Large-Time Asymptotics for Solutions of a Generalized

Burgers Equation with Variable Viscosity

By Ch. Srinivasa Rao and Engu Satyanarayana

In this paper, we discuss the large-time asymptotics for periodic solutions of

a generalized Burgers equation with variable viscosity (GBEV). Large-time
asymptotics for the solutions of the GBEV depend on the parameters present
in the partial differential equation and also the period of the solution of the
GBEV. Large-time asymptotic expansions of the solutions are obtained by
improving the solution of the linearized GBEV for certain parametric regions
via a perturbative approach. These constructed large-time asymptotics are
compared with the corresponding numerical solutions and are found to be
in good agreement for large time. For certain other parametric region, our
numerical study suggests that the solution of the inviscid GBEV describes the
large-time behavior of the periodic solutions of the GBEV.

1. Introduction

In this paper, we discuss the large-time asymptotics for solutions of a

generalized Burgers equation with variable viscosity (GBEV), namely,
δ
u t + uu x = uxx , 0 < x < l, t > 0, (1)
(t + 1) M
subject to the initial and boundary conditions
u(x, 0) = u 0 (x), 0 ≤ x ≤ l, (2)

Address for correspondence: Dr. Ch. Srinivasa Rao, Department of Mathematics, Indian Institute of
Technology Madras, Chennai 600036, India; e-mail: chsrao@iitm.ac.in

DOI: 10.1111/j.1467-9590.2011.00509.x 1
STUDIES IN APPLIED MATHEMATICS 0:1–23

C 2011 by the Massachusetts Institute of Technology
2 Ch. S. Rao and E. Satyanarayana

u(0, t) = u(l, t) = 0, t ≥ 0. (3)

Here M ≥ 0, δ > 0, and l > 0 are constants. Further, u 0 (x) is continuous and
takes the value zero for x = 0 and x = l.
2
For the parametric regions (i) 0 ≤ M < 1 and (ii) M = 1, δ > πl 2 , the
large-time asymptotics for solutions of (1)–(3) are constructed via a perturbative
approach assuming that a solution of the linearized equation
δ
ut = uxx , 0 < x < l, t >0 (4)
(t + 1) M
of (1) satisfying (3) describes the large-time behavior of the solutions of (1)–(3).
The large-time asymptotic solutions of (1)–(3) constructed here are compared
with the numerical solutions of (1)–(3) obtained by a numerical scheme due to
Dawson [1]. They agree very well for large time. The perturbative approach
used for constructing large-time asymptotic solutions is similar to the work of
Sachdev et al. [2, 3]. It may be noted that we do not use the initial profile
in the construction of the large-time asymptotic solutions of (1)–(3). Our
numerical study shows that the large-time asymptotics of (1)–(3) constructed
here describe the large-time behavior of the solutions of (1)–(3) for different
initial conditions. However there is a constant, called the old-age constant, in
the constructed asymptotic solutions of (1). This constant needs to be found by
matching with the relevant numerical solution of (1)–(3) or by some other
means. It may depend on the parameters present in the partial differential
equation and the initial and boundary conditions. Our numerical study suggests
that the large-time behavior for solutions of (1)–(3) for the parametric region
M > 1 is described by the solution x/τ of the inviscid generalized Burgers
equation, namely,
u τ + uu x = 0, τ = t + 1. (5)
2
The parametric region M = 1, δ ≤ πl 2 will be dealt elsewhere.
Equation (1) with M = 0 becomes the most celebrated Burgers equation
u t + uu x = δu x x , 0 < x < l, t > 0. (6)
The exact periodic solution of (6) subject to (3) is given by
⎡ ∞  2 2   nπ x ⎤
 n π δt
⎢ nexp − 2
sin ⎥
4π δ ⎢
⎢ n=1
l l ⎥
⎥
u(x, t) = ⎢   ⎥ (7)
l ⎣ ∞
n π δt
2 2 
nπ x ⎦
1+2 exp − cos
n=1
l2 l

(see Sachdev [10], p. 33). It may be noted that the exact solution (7) of (6) is
obtained from the solution of (6) subject to the boundary conditions (3) and
the initial datum
 Solutions of a Generalized Burgers Equation 3

πx
u(x, 0) = u 0 sin , 0 ≤ x ≤ l, (8)
l
under the restriction that δ is sufficiently small. Here u 0 > 0 is a constant.
Expanding the solution (7) of the Burgers equation (6) as a series in descending
exponential functions for sufficiently large time, we obtain large-time asymptotic
expansion for periodic solutions of (6) subject to (3) and (8) as
πx  
4π δ −π 2 δt/l 2 −2π 2 δt/l 2 2π x
u(x, t) = e sin −e sin
l l l
    
