shows that (4.32) has a solution that is linear in 77. It
is easily found to be

F(V) = 2~~7?- (4-33)

Thus the solution of (2.20) in this case is

x2
z(x,y) =2^. (4.34)
214 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

Now let r] = r/o be a shock of finite strength so that y = x/t]q = yo(x). Then, Eqs.
(2.29)-(2.30) give

(4-35)
Solving (4.35) for F'(r]o), we get
pi, \ _ 273?lo ~ 272t?q -37 + 1
770 oit?o7(72 - 1)
Equations (4.35) and (4.36) give r]0 = ^2/(7(7 —1)). The shock velocity as obtained
from (2.28) is
U = — y'0(x) = ao\ 1^——-= constant. (4.37)
7 V 27
3. fci = a,2 — <24—2ai = 0.
The characteristic equations (4.20) now integrate to give
z(x,y) =-^+e2aiy/a6F(r/), t)(x, y) = {aix + a3)e~aiy/a6. (4.38)
Z&i
In this case, (2.20) transforms to

772 -fa
, 27+1p,2
1——F — a\ 7 + 1 rjrw — o
2a\ 7 - 1 r„ F" + (2a, - F'2
27 7 7
(4.39)
— —3 J r)F' + 2)2- — )F = 0.
d\ J \ &i
It is easy to check that
Fiv) = + as, (4-40)
with ag a constant, is a solution of (4.39). Thus the solution of (2.20) as obtained from
(4.38) and (4.40) is
z(x,y) = -7^- + e2aiV + a3)2e~2aiy + a8^ . (4.41)

Now if we take 05 = <23= 0, the strong shock condition (2.33) and (4.38) with as = 03 = 0
imply that F(0) = 0. Therefore, it follows from (4.40) and F(0) = 0 that a§ = 0. Thus,
the solution of (2.20) satisfying the strong shock condition (2.33) is z(x,y) = a\x2/2. In
this case the sound speed and the density as obtained from (2.34)-(2.35) and (4.41) are
1 2/(7-1)
a2 = 27(7-1)(aoais)2 and P= _ 1)7/(7-y(-^)/(ai(^1)). (4.42)

4. k = 0, ck\ = ab, a2 = dk\, and b2 = 2lk\.

The characteristic equations (4.20) in this case are easily integrated. We get the
solution
. .(x + b/ki)2 x + b/ki
z(x,y) = », ,, s +F(tj , r)(x,y) = /, • (4-43)
2 (y + a/ki) y + a/kx
On substituting (4.43), Eq. (2.20) reduces to F" = 0. Therefore, F(rf) = Ao+ Ai77,where
Ao and Ai are arbitrary constants. Thus the solution of (2.20) in the present case is
. , (x + b/ki)2 , x + b/ki tJ.
z(x,y) = _L/ , \ + Ai - ' + A0- 4.44
2{y + a/k\) y + a/ki
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 215

5. Solutions by direct similarity method. Following the work of Clarkson and

Kruskal [17], we seek solutions of (2.20) in the form

z(x,y) = a{x,y) +/3{x,y)H{r){x,y)), (3{x,y)^0. (5.1)

We substitute (5.1) in (2.20) and require the resulting equation to be an ordinary differ-
ential equation governing the function H(rj):

{.H'2 + TiHH' + T2H' + r 3H2 + T±H + T5)H" + T6H'3 + {T7H + T8)H'2

(5.2)
+ (r 9h2 + r 10h + Tn)H' + r12#3 + r13tf2 + r14tf + r15 = o.

The functions Tra = r„(ry), n = 1,2,..., 15 are introduced according to

2px=0Ti, (5.3)
2u>axr/x+ (2 - n)rjy = u0r)lr2, (5.4)
3, (5.5)
2u)ax(3xr)x + 2(3xr)y - n(3yr]x = Ljbt2rfxTA, (5.6)
uotlrfc+ 2axr)xr)y- nayr]l + rfc= uf32r]*T5, (5.7)
Pvxx + 2/3*7?*= 0rgr6, (5.8)
ZPPxVxx + 4Plr]x + pf3xxr]x = /?27?3r7, (5.9)
ALo/3axr]xT]xx - 2n(3r]yr]xx + 4f3r]xT]xy + 8coax/3xr]x

-4(n - l)Pxr]xr)y+ Afjyril+ 2uf3axxTfx

- <p(y)f3r]2x
= 2u>p2ri^T8, (5.10)
PxiPPxVxx + 2P%r)x+ 2/3/3xxT]x)= (5-11)

Auj(3q.x(3xtjxx 2n(3(3yr)xx -H "I-

+ 4a>f3axpxxrjx + 4f3xrjy - 4(n - l)/3xf3yrjx + 4a)(3axx/3xrjx

- 2(j>{y)0Pxr]x
+ Af5(3xyr)x
- 2nf3(3xxr)y= 2a;/?3r/^ri0, (5.12)