−3π 2 δt/l 2 πx 3π x
+e sin + sin + · · · as t → ∞. (9)
l l
We show that the large-time asymptotic solution for (1)–(3) constructed here (for
the parametric regions 0 ≤ M < 1) contains (9) as a special case when M = 0.
A special case of (1) with M = 1/2, namely,
δ
u t + uu x = √ uxx
t +1
was derived by Enflo and Rudenko [4] from Khokhlov, Zabolotskaya, and
Kuznetsov (KZK) equation while studying a plane wave in the center of a
bounded nonlinear acoustic beam (see also Enflo and Hedberg [5], p. 215).
Crighton [6] derived generalized Burgers equations (GBEs) of the form

u t + uu x = ν(t)u x x ,

where (i) ν(t) = t + t0 , (ii) ν(t) = exp(t/t0 ), (iii) ν(t) = (t0 ± t)−1 . These
GBEs are referred to as cylindrical far-field GBE, spherical far-field GBE, and
exponential horn GBE, respectively. Note that the GBE (iii) corresponds to (1)
with M = 1 (see also Sionóid and Cates [7]). Cates [8] transformed the GBE

u t + uu x = ε(t)u x x (10)

into

wτ + wwξ = ε(τ/(1 − τ ))wξ ξ (11)

via the transformation

u(x, t) = (1 + t)−1 w(x(1 + t)−1 , 1 − (1 + t)−1 ) + x(1 + t)−1 , (12)

x t
ξ= , τ= . (13)
1+t 1+t
An interesting observation of Cates [8] is that the cylindrical far-field GBE
(ν(t) = t + 1) transforms to the exponential horn GBE
4 Ch. S. Rao and E. Satyanarayana

1
wτ + wwξ = wξ ξ
1−τ
via the transformation (12) and (13). Thus, these GBEs are equivalent.
For a related study, we may refer to [9–15].
Assume that u 1 (x) is a continuous function on R satisfying the following
conditions:
(i) u 1 (0) = u 1 (l) = 0,
(ii) u 1 (x) is periodic with period 2l,
(iii) u 1 (x) is anti-symmetric in the interval (− l, l).
Because the GBE (1) and the function u 1 (x) are invariant under the
transformations
(i) x → −x, u → − u,
(ii) x → x + 2l, u → u,
the solution of the GBE (1) subject to the initial profile u 1 (x) is given by the
solution of (1) subject to (3) with the initial profile u 1 (x) restricted to the
interval [0, l]. Because of this reason, without the loss of generality, we may
refer to the solutions of (1)–(3) as periodic solutions.
The organization of this paper is as follows. We construct large-time
asymptotic periodic solutions of (1)–(3) for the parametric regions
0 ≤ M < 1 and M = 1, δ > l 2 /π 2 in Sections 2 and 3, respectively. The
conclusions of the present paper are set forth in Section 4.

2. Large-time asymptotics for periodic solutions of (1) when 0 ≤ M < 1

In this section, we construct large-time asymptotics for periodic solutions of

(1)–(3) via a perturbative approach. The asymptotics of (1)–(3) constructed
here are compared with numerical solutions of (1)–(3) obtained by Dawson’s
[1] numerical scheme for a specific initial condition.
We assume that, for large time, the diffusion term dominates the convection
term of (1) and hence the periodic solution of (1) subject to (2)–(3) is
asymptotic in the limit t → ∞ to the old-age solution of (1), namely,
πx
u(x, t) = Ae−τ (t) sin (14)
l
when 0 ≤ M < 1. Here
π 2δ
τ (t) = (t + 1)1−M (15)
l 2 (1 − M)
and A is old-age constant.
 Solutions of a Generalized Burgers Equation 5

Note that the function given in (14) is a solution of the linear partial
differential equation (4) satisfying the boundary conditions (3).
We seek u(x, t) in the form
πx
u(x, t) = Ae−τ (t) sin + (x, t) as t → ∞. (16)
l
Here (x, t)  O(e−τ (t) ) as t → ∞. The correction term (x, t) takes into
account the effect of the nonlinear term in (1) for large time. Substitution of
(16) into (1) gives the following partial differential equation for (x, t):
δ πx Aπ −τ (t) πx
−τ (t)
t −  xx +  x + Ae sin  x + e cos 
(t + 1) M l l l
 
A2 π −2τ (t) 2π x
=− e sin . (17)
2l l
Ignoring the higher order terms, for sufficiently large t, we get
 
δ A2 π −2τ (t) 2π x
t − x x ≈ − e sin . (18)
(t + 1) M 2l l
Substituting the particular form
 
2π x
(x, t) ≈ f (t) sin (19)
l
in (18) yields
A2 π −2τ (t)
f
(t) + 4τ
f (t) ≈ − e . (20)
2l
The general solution of (20) is given by

A2 π −4τ (t)
f (t) ≈ − e e2τ (t) dt + ce−4τ (t) , c is the integration constant
2l

A2 π −4τ (t) (21)
∼− e e2τ (t) dt as t → ∞.
2l
Applying integration by parts for the integral in (21), we arrive at

A2l −2τ (t) M M(2M − 1)
f (t) ∼ − e (t + 1) M − (t + 1)2M−1 + (t + 1)3M−2
4π δ a a2

M(2M − 1)(3M − 2) (22)
− (t + 1)4M−3
+ · · · (up to [m] terms)
a3
as t → ∞; here a = 2π 2 δ/l 2 , m = 1/(1 − M) and [m] is the greatest integer
less than or equal to m. It may be observed that (i) if m is a positive integer,
then (m + 1)th term and the subsequent terms in the bracket of (22) become
zero (i.e., for M = 0, 1/2, 2/3, . . . , the right-hand side (RHS) expression of
6 Ch. S. Rao and E. Satyanarayana

Equation (22) contains only finite number of terms) and (ii) if m is not a
positive integer (i.e., M = 0, 1/2, 2/3, . . .), then those terms with negative
powers may be neglected as t → ∞ . Therefore, the function  for (18) takes
the general form
∞  nπ x  
−n 2 τ (t) 2π x
(x, t) ∼ cn e sin + f (t) sin as t → ∞. (23)
n=2
l l