2u)(3axr)xx-\-2n(3ayr]xx ^,PVyy~^~/^P^x^lxy~^~^^^xPxVx

~~^~^Qx0xVy
"I"^&xfiyT]x ^PQix&xxVx ^HQyflxVx
+4(3yr)y + 4:(3axyr)x-2n(3axxrjy - 2(/)(y)P(axrjx+riy) = 2u/33r)xTlu (5.13)
p2xpxx = /33^r12, (5.14)
Aiuaxpxpxx+2LdaxxPl+Apxpxy-2nPyPxx-(t)(y)pl = 2loP3t]xTi3, (5.15)
2aJOixfixx"1"^^^x^xxPx ^Q-xfixy 2flOLy(3XX
H-4flxOixy
—2naxxf3y
+ 2f3yy- 2</>{y)(ax/3x
+ (3y)= fafPrfcYw, (5.16)
^^yy~^~4c^a;^xy~\~e^^^-x^-xx
2?i0Cy0LXx0(l/)(^x~\~2ocy) — 2idf3 TJxT\^, (5.IT)

where uj = (7 + l)/(27) = 1 —n/2.

216 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

Now with the help of the following three remarks we solve Eqs. (5.3)-(5.17) for the
functions a(x,y), /3(x,y), r](x,y), and rn(r?), n = 1,2, ...,15. Remarks: (1) If a(x,y)
has the form a(x,y) = a(x,y) + (3(x,y)£l(r7), then we may choose 17 = 0. (2) If f3(x,y) is
found to have the form f3(x,y) = (3(x,y)tt(ri), then we may put Q(r}) = 1. (3) If r)(x,y)
is determined from an equation of the form /(r?) = fj(x,y), where f(r)) is any invertible
function then we may take 77without loss of generality.
Setting 1^(77) = 2Q'1(r])/n1(r]) in (5.3) and integrating with respect to x leads to
P = /3(y)Qi(r)), where j3(x, t) > 0 is a function of integration. Using Remark 2 we choose
r^i (77)= 1 so that
(3= (3(y), ' (5.18)
where we have dropped the hat on (3. Putting Te(r]) = —(77)/(77) in (5.8) and
integrating twice with respect to x yields S76(77)= x9(y) + ip{y), where 9(y) and ip{y)
are two functions of integration. In view of Remark 3 we choose ^(t?) = V so that
rj(x,y) = x6(y) + ip(y). Hence we obtain

r}(x,y) = x0(y) + ip{y). (5.19)

Putting P2(?/) = 2£l'2{ri) in (5.4) and integrating with respect to x yields a = (30,2 —
(x/(29))(x9' + 2ip') + A(y), where A(y) is a function of integration. Using Remark 1 we
choose f22(?7)= 0. Thus we have

a(x,y) = Vj-j[x9'{y) + 2ip'{y)] + X(y). (5.20)

It is now easily seen from (5.3)-(5.5), (5.8), (5.9), (5.11), (5.12), (5.14), (5.15), and (5.18)
that
r„(77) = 0, n= 1,2,3,6,7,9,10,12,13. (5.21)
Equation (5.6) requires that T^) = k = constant so that

P' = —uJok92(32, ujq = uj/n. (5.22)

On inserting (5.19)-(5.20), (5.7) becomes

(x9' + ip')2 - 9{x29" + 2xip") + 9'{x29' + 2xip') + 292\' = —2c^0/326»4r5(77). (5.23)

Since the left-hand side of (5.23) is quadratic in x, the function r5(77) must assume the
form r5 (77)= arj2 + 677+ 0, where a, 6, and c are constants; with this form of Ts^), Eq.
(5.23) is satisfied provided

99" - 29'2 = 2cllo0I3296, (5.24)

ip" - 29~l9'ip' = L00p294(2atp + b), (5.25)
26>2A'+ ip'2 = -2 io0p294{aip2 + bip + c). (5.26)

Equation (5.10) requires that r8(?7) = I so that

<p{y)= 2(2 - uj)9~19' - 2w(2k0 + l)92(3, k0 = k/n. (5.27)

 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 217

It is easy to verify that, in view of (5.24)-(5.25), Eqs. (5.13) and (5.17) are satisfied
provided that Tn{r]) — (—l/n)(2ar) + 6), T15(77)= —2a>o(2k/n + l){arj2 + brj) + d, and

—26~16'X' = {2kuj%{a'ip2
+ bxl>)
+ 2uu)O{2ko+l)c+ujd}O4l33-2ujol3202ip'{2ailj+b),(5.28)

where d is a constant. Equation (5.16) requires that r14(77) — h = constant, implying

that
j3" _ 26T16>'/3'= [hto+ (2ioco0k){2k/n+ l)}04/33. (5.29)
Thus we have derived six equations, namely (5.22), (5.24)-(5.26), (5.28), and (5.29),
governing the four functions 9(y), 0(y), ip(y), and A(y). If the constants h and d
are chosen to be h = (—2kwo){k/n + I) and d = (—2ivo)(k/n + l)c, then this over-
determined system becomes determinant in the following sense: Equation (5.29), with
h = (—2ku)o)(k/n + I), may be obtained from (5.22) by once differentiating the latter
with respect to y, similarly, Eq. (5.28), with d = (—2coo)(k/n + l)c, may be obtained from
(5.26) by once differentiating the latter with respect to y and using (5.22) and (5.25).
Thus, the solution (5.1) of (2.20) has the form

z(x,y) = A(y) - [x9'(y) + 2ip'(y)] + (3(y)H{r)), (5.30)

r)(x,y) = x6(y) +tp(y), (5.31)

where the functions (3(y), 9(y), ip{y), and A(y) are governed by the determinate system
of ordinary differential equations (5.22), (5.24)-(5.26).
Substituting for (77),n = 1,2,..., 15, thus found, in (5.2), we obtain the following
second-order nonlinear, non-autonomous ordinary differential equation for

(H'2 + kH + ar]2 + br] + c)H" + IH'2 - i(2arj + b)H'

—2kcoo(ko + l)H —2uj0(ko + l)(ar]2 + brj + c) = 0. (5.32)

We may seek a solution of (5.32) in the quadratic form

H(ri) = Art2 + BV + C. (5.33)

The constants A, B, and C are given by the algebraic equations

4A3 + {21+ k)A2 + cl>(q,

—/cq —lko)A —aujo^ko-i-1) —0, (5.34)
[4A2+ (2l + k)A —a/n —kkoUo—kluJo\B+b(l —l/n)A—bLL)o(ko+l) = 0, (5.35)
2k(A —koujQ—Iujq)C (2A -f-l^B~ —{b/vi)B-\-2cA —2c^o(^o~t~0^ = (5.36)

A solution of the algebraic equations (5.34)-(5.36) will be obtained for certain simple
cases.
Now we return to the system (5.22), (5.24)-(5.26). Equations (5.22) and (5.24) are
coupled and nonlinear and once these are solved for the functions f3(y) and 6(y), then
(5.25) for ij}(y), being linear and inhomogeneous, may be solved. The function X(y) is
218 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

then easily obtained from (5.26). Some simple solutions of this system are obtained below
and the solution of (2.20) is written out in each case:
1. A solution of Eqs. (5.22), (5.24)-(5.26) is easily found to be

%) = m0(miy + m2)ni, P(y) = nm'fc^2+ ^ (miV + m2)~2"1~1!

Um° (5.37)
,f \ ,/0 \( \ (bz - 4ac)nm1(2n1 + 1) ,-2ni-i
^(y) = -b/2a, A(y) = —^ (mlV + m2)

where mo, mi, and to2 are arbitrary constants and nj satisfies the quadratic n2 + ni +
(2an/(8an + u>k2)) = 0. On inserting (5.37), Eqs. (5.30)-(5.31) give the following simi-
larity form of the solution of (2.20):

z(x,y) = 44ao;Fm^ni + ^ ^miV + TO2)~2ni_1 ~ ^nimi(mi2/ + m2)-1x2

r](x, y) = m0(miy + m2)mx - 6/(2a).

(5.38)
Equations (5.33) and (5.38) now yield the solution

—n 1 n . 9 ,
z(x,y) = mx + -—(1 + 2ni)m0^4 (m-iy + m2) 1a:2
2 kto

+ 7^(1 + 2ni)TO„ --A + B (miy + m2)~ni_1a; (5.39)

kcu a
b2A bB +C„ —b2 —Aac n ■- * 2„, \-2m-i
+ , (1 + 2ni)m0mi(m1y + m2)
4a2 a 4ak kui

The function cp(y) is given by

4>(y)= 2mi[ni(2 —w) —k^l{2ko + l)(l -I-2m))(m\y + m2)_1. (5.40)

2. b = c = 0.
In this case, the solution of the system (5.22), (5.24)-(5.26) is given by

8{y) — m0e_miJ/, 0(y) = i%) = d0e"miy,

w , , d?,m? / 8an\
(5.41)
A(?/)" A°" ( +^ ^

where too, toi, do, and Ao are arbitrary constants. Using (5.41) in (5.30)-(5.31), we
obtain the following similarity form of the solution of (2.20):

, , , dnm? 8an\ 1 2 2 m^do 2nmi 2m,VTT, N

^•9) = A»- iqf ('+ rf) o+2 + (5.42)
r)(x, y) = (m0x + d{))e~miy.
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 219

Thus, the solution of (2.20) as obtained from (5.33) and (5.42) is

»-*-^ -if 0+s)»- m 1

do - ^-(2d0A + BemiV) x + mi I
1 2nA\ 2
(5.43)
+ — — a; .
m0 (OK w/c /

Here, the function (j>(y)= 2mi (uj + IkQ1).