Here f (t) is as in (22) and in the summation of (23), n has to vary from 2 to ∞
because of the requirement that
 
(x, t)  O e−τ (t) as t → ∞.
It follows from (23) that
 
2π x
(x, t) ∼ f (t) sin as t → ∞. (24)
l
Thus, the large-time asymptotic periodic solution of (1)–(3) for 0 ≤ M < 1 is
given by
πx  
−τ (t) 2π x
u(x, t) = Ae sin + f (t) sin + ··· (25)
l l
π δ 2
as t → ∞ ; here τ (t) = l 2 (1−M) (t + 1)1−M and f (t) is as in (22).
Motivated by the form of the asymptotic solution (25) for (1)–(3), we seek
the large-time asymptotic periodic solution u of (1)–(3) as follows:
u(x, t) = e−τ (t) f 1 (x, t) + e−2τ (t) f 2 (x, t) + e−3τ (t) f 3 (x, t)
+ · · · + e−nτ (t) f n (x, t) + · · · as t → ∞. (26)
Substituting the expression (26) for u into (1) and then equating the
coefficients of e−nτ (t) , n = 1, 2, 3, . . . to zero yield the following system of
linear second order partial differential equations for the unknown functions f n :
l2

f 1,t − τ
(t) f 1 − τ (t) f 1,x x = 0, (27)
π2
l2

f 2,t − 2τ
(t) f 2 − τ (t) f 2,x x = −[ f 1 f 1,x ], (28)
π2
l2

f 3,t − 3τ
(t) f 3 − τ (t) f 3,x x = −[ f 1 f 2,x + f 2 f 1,x ], (29)
π2
...
l2

f n,t − nτ
(t) f n − τ (t) f n,x x = −[ f 1 f n−1,x + f 2 f n−2,x + · · · + f n−1 f 1,x ],
π2
(30)
....
 Solutions of a Generalized Burgers Equation 7

The solution f 1 of (27) satisfying the relevant boundary conditions

f 1 (0, t) = f 1 (l, t) = 0 (31)
is given by
∞
  nπ x
An e−(n −1)τ (t)
2
f 1 (x, t) = sin
n=1
l
πx
∼ A1 sin as t → ∞. (32)
l
In view of (32) and (28), we have
 
l2
A21 π 2π x
f 2,t − 2τ
(t) f 2 − τ (t) f 2,x x = − sin (33)
π2 2l l
as t → ∞. Motivated by the RHS expression of (33), let us choose the
particular solution
 
2π x
f 2 p (x, t) = χ (t) sin . (34)
l
Then χ (t) satisfies
A21 π
χ
(t) + 2τ
(t)χ (t) + = 0. (35)
2l
Solving (35), we arrive at

A21l M M(2M − 1)
χ (t) ∼ − (t + 1) M − (t + 1)2M−1 + (t + 1)3M−2
4π δ a a2

+ · · · (upto [1/(1 − M)] terms) as t → ∞ (36)

≡ C1 (t + 1) M + C2 (t + 1)2M−1 + C3 (t + 1)3M−2 + · · · as t → ∞,(37)

where a = 2π 2 δ/l 2 . The solution f 2 of (33) satisfying the relevant boundary
conditions
f 2 (0, t) = f 2 (l, t) = 0 (38)
is given by
∞
  nπ x  
−(n 2 −2)τ (t) 2π x
f 2 (x, t) = αn e sin + χ (t) sin , (39)
n=2
l l
where α n are constants. In the summation of (39), n cannot start from 1 because
of the requirement that
f 2 (x, t)e−2τ (t)  O(e−τ (t) ) as t → ∞.
8 Ch. S. Rao and E. Satyanarayana

It is easy to see that

 
2π x
f 2 (x, t) ∼ χ (t) sin as t → ∞, (40)
l
where χ is given by (36). Making use of (32) and (40), Equation (29) becomes
   π x 

l2
A1 π 3π x
f 3,t − 3τ (t) f 3 − 2 τ (t) f 3,x x = − χ (t) 3 sin − sin
π 2l l l
(41)
as t → ∞ . Motivated by the RHS expression of (41), we attempt the particular
solution of (41) in the form
  πx
3π x
f 3 p (x, t) = φ(t) sin + ψ(t) sin . (42)
l l
Substituting (42) into (41) and comparing the coefficients of sin( 3πl x ) and
sin( πlx ), respectively, we get the following inhomogeneous first order ordinary
differential equations for φ and ψ, respectively:
3A1 π
φ
+ 6τ
φ +
χ (t) = 0, (43)
2l
A1 π
ψ
− 2τ
ψ − χ (t) = 0. (44)
2l
Solving (43) and (44), we arrive at

3A1 π −6τ (t)
φ(t) = − e χ (t)e6τ (t) dt + c1 e−6τ (t) , (45)
2l

A1 π 2τ (t)
ψ(t) = e χ (t)e−2τ (t) dt + c2 e2τ (t) ; (46)
2l
here c1 and c2 are integration constants. In view of (36), we obtain

A2l 3
e−6τ (t) χ (t) e6τ (t) dt ∼ − 1 3 2 (t + 1)2M + O((t + 1)3M−1 ) as t → ∞,
24π δ
(47)

A21l 3
e 2τ (t)
χ (t) e−2τ (t) dt ∼ (t + 1)2M + O((t + 1)3M−1 ) as t → ∞.
8π 3 δ 2
(48)
Inspired by Equations (47) and (48), we write the particular solutions for φ
and ψ as follows:
φ p (t) = d1 (t + 1)2M + d2 (t + 1)3M−1 + d3 (t + 1)4M−2 + · · · as t → ∞,
(49)
 Solutions of a Generalized Burgers Equation 9

ψ p (t) = d˜1 (t + 1)2M + d˜2 (t + 1)3M−1 + d˜3 (t + 1)4M−2 + · · · as t → ∞.