With a\ — m0, m\ = a\/as, Ao = —05/(201), 8an + uk2 — 0, and 07 = 2(w + 2 +
2/fcp 1)oi, we recover the similarity form (4.38), obtained by the group-theoretic method
provided that
='moFM + ~-V2- (5-44)
3. k —0 and b2 = 4oc.
It is easy to check that the solution of Eqs. (5.22), (5.24)-(5.26) is now given by

%) = m0(miy + m2) 1/2, f3(y)= ~ J

tUq V °au; (5.45)
V>(j/)= -6/ (2a), A(y) = A0,
where mo, mi, m2, and A0 are arbitrary constants. Putting (5.45) in (5.30) (5.31), we
get

z(x,y) = A0 + |mi(mij/ + m2) 1x2 + ^k\ ~~H(r)),

4 tUq V Sao; (5.46)
r/(x, y) = m0{miy + m2)~1/2x - 6/(2a).

On inserting (5.33), Eq. (5.46) gives the solution

1 / n (miy + m2) xx2

z(x,j/) - mi
4^ V 8aw
bA +B„
—
+ m0mi (5-47)
a
r62A ^ n
+ A0 + m§mi 7T + <?
4o2 a 8auj

In this case, the function <j>(y)is given by

<t>(y) = ~rn1 2 —u>+ (2fco+ /) 1 (mi2/ + m2) l. (5.48)

4. k = b —c = 0.
Here, the solution of the system (5.22), (5.24)-(5.26) is found to be

mi I n
%) = merrily + m2) 1/2, /%) =
771qV 8acu'
(5.49)
V'(y) = do(miy + m2)"1/2, A(y) = ^Jr(™i2/ + m2)~\
220 P. L. SACHDEV, S. DOWERAH. B. M. VAGANAN, and V. PHILIP

where mo, mi, TO2, and do are arbitrary constants. On using (5.49), Eqs. (5.30)-(5.31)
yield

z(x,y) = -^—T]2 +
y ,yl 2too' too V 8^ (5.50)
T) = {m0x + d0)(miy + m2) 1/2.

Equations (5.33) and (5.50) now lead to the solution

1 I n
z(x,y) = toi (m1y + m2) 1x2
4 2 V 8aw

^d0m0mi ^ (miy + m2)-1^ a; (5.51)

1 TOqTOi
— 2 - Wo
—
+ do (TOiy + m2)_1 + C,

where C is a free constant. The function <f>(y)appearing in (2.20) is now found to be

,n /nw
<t>{y)= -mi / —— (miy + m2)
2 —ijJ+ (2fco+ Z)■* . (5.52)

With mi = 2to0, do = a3, to2 = a6, and a7 = —2ai[2 + !wi/(-w + n/(8aw))], we

recover the similarity form (4.21), obtained by the group-theoretic method provided that

f(v) = 7^-v2
2mo + —
too \r~HW-
V (5-53)

5. o = b = c = 0.
In this case, the solution of Eqs. (5.22), (5.24)-(5.26) is given by

8{y) = TO0(miy + TO2)~1, /3(y) =--^~(miy + m2), ip(y) = X(y) = 0, (5.54)

where Too, mi, and to2 are free constants. On inserting (5.54), Eqs. (5.30)-(5.31) give

z{x,y) = r^-(rnly + m2)~1x2 - nmi2(miy + m2)H(r)),

2 ukm£ (5.55)
r]{x,y) = mox(miy + m2)~l.

Thus, the solution of (2.20) is now given by

*(*,»)="». Q - +m„rV - - ^(miy + ma), (5.56)

 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 221

where A satisfies the quadratic 4A2 + (2l+k)A —kuio(k/n+l) = 0 and C = —(l/(2fc))[(2/+

A)/{A — uio(k/n + 1)}\B2; the constant B is arbitrary. The function (f>{y) — 2mi[w +
kg 1l](m1y + m2)_1.

With mo = mi = 1, m2 = 0, and 07 = 2/cq 1(u)k0 + l)a 1, we get the similarity form

(4.31), obtained by the group-theoretic method, provided that

1 77 1
a\F{rj)=-r] t~h(v)- (5-57)
2 uik rj

6. a = 0.
In the present case, we obtain two distinct solutions of the system (5.22), (5.24)—(5.26):
(i) One solution of the system (5.22), (5.24)-(5.26) is given by

0{y)=mQ, /3(y) = (miy + m2)~1,

Tib
VKV)= di + d0y - log(miy + m2),

.. , , <& rim i ( nb2 \ . . ,

A(s)= ~ +

+ "^2 (rfoV+ rfi)(miV + w2)~1 (5.58)

- (TOly + TO2)_1 log(miy + m2),

where mo, mi, m2, do, di, and Ao are arbitrary constants. Therefore the similarity form
of the solution of (2.20) as obtained from (5.30)-(5.31) and (5.58) is

. d?, nm\ ( nb2 \ ,

Z(X'V) = " + UkV„ (C " ) (mi» + m2>

+ y + <h){mxy + m2y1
lo
m\n2b2
^^2 (miy + m2) 1\og(mxy + m2) (5-59)
U!

1 ( , nbmi
No - (<m1y + m2) ^ x + (mi?/ + m2) ^(jj),
mo V UJ^2

r/(x, y) = m0x + di + d0y - ^ log(mij/ + m2).

222 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN. and V. PHILIP

Putting (5.33) in (5.59), we obtain

. , nmi A. , 2
z{x,y) = — (miy + m2) x
u>k
nmi (b „ 2bnA, . . „ ,. , ,.
+ 7 + B — log(miy + m2) + 2^(d0y + dx)
Lokmo uk2

- — (m1y + m2) (mi?/ + m2)~lx

too J
nmi r*
1 /' ' ' \
n&2 b
+ k{C'2^) + k{doy + dl)
ujkrriQ
nb2 n2b2A 2
" Zkhn2 l°g{miy + m2) + ^T[1°g(miy + m2)1
2,TibA
+ A(d0y + d\)2 - (doU + di) log(miy + m2)

log(miy + to2) + B(d0y + dx) + C (m\D + m2) 1

dl-~y + A0.
2toq
(5.60)
The function <^(y) = —{2nm\/k)(2ko + l)(rri\y + m2)_1.
(ii) Another solution of the system (5.22), (5.24)-(5.26) is found to be
777771
0(y) = m0(mi2/ + m2) , 0(y) = ~^fcm2 i171^ + m2)>

ijj(y) = di - —(miy + m2)~l + log(miy + to2),

777-1 ^
dn ,_i nmj / Jib2 \ , (5.61)
A(i/>= A°+ 2^5(m,!/+m2) + (2^ - -c) <""»+
m\n2b2 ,
- ^2fc4m2(ml^ + TO2) log(TOiy + TO2),

where mo, toi, m2, do> di, and Ao are arbitrary constants. On inserting (5.61), Eqs.
(5.30)-(5.31) give

777771 /
z(a:,j/) = A0+ 0J^° 2 {m\y + m2) 1 + "^''2 f - 6di - c ) (mjj/ + to2)
2t77i777qV *Ay ' ' u;/c2777q \2cj/c2

777l77^6^
ui2kAm2 ^miy + m^ log^mi2/ + m2>

1
([-mam^x2 + 2d(ix}(miy + mi) 2+ fj (5.62)
2mt
nTOl / \ TT / \
^(,nlV + ,n,)H(m,

rj(x,y) =to0(toiJ/ + to2) xx + dx - — (miy + to2) 1+ log (mi 2/ + to2).

7771 (j/t
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 223

Substituting (5.33) in (5.62), we obtain the following solution of (2.20):

z(x, y) = m 1 Q - (mi y + m2)~1x2

+ - ^(miy + m2)~l + 2diA + B + log(miy + m2)^j

+ ^o (S?-1)(ml!' + m2^,

+ z^(^-bdl^c+d'A+d'B+c),mty+m2)
bn2 (
+ ^^%^(v2d0^41og(miy+m2)

—mi (6 + 2d\A + B)(miy + m2) log(m\y + m2)

- ~^p~(TOiV + rn2)[log(miy + m2)]:

+ Z^k{2dlA + B) + X°-
(5.63)
The function <j>(y)in (2.20) is now given by <f>(y)= 2m\(w + kQll)(m\y + m2)_1.
7.a = b = c= k = 0.
Here, the solution of Eqs. (5.22), (5.24)-(5.26) is found to be

0(y) = m0(miy + m2)~\ (3(y) = du t/j(y) = d0(miy + m2)~l,

w , mxdl, (5.64)
Ky) = -^2(rn\y + rn2) ,

where my, toi, to2, do, and di are arbitrary constants. In this case, the similarity form
of the solution of (2.20) as obtained from (5.30)-(5.31) and (5.64) is

z(x,y) = ^\(m0x + doftm^y + m^'1 +dtH(ri),

2mo (5.65)
r)(x, y) = (m0x + da)(miy + m2)_1.

Equations (5.33) and (5.65) give the solution

z(x>y) = \(mi- ld\ml(m\y + m2)_1)(mxy + m2)_1i2

2
domi
+ + m0dx (-ld0(miy + m2) 1 + B) (miy + m2) xx (5.66)
m0

+ + d^di (-^(miy + m2)_1 + B^j (miy + m2) 1 + C,

224 P. L. SACHDEV, S. DOWERAH, B. M. VAGANAN, and V. PHILIP

where B and C are arbitrary constants. The function (f>{y)is found to be

4>{y)= —2mi(2 —u))(m\y + r^)-1 —2codiml(2ko + l)(miy + m2)-2. (5.67)

The solution (4.44) obtained by the group-theoretic method is easily deduced from
(5.66) if we choose mo = m\ = d\ = 1, I = 0, and 2B = —d0 in the latter. In the present
case, F(r]) = H(r)).