(50)

Substituting (49)–(50) in (43) and (44), respectively, and solving for di , d˜i ,
we arrive at
C˜1 C̃i − di−1 [i M − i + 2]
d1 = , di = , i ≥ 2, (51)
ã ã
C¯1 C̄i − d˜i−1 [i M − i + 2]
d˜1 = , d˜i = , i ≥ 2, (52)
ā ā
where
6π 2 δ 3A1 π
ã = , C̃i = − Ci , i ≥ 1, (53)
l2 2l
2π 2 δ A1 π
ā = −
2
, C̄i = Ci , i ≥ 1, (54)
l 2l
and Ci are as in (37). In view of Equations (45) and (49),
φ(t) ∼ d1 (t + 1)2M + d2 (t + 1)3M−1 + d3 (t + 1)4M−2 + · · · as t → ∞.
(55)
Because
f 3 (x, t) e−3τ (t)  O(e−2τ (t) (t + 1) M ) as t → ∞, (56)
c2 in (46) must be zero and hence

A1 π 2τ (t)
ψ(t) = e χ (t)e−2τ (t) dt
2l
∼ d˜1 (t + 1)2M + d˜2 (t + 1)3M−1 + d˜3 (t + 1)4M−2 + · · · as t → ∞.
(57)
Then the solution f 3 for (41) subject to the relevant boundary conditions
f 3 (0, t) = f 3 (l, t) = 0 (58)
is given by
∞
  nπ x   πx
3π x
kn e−(n −3)τ (t)
2
f 3 (x, t) = sin + φ(t) sin + ψ(t) sin ,
n=2
l l l
(59)
where kn are constants, φ and ψ are as in (55) and (57), respectively. Thus,
  πx
3π x
f 3 (x, t) ∼ φ(t) sin + ψ(t) sin as t → ∞. (60)
l l
10 Ch. S. Rao and E. Satyanarayana

In the similar way, we can solve the linear partial differential Equation (30) for
f n , where n = 4, 5, . . . .
Finally, the large-time asymptotic periodic solution of (1)–(3) is given by
πx  
−τ (t) −2τ (t) 2π x
u(x, t) = A1 e sin +e χ (t) sin
l l
   
−3τ (t) 3π x πx
+e φ(t) sin + ψ(t) sin + ··· (61)
l l
as t → ∞; here A1 is the old-age constant and the functions χ , φ and ψ are
given by (36), (55), and (57), respectively, and τ (t) is as in (15).
It is to be noted that we have not used the initial condition (2) anywhere
in the process of constructing the large-time asymptotic periodic solution of
(1)–(3). It can be observed that the solution (61) of (1)–(3) has exponential
decay as t → ∞.
It can be seen that the asymptotic solution (61) for M = 0 reduces to
 π x (A e−γ )2  
−γ −γ t 1 −2γ t 2π x
u(x, t) = (A1 e )e sin − e sin
l 4π δ/l l
−γ 3     
(A1 e ) −3γ t 3π x πx
+ e sin + sin + ··· (62)
(4π δ/l)2 l l

as t → ∞; here γ = π 2 δ/l 2 . If we choose the old-age constant A1 to be 4πδ

l
eγ,
the large-time asymptotic solution (62) becomes the asymptotic solution (9) of
the Burgers equation (6).
For the sake of numerical study, we take M = 0.5. In this case, the
asymptotic periodic solution u of (1) satisfying (3) is given by (61) with

2π 2 δ √
τ (t) = t + 1, (63)
l2
√
A21l(l 2 − 4π 2 δ t + 1)
χ (t) = , (64)
16π 3 δ 2
 √ 
5l 6
5l 4
t + 1 l 2
3A31 − + − (t + 1)
432π 4 δ 2 36π 2 δ 3
φ(t) = − , (65)
16π 2 δ 2

 √ 
l6 l4 t + 1
A31 + + l (t + 1)
2
16π 4 δ 2 4π 2 δ
ψ(t) = . (66)
16π 2 δ 2
 Solutions of a Generalized Burgers Equation 11

0.2

1 0.18

0.16
0.8
0.14

0.12
0.6
u(x,t)

u(x,t)
0.1

0.08
0.4
0.06

0.2 0.04

0.02

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 1. [Left] Initial profile given in (67) at t = 0; [right] numerical solution (dashed) of
(1) subject to (67) and (68), asymptotic solution (61) (solid), and the linear solution (69)
(dashdotted) with M = 0.5, δ = 0.5 at t = 5.