6. Solutions quadratic in x. The piston driven flows characterised by the boundary

condition (2.32) may be described by the solution

z(x,y) = F0(y) + G(y)x + \H{y) x2. (6.1)

Substituting (6.1) in (2.20) and equating the coefficients of different powers of x to zero,
we get the following system of nonlinear ordinary differential equations:

H" + [(4 - n)H - (j)]H' + (2 - n)H3 - (f>H2= 0, (6.2)

G" + [(2 - n)H - 4>}G'+ [(2 - n)H2 - <j>H
+ 2H']G = 0, (6.3)
F," - (nH + - \ [<f>
- (2 - n)H}G2 + 2GG' = 0. (6.4)

We choose cf>= 4A0H, where Aq is a constant. We may then solve (6.2) to obtain

H = (Co + Boy)'1, Bo=u-2Ao, (6.5)

where Co is a constant. Now, (6.3) is easily solved:

G(y) = D0(C0 + B0y)m\ m1=-l-2B^. (6.6)

A solution of (6.4) is
F0(y) = lD2(Cf) + B0y)2rn^1. (6.7)

Thus, the solution of (2.20) in the present case is

z(x,y) = \D2{C0 + Boy)2nil+1 + D0(C0 + BQy)mix + |(C0 + Boy)~lx2, (6.8)

and the sound speed and the density are given by

a2 = 'y^y — l)ao[|£>o[l + |(2n2 + l)i?o](Co + Boy)2"2

+ Dq{ti2Bq + l)(Co + Boy)"2 1x + ^(1
2V. — Bq)(Cq + Boy) 2x2],
~ J,

a3"-'-"
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 225

7. Solutions by equation-splitting. In this section we obtain two solutions of

(2.20) by splitting it into two equations and hence solving them.
1. We split equation (2.20) into

„2
u>zx —nzy = 0, (7.1)
1 ,2
Zyy + 2ZXZXy (fiiy) + 2 ZXJ — 0- (7.2)

Manipulating equations (7.1)-(7.2) we get

^yy 2kJi^(t/)Zy — 0, uj\ — l/(2u^). (7.3)

We may write (7.3) as

9 (* 1 (, , 1
dy \Zx u0Zy)~ 2Ul fZy V1 + o;0) Zyv~ (7'4)

Equations (7.1) and (7.4) require that

2uJlf-1f'Zy - (1 + l/Uo)Zyy = 0- (7-5)

Integrating (7.5) twice with respect to y, we have

z = v(x) J fa2(t)dt + w(x), a2 = 27/(37 - 1), (7.6)

where v(x) and w(x) are functions of x. Substitution of (7.6) in (7.1) leads to the
determination of the functions v(x) and w(x):

v(x) = (bnx + b2)2, w(x) = b0bix2 + 2bib2x + b3, (7.7)

where bo, b\, b2, and 63 are arbitrary constants. The function / is obtained from (7.1),
(7.6), and (7.7):
/1 \ (~37+l)/7
• <7-8)

where A is a constant. Thus the solution of (2.20) is

z(x>y) = 2A~^(boX + b2^ (2~ 2uJ°/2Ay) + bohx2+ 2b1b2x + b3, (7.9)

In the present case the equation of state (2.7) is

/ 1 \_37+1

^-^y) • <"»>

2. Another solution of (2.20) is obtained if we split it as

nzxx + <t>(y)= 0 and zyy + 2zxzxy - ^(j)(y)zl = 0. (7-11)

226 P. L. SACHDEV, S. DOWERAH. B. M. VAGANAN, and V. PHILIP

Integrating the first of (7.11) twice with respect to x we get

z(x,y) = — ^(p(y)x2 + V{y)x + W(y) (7.12)

n

Substitution of (7.12) in the second of (7.11) leads to the following system of nonlinear
ordinary differential equations for V(y), and W(y):

n2(t)"- AnW + 1<j?= 0, (7.13)

nV-2n-f(^')+2^
dy
= 0, (7.14)
n2W" —2nVV' + 4>V2= 0. (7.15)

The solution of the system (7.13)—(7.15) is

j./ \ = 7 ~ 1 P
4>(y) 7 a + f3y
V{y) = b4(a + f3y)~l +b5,
,2
W(y) = (b6y + b7) + ~^{a + f3y)[~1 + log(a + f3y)] ^716^
(7- 1) P
- + (3y) 1

where 64, 65, bg, and bj are free constants.