We now solve, numerically, the variable coefficient Burgers equation (1) with
M = 0.5 and δ = 0.5 subject to the initial–boundary conditions

u(x, 0) = sin x, x ∈ [0, π ], (67)

u(0, t) = u(π, t) = 0, t ≥ 0. (68)

We chose t = 0.0001, x = 0.0062 and used Dawson’s [1] scheme for

finding the numerical solution of the GBE (1) with the initial and boundary
conditions (67) and (68). We chose l = π for the sake of simplicity. As the
initial profile evolves in time under the generalized Burgers equation (1),
the solution of (1) for sufficiently large time behaves like the solution √(14)
of the linearized equation (4) for 0 ≤ M < 1. We computed sin(x u max
max )
e2δ t+1 ,
where xmax is the value of x when u attains its maximum value u max on (0,
π ) at different times and chose the converged value for A1 . Convergence
of A1 means that the nonlinear terms are negligible, that is, we reached
the linear regime. Here A1 is 2.06. The agreement between numerical and
asymptotic solutions is very good. Figures 1–3 show the asymptotic solution
(61) with τ (t), χ (t), φ(t), and ψ(t) given by (63)–(66), numerical solution of
the variable coefficient Burgers equation (1) subject to (67) and (68) with
M = 0.5 and δ = 0.5 and the solution
√
u lin (x, t) = A1 e−2δ t+1
sin x (69)

of linearized form (4) at various times. We observe that our asymptotic solution
(61) starts agreeing with the relevant numerical solution much before the linear
regime.
12 Ch. S. Rao and E. Satyanarayana

0.08 0.04

0.07 0.035

0.06 0.03

0.05 0.025
u(x,t)

u(x,t)
0.04 0.02

0.03 0.015

0.02 0.01

0.01 0.005

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 2. Numerical solution (dashed) of (1) subject to (67) and (68), asymptotic solution (61)
(solid), and the linear solution (69) (dashdotted) with M = 0.5, δ = 0.5 at t = 10, 15,
respectively.

x 10
0.025 1.8

1.6

0.02
1.4

1.2
0.015
1
u(x,t)

u(x,t)

0.8
0.01
0.6

0.4
0.005

0.2

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 3. Numerical solution (dashed) of (1) subject to (67) and (68), asymptotic solution (61)
(solid), and the linear solution (69) (dashdotted) with M = 0.5, δ = 0.5 at t = 20, 50,
respectively.

3. Large-time asymptotics for periodic solutions of (1)

when M = 1 and δ > l 2 /π 2

In this section, we construct large-time asymptotics for periodic solutions

of (1)–(3) for M = 1 and δ > l 2 /π 2 via the perturbative approach as in
Section 2. The large-time asymptotics of (1) constructed here are compared
with numerical solutions of (1)–(3) for specific initial conditions.
The initial–boundary value problem (1)–(3) with M = 1 becomes
δ
u t + uu x = uxx , 0 < x < l, t > 0, (70)
t +1

u(x, 0) = u 0 (x), 0 ≤ x ≤ l, (71)

 Solutions of a Generalized Burgers Equation 13

u(0, t) = u(l, t) = 0, t ≥ 0. (72)

We assume that, for large time, the convection term gets dominated by the
diffusion term in (70) and then the periodic solution of (70) subject to (71) and
(72) is asymptotic to function
πx
u(x, t) = A0 (t + 1)−δπ /l sin
2 2
(73)
l
for δ > l 2 /π 2 . Here A0 is old-age constant. Note that the function given in
(73) is a solution of the linearized partial differential equation
δ
ut = uxx , 0<x <l (74)
t +1
of (70) subject to the initial and boundary conditions (71) and (72). We seek
the asymptotic periodic solution u of (70) in the form
πx
u(x, t) = A0 (t + 1)−k sin + (x, t) as t → ∞, (75)
l
where k = δπ 2 /l 2 and (x, t)  O((t + 1)−k ) as t → ∞. Substitution of (75)
into (70) gives the following partial differential equation for (x, t):
δ πx A0 π πx
t − x x + x + A0 (t + 1)−k sin x + (t + 1)−k cos 
t +1 l l l
 
A2 π 2π x
= − 0 (t + 1)−2k sin . (76)
2l l
Ignoring the higher order terms in (76), we get
 
δ A20 π −2k 2π x
t − x x ≈ − (t + 1) sin (77)
t +1 2l l
as t → ∞. Let
 
2π x
(x, t) = f (t) sin as t → ∞. (78)
l
Substitution of the expression for  given in (78) into (77) yields
f (t) A2 π
f
(t) + 4k = − 0 (t + 1)−2k . (79)
t +1 2l
Solving (79) for f , we arrive at
A20 π
f (t) = c̄1 (t + 1)−4k − (t + 1)1−2k , (80)
2l(1 + 2k)
where c̄1 is the integration constant. Then the correction term  satisfying the
relevant boundary conditions
14 Ch. S. Rao and E. Satyanarayana

(0, t) = (l, t) = 0 (81)

is given by
∞
  nπ x
c̃n (t + 1)−n k sin
2
(x, t) =
n=2
l
  
A20 π 2π x
+ c̄1 (t + 1)−4k − (t + 1)1−2k
sin , (82)
2l(1 + 2k) l
where c̃n are constants and the summation in (82) started from n = 2 because
of our assumption that
(x, t)  O((t + 1)−k ) as t → ∞. (83)
It follows from (82) that
 