Thus the solution of (2.20) in the present case is

z(x,y) = — T- 1 P xJ2 + (64(a +/3y)

fl.A-l +65)2;
n 27 a + (3y
7 6?
+ (&6y+ b7) + — (a + /3y)[-l + log(a + 0y)
(7- 1) (7.17)

—^pb24{a + Py)-1

In this case, the equation of state (2.7) becomes

p = (7-1)7(q + /3j/)1_7/97. (7.18)

8. Results and Conclusions. Equation (2.20) derived in Sec. 2 from the conser-
vation form of the one-dimensional gas dynamic equations (2.1)-(2.3) does not seem to
have attracted much attention. This complicated partial differential equation, with sev-
eral linear and nonlinear terms and with a function 4>{y) of the independent variable
multiplying two of them, does not appear to possess any symmetries. Yet, the details of
Sees. 2-7 show that it allows considerable analysis, symmetries, and several families of
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 227

solutions. It also has first integrals (3.11) which allow a generalization (3.12) of the Rie-
mann invariants. One of the first integrals is then treated as first-order partial differential
equation, yielding the solutions (3.17) and (3.22).
Equation (2.20) is a "good" test equation to compare which of the two approaches
exploiting symmetries—the classical Lie group method and the direct similarity approach
proposed by Clarkson and Kruskal [17]—leads to more general similarity solutions. The
similarity form (5.30)-(5.31) of the solution of (2.20), obtained by the direct similarity
method, seems to represent a much larger family of solutions. The similarity function
appearing in the form (5.30)-(5.31) of the solution, is given by the nonlinear
ordinary differential equation (5.32) while the other functions involved in the similarity
form, namely, 9{y), (3(y), ip(y), and A(y) may be found from the determinate set of
nonlinear coupled ordinary differential equations (5.22), (5.24)-(5.26).
Even though it does not seem possible to analytically solve the determinate system
(5.22), (5.24)^(5.26), and (5.32), it may be solved numerically, if suitable initial condi-
tions, pertinent to them, can be identified. The generality of this system is confirmed by
the fact that all the solutions obtained through the classical Lie group method can be
recovered as special cases of this system. The solutions obtained by the two methods ap-
pear distinct: each of the ordinary differential equations (4.22), (4.32), and (4.39) when
compared with Eq. (5.32) obtained by the direct similarity method and governing the
similarity functions, appear different, but, as we have shown in Sec. 5, a simple change of
variables (see (5.44), (5.53), and (5.57)), reconciles these apparently distinct equations.
There are five distinct classes of solutions obtained by using the direct similarity
approach. For all five classes, the solutions have the form

z{x,y) = Q0(y) + Qi(y)x + Q2{y)x2,

with Q2(y) = cr(miy + m2)^1, where a, mi, and m2 are arbitrary constants. The
functions Qi{y) and Qo(j/) have the following five distinct forms:
Class (1):

n
My) = + 2n{)ml --A
a
+B (miy + m2) — Til ~ 1

b2A bB b2 —4ac
Qo(y)~- 4a2 a Aak
^(1 + 2ni)m2)m1(miy + m2) 2ni

Class (2):

r\ i \ mi
Qi{y)= —
m0
d0-~{2d0A
UJK
+ Bemiv)

,, , . . 2nmidr)A dim2, ( 8an\ 2nm1o?n,r, „

Q'{y>
= A»- " M V + ^J^taT(s +Ce }<
228 P. L. SACHDEV, S. DOWERAH. B. M. VAGANAN, and V. PHILIP

Class (3):

nm\
Qi(y)= ujk2mi
b(miy + m2)~1 + 2dyA + B + —— log(miy + m2)

dQ (2nA \ ,

-1
Qoiv)- — 2 f ^ - -7-7 I (mij/ + m2)
771x7710\2 Ujk2 J
77,7721 X Til)2 \

bn2
+ ^2fc4m2 (2<Mlog(miy+m2) - mi (6+2di,4 + .B)(miy+TO2) log(mij/+m2)

bnm\A, ... . ,,9. ndn . , ,

ujk2^ V + m2)[log(miy + m2)] ) + ~^j^(2diA + B) + A0.

C/ass (4):

^to. / 6 26nA, . ,.,

Qi(y)= ujkm0 \ k
7 + B -y- log(miy + m2) + 2A(d0y + di)
uk2
1
- —(toij/ + m2) (miy + m2)_1,
m0

^ nm\ [1 / n&2 \ b., , .

Q»(l,)= Jta! .fc(C" + +
nb2 . n'2b2A 2
log(miy + m2) H ^rr[log(miy + m2)J
wfc3m62o u;2fc4

+ .4(d0y+ dj)2 - -j^r(doy + dl)\og(m1y + m2)

log(miy + m2) + S(d0y + di) + C (m!y + m2) 1

dg
2^f+ V
Ciass (5):

Q\(y)= d(,mi + m0di (-ld0(m\y

rn0
+ m2) 1 + 5) (mij/ + m2) \

Qo(y)= +d0di ^-^(toi2/ + to2) 1 + (miy + m2)_1 + C.