A20 π 2π x
(x, t) ∼ − (t + 1)1−2k
sin as t → ∞. (84)
2l(1 + 2k) l
The condition on  (see (83)) is satisfied because of our assumption that k >
1. Therefore, the asymptotic periodic solution u of (70)–(72) is
πx  
−k −(2k−1) 2π x
u(x, t) = A0 (t + 1) sin − B̃(t + 1) sin + · · · (85)
l l
A2 π
as t → ∞, where B̃ = 2l(1+2k)
0
and k = δπ 2 /l 2 .
Motivated by the form of asymptotic solution (85) for (70) satisfying (72),
we seek the large-time asymptotic periodic solution u of (70) subject to (71)
and (72) as follows:
u(x, t) = (t + 1)−k f 1 (x, t) + (t + 1)−2k f 2 (x, t)
+ (t + 1)−3k f 3 (x, t) + · · · + (t + 1)−nk f n (x, t) + · · · (86)
as t → ∞ and k = δπ 2 /l 2 . Substituting the expression (86) for u into (70)
and equating the coefficients of (t + 1)−nk , n = 1, 2, 3, . . . to zero yield the
following system of linear second order partial differential equations for f n :
(t + 1) f 1,t − k f 1 − δ f 1,x x = 0, (87)

(t + 1) f 2,t − 2k f 2 − δ f 2,x x = −(t + 1)[ f 1 f 1,x ], (88)

(t + 1) f 3,t − 3k f 3 − δ f 3,x x = −(t + 1)[ f 1 f 2,x + f 2 f 1,x ], (89)

···
 Solutions of a Generalized Burgers Equation 15

(t + 1) f n,t − nk f n − δ f n,x x = −(t + 1)[ f 1 f n−1,x + f 2 f n−2,x + · · · + f n−1 f 1,x ],

(90)
···
The general solution f 1 of (87) subject to the relevant boundary conditions
f 1 (0, t) = f 1 (l, t) = 0 (91)
is given by
∞
  nπ x
An (t + 1)−(n −1)k
2
f 1 (x, t) = sin
n=1
l
πx
∼ A1 sin as t → ∞. (92)
l
In view of (92), Equation (88) reduces to
 
A21 π 2π x
(t + 1) f 2,t − 2k f 2 − δ f 2,x x =− (t + 1) sin (93)
2l l
as t → ∞. Motivated by RHS expression of (93), we seek the particular
solution f 2p for (93) as
f 2 p (x, t) = (t + 1)g(x). (94)
Because of (72), the boundary conditions for f 2p are
f 2 p (0, t) = f 2 p (l, t) = 0. (95)
Then (94) gives
g(0) = g(l) = 0. (96)
Substitution of the expression for f 2p given in (94) into (93) yields
   

2k − 1 A21 π 2π x
g + g= sin . (97)
δ 2lδ l
The general solution of (97) is given by
     
2k − 1 2k − 1 A21 π 2π x
g(x) = c3 cos x + c4 sin x − sin .
δ δ 2l (1 + 2k) l
(98)
Here c3 and c4 are constants. The boundary condition (96)1 gives c3 = 0. This
result along with the condition (96)2 and (98) gives
1
either c4 = 0 or = 2 − n2, n is any integer. (99)
k
Recall that k = δπl 2 > 1. This, in turn, implies that 0 < k1 < 1 and hence there
2

exists no integer satisfying (99)2 . Hence, we must choose c4 = 0 and the

16 Ch. S. Rao and E. Satyanarayana

general form of the solution f 2 of (93) is

∞
  nπ x
Bn (t + 1)−(n −2)k
2
f 2 (x, t) = sin
n=2
l
 
A21 π 2π x (100)
− (t + 1) sin ,
2l(1 + 2k) l
where Bn are constants. In the summation of (100), n varies from 2 to ∞
because of the requirement that
(t + 1)−2k f 2 (x, t)  O((t + 1)−k ) as t → ∞.
Retaining the dominant term in (100), we get
 
2π x
f 2 (x, t) ∼ B(t + 1) sin as t → ∞, (101)
l
A2 π
where B = − 2l(1+2k)
1
. Note that (t + 1)−2k f 2 (x, t)  O((t + 1)−k ) as t → ∞,
because of the condition k > 1. Making use of (92) and (101), Equation (89)
becomes
   π x 
3π x
(t + 1) f 3,t − 3k f 3 − δ f 3,x x = − D̃(t + 1) 3 sin
2
− sin (102)
l l
as t → ∞. Here D̃ = A12lBπ . The RHS expression of (102) suggests the
particular solution f 3p to be of the form

f 3 p (x, t) = (t + 1)2 h(x). (103)

Substitution of the expression (103) into (102) leads to the second order linear
ordinary differential equation
     π x 
3k − 2 D̃ 3π x
h

+ h= 3 sin − sin . (104)

δ δ l l
The general solution of (104) is given by
   
3k − 2 3k − 2
h(x) = c5 cos x + c6 sin x
δ δ
πx  
D̃ 3 D̃ 3π x (105)
+ sin − sin ,
2(1 − k) l 2(1 + 3k) l
where c5 and c6 are constants. Because of (72), the relevant boundary conditions
for f 3p are
f 3 p (0, t) = f 3 p (l, t) = 0. (106)
 Solutions of a Generalized Burgers Equation 17

Then (103) gives

h(0) = h(l) = 0. (107)

The boundary conditions (107) yield c5 = 0 and

2
either c6 = 0 or = 3 − n2, n is any integer. (108)
k

As 0 < k2 < 2, there exists no integer satisfying (108)2 . Thus, c6 = 0. Hence,

the general solution for f 3 is
∞
  nπ x
Cn (t + 1)−(n −3)k
2
f 3 (x, t) = sin
n=2
l
πx  
D̃ 1 3 3π x
+ (t + 1)2 sin − sin , (109)
2 1−k l 1 + 3k l

where Cn are constants. In the summation of (109), n varies from 2 to ∞

because of the assumption that

(t + 1)−3k f 3 (x, t)  O((t + 1)−(2k−1) ) as t → ∞.

It follows from (109) that

πx  
D̃ 1 3 3π x
f 3 (x, t) ∼ (t + 1)2
sin − sin (110)
2 1−k l 1 + 3k l

as t → ∞. In the similar way, we can solve the linear partial differential

Equation (90) for f n , where n = 4, 5, . . .
Thus, the large-time asymptotics for periodic solution u of (70)–(72) is
πx  
−k −(2k−1) 2π x
u(x, t) = A1 (t + 1) sin + B(t + 1) sin
l l
πx  
D̃ −(3k−2) 1 3 3π x
+ (t + 1) sin − sin
2 1−k l 1 + 3k l
+ ··· as t → ∞, (111)

A2 π
where k = δπ 2 /l 2 , B = − 2l(1+2k)
1
, D̃ = A12lBπ and A1 is the old-age constant.
It may be noted that the initial condition (71) is not used while constructing
large-time asymptotics for periodic solutions of (70). It may be observed that
the large-time asymptotic solution u given in (111) for (70)–(72) has algebraic
decay for sufficiently large t.
18 Ch. S. Rao and E. Satyanarayana

We have solved numerically the variable coefficient Burgers equation (70)

with δ = 1.2 subject to the initial and boundary conditions:

⎧
⎪ 2x π
⎨ , if 0 ≤ x ≤ ,
u(x, 0) = π
2 (112)
⎪
⎩2 1 − x , π
if ≤ x ≤ π,
π 2

⎧
⎪ 4x π
⎪
⎪ , if 0 ≤ x ≤ ,
⎪
⎪ π 4
⎪
⎪
⎪
⎪
⎪ 4 x π π
⎨ 1− , if ≤x≤ ,
u(x, 0) = 3 π 4 2 (113)
⎪
⎪ 4x π 3π
⎪
⎪ , if ≤x≤ ,
⎪
⎪ 3π 2 4
⎪
⎪ 
⎪
⎪
⎩4 1 − x , if
3π
≤ x ≤ π,
π 4

u(x, 0) = sin x, x ∈ [0, π ], (114)

u(0, t) = u(π, t) = 0, t ≥ 0. (115)

As in Section 2, we used Dawson’s [1] scheme for finding the numerical

solutions of the GBE (70) satisfying the boundary conditions (115) and the
initial conditions (112), (113), and (114), respectively.
We computed sin(x u max
max )
(t + 1)δ , where xmax is the value of x when u attains its
maximum value u max on (0, π ) at different times and chose the converged
value for A1 . The values of the old-age constant A1 corresponding to the
initial profiles (112), (113), and (114) are 0.784, 0.944, and 0.953, respectively.
The agreement between numerical and asymptotic solutions is quite good for
all the three initial profiles when t is large. For the sake of illustration, we
present below in Figures 4–7 the comparison of the asymptotic solution (111),
numerical solution of (70) subject to (113) and (115), and the solution of the
linearized equation of (70) given by

u(x, t) = A1 (t + 1)−δ sin x (116)

when δ = 1.2 at different times.

 Solutions of a Generalized Burgers Equation 19

0.9

1
0.8

0.7
0.8
0.6

0.6 0.5

u(x,t)
0.4

0.4
0.3

0.2
0.2
0.1

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x

Figure 4. [Left] Initial profile given by (113) at t = 0; [right] numerical solution (dashed) of
(70) subject to (113) and (115), asymptotic solution (111) (solid), and the linear solution
(116) (dashdotted) with δ = 1.2 at time t = 0.1.
0.8 0.7

0.7 0.6

0.6
0.5

0.5
0.4
u(x,t)

u(x,t)

0.4
0.3
0.3

0.2
0.2

0.1 0.1

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 5. Numerical solution (dashed) of (70) subject to (113) and (115), asymptotic solution
(111) (solid), and the linear solution (116) (dashdotted) with δ = 1.2 at times t = 0.2, 0.5,
respectively.

0.45 0.06

0.4
0.05
0.35

0.3 0.04

0.25
u(x,t)

u(x,t)

0.03
0.2

0.15 0.02

0.1
0.01
0.05

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 6. Numerical solution (dashed) of (70) subject to (113) and (115), asymptotic solution
(111) (solid), and the linear solution (116) (dashdotted) with δ = 1.2 at times t = 1, 10,
respectively.
20 Ch. S. Rao and E. Satyanarayana

x 10 x 10
4 1

0.9
3.5
0.8
3
0.7
2.5
0.6
u(x,t)

u(x,t)
2 0.5

0.4
1.5
0.3
1
0.2
0.5
0.1

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 7. Numerical solution (dashed) of (70) subject to (113) and (115), asymptotic solution
(111) (solid), and the linear solution (116) (dashdotted) with δ = 1.2 at times t = 100, 300,
respectively.
0.7

1
0.6

0.8 0.5

0.4
0.6
u(x,t)

u(x,t)

0.3
0.4

0.2

0.2
0.1

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 8. [Left] Initial profile given by (114) at t = 0; [right] numerical solution (dashed) of
(1) subject to (114), (115) with M = 1.5, δ = 0.5 at time t = 5. Solid profile corresponds to
the inviscid solution (117) of (1) at t = 5.