Other approaches—solutions from the first integrals and the assumed quadratic (in x)
form of the solution in Sec. 6—do not lead to any new solutions; these are, in fact, special
cases of the general form (5.30)-(5.31) obtained from the direct similarity approach. The
only exception is the adhoc approach which splits Eq. (2.20): it leads to one new solution,
namely (7.17), not obtained from any other approach.
 NONLINEAR PARTIAL DIFFERENTIAL EQUATION OF GAS DYNAMICS 229

Referring now to the boundary conditions for these solutions, arising from the physical
problem, namely, (2.29)-(2.30) at the shock curve y = yo(x), and (2.32) at the piston
y — 0, we have given a few illustrative examples, (see (4.29), (4.34), and (4.41) with
«3 = 05 = a,g = 0). The solutions (4.29) and (4.41) with 03 = 05 = as = 0, satisfy the
condition (2.33) for a strong shock while the solution (4.34) satisfies the conditions (2.29)-
(2.30) for a shock of finite strength. It may be possible to satisfy boundary conditions
at the shock for other classes of solutions that we have found. More general solutions
with strong shocks can be obtained by solving the ordinary differential equations such as
(4.22), (4.32), and (5.32)) subject to initial conditions. For example, for Eq. (4.22), we
find that F(0) = 0 and F'(0) = 0 so that a numerical solution of (4.22) may be found
using these initial conditions.
We conclude that the nonlinear partial differential equation (2.20) arising from the con-
servation form of one-dimensional gas-dynamic equations displays many symmetries and
possesses a great variety of solutions. Other partial differential equations for "potential-
like" functions (see Steketee [8] and Ardavan-Rhad [10]) arising from one-dimensional
gas-dynamic equations may be treated by the methods expounded in the present paper.

References
[1 R. Courant and K. O. Friedrichs, Supersonic Flow and Shock Waves, Interscience, New York, 1948
[2 L. I. Sedov, Similarity and Dimensional Methods in Mechanics, Academic Press, New York, 1959
[3 J. B. Keller, Spherical, cylindrical and one-dimensional gas flows, Quart. Appl. Math. 14, 171-184
(1956)
[4 G. C. McVittie, Spherically symmetric solutions of the equations of gas dynamics, Proc. Roy. Soc.
London Ser. A 220, 339-355 (1953)
[5 M. H. Martin, The propagation of a plane shock into a quiet atmosphere, Canadian Journal of
Mathematics 5, 37-39 (1953)
[6 G. S. S. Ludford and M. H. Martin, One dimensional anisentropic flows, Comm. Pure Appl. Math.
7, 45-63 (1954)
[7 J. A. Steketee, Unsteady rectilinear flows of a non-homentropie gas, Acta Astronaut. 6, 413-434
(1979)
[8 J. A. Steketee, Transformations of the equations of motion for the unsteady rectilinear flow of a
perfect gas, J. Engrg. Math. 10, 69-94 (1976)
[9 P. Smith, Anisentropic rectilinear gas flows, Appl. Sci. Res. A12, 66-72 (1964)
[10 H. Ardavan-Rhad, The decay of a plane shock wave, J. Fluid Mech. 43, 737-751 (1970)
[11 M. D. Ustinov, Ideal gas flow behind a finite-amplitude shock wave, Izv. Akad. Nauk. SSSR, Mekh.
Zhid. Gaza. 2, 88-90 (1967)
[12 P. L. Sachdev and A. Venkataswamy Reddy, Some exact solutions describing unsteady plane gas
flows with shocks, Quart. Appl. Math. 40, 249-272 (1982)
[13 P. L. Sachdev, Neelam Gupta, and D. S. Ahluwalia, Exact analytic solutions describing unsteady
plane gas flows with shocks of arbitrary strength, Quart. Appl. Math. 50, 677-726 (1992)
[14 M. D. Ustinov, Approximate solution to nonself-similar problem of motion of a piston after an
impact, Izv. Akad. Nauk. SSSR, Mekh. Zhid. Gaza. 6, 167-171 (1982)
[15 M. D. Ustinov, Motion of a piston under the influence of gas pressure in the presence of an initial
temperature gradient, Izv. Akad. Nauk. SSSR, Mekh. Zhid. Gaza. 2, 177-180 (1984)
[16 M. D. Ustinov, Some one-dimensional unsteady adiabatic gas flows with plane symmetry, Izv.
Akad. Nauk. SSSR, Mekh. Zhid. Gaza. 5, 96-104 (1986)
[17 P. Clarkson and M. D. Kruskal, New similarity reductions of the Boussinesq equation, J. Math.
Phys. 30, 2201-2213 (1989)
[18 G. W. Bluman and J. D. Cole, The general similarity solution of the heat equation 18, 1025-1042
(1969)

View publication stats