Remark. Based on our numerical study, we expect that the large-time

asymptotic behavior of the solutions of (1)–(3) on [0, l] for M > 1 is given by
the solution
u(x, t) = x/(t + 1) as t → ∞ (117)
of the inviscid equation, namely,
u τ + uu x = 0, (118)
where τ = t + 1. Figures 8 and 9 show the evolution of the initial profile
(114) under the GBE (1) for M = 1.5, δ = 0.5 to the inviscid solution (117).

Remark. Parker [16] studied an initial value problem posed for a GBE
on infinite interval. He made use of Cole–Hopf like transformation and then
 Solutions of a Generalized Burgers Equation 21

0.35 0.07

0.3 0.06

0.25 0.05

0.2 0.04
u(x,t)

u(x,t)
0.15 0.03

0.1 0.02

0.05 0.01

0 0
0 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5
x x

Figure 9. Numerical solution (dashed) of (1) subject to (114), (115) with M = 1.5, δ = 0.5
at times t = 10, 50, respectively. Solid profiles correspond to the inviscid solution (117) of (1)
at t = 10, 50, respectively.

linearized the resulting differential equation as well as the resulting initial

condition. We deal with the resulting initial condition as it is without linearizing
it unlike in Parker’s [16] work. Following Parker’s [16] approach, it is easy to
arrive at the approximate solution of the GBE (1) subject to the initial profile
(8) and the boundary conditions (3) as
w̄x (x, T )
u(x, t) = −2δ , (119)
1 − (t + 1) M (1 − w̄(x, T ))
where
∞
  
jπ x
w̄(x, T ) ≈ c0 + c j exp(−( jπ ) T /l ) cos
2 2
, (120)
j=1
l

 l   πs 
1
c0 = exp k̃ cos − 1 ds,
l 0 l
=e −k̃
I0 (k̃), (121)

 l   πs   
2 jπ s
cj = exp k̃ cos − 1 cos ds,
l 0 l l
= 2e −k̃
I j (k̃). (122)

δ
π
Here T = 1−M ((1 + t)1−M − 1), k̃ = 2πδ
lu 0
and In (x)(= π1 0 ex cos θ cos(nθ )dθ )
are the modified Bessel functions of the first kind satisfying
x 2 y

+ x y
− (x 2 + n 2 )y = 0.
22 Ch. S. Rao and E. Satyanarayana

It is easy to see that for M = 0, the solution (119) reduces to

w̄x (x, δt)

u p (x, t) = −2δ . (123)
w̄(x, δt)

In fact, the solution u p (x, t) given in (123) (see (120)–(122)) is the known
exact periodic solution (see Sachdev [10], p.32) of the Burgers equation (6)
subject to (8) and (3).
Our numerical study showed that this approximate solution (119)–(122)
agrees well with the relevant numerical solution for M sufficiently small.

4. Conclusions

In this paper, we have studied the solutions of a GBEV (1) for large time. The
large-time asymptotics for periodic solutions of (1)–(3) have been constructed
by a perturbative approach in the manner of Sachdev et al. [2, 3] for 0 ≤ M <
1. It has been assumed that the relevant solution of linearized partial differential
equation of (1) satisfying (3) describes the large-time behavior of the solutions
of (1)–(3). Then large-time asymptotic expansion for solutions of (1)–(3) have
been obtained by including the effect of the nonlinear term. The large-time
asymptotic expansion of the solution of (1)–(3) contains the solution (9) of
the Burgers equation as a special case for M = 0. Similarly, we have also
constructed large-time asymptotics for the GBEV (1) subject to (2)–(3) for
the parametric region M = 1, δ > l 2 /π 2 . Large-time asymptotic expansions
obtained in the parametric regions 0 ≤ M < 1 and M = 1, δ > l 2 /π 2 for
the solutions of (1)–(3) are compared with relevant numerical solutions of
(1)–(3) for specific initial conditions and are found to be in good agreement.
Our numerical study suggests that the large-time behavior of solutions for the
initial–boundary value problem (1)–(3) is given by the solution of the inviscid
GBE for the parametric region M > 1. The parametric region M = 1, δ ≤
l 2 /π 2 will be dealt elsewhere.

References
1. C.N. DAWSON, Godunov-mixed methods for advective flow problems in one space
dimension, SIAM J. Numer. Anal. 28:1282–1309 (1991).
2. P.L. SACHDEV, B.O. ENFLO, CH. SRINIVASA RAO, B. MAYIL VAGANAN, and POONAM GOYAL,
Large-time asymptotics for periodic solutions of some generalized Burgers equations,
Stud. Appl. Math. 110:181–204 (2003).
3. P.L. SACHDEV, CH. SRINIVASA RAO, and B.O. ENFLO, Large-time asymptotics for periodic
solutions of the modified Burgers equation, Stud. Appl. Math. 114:307–323 (2005).
 Solutions of a Generalized Burgers Equation 23

4. B.O. ENFLO and O.V. RUDENKO, Evolution of a shock wave in the center of a bounded
sound beam, in 17th Scandinavian Symposium in Physical Acoustics (M. WESTRHEIM
and H. HOBAEK, Eds.), pp. 128–131, 1994.
5. B.O. ENFLO and C.M. HEDBERG, Theory of Nonlinear Acoustics in Fluids, Kluwer
Academic Publishers, Dordrecht, 2002.
6. D.G. CRIGHTON, Basic nonlinear acoustics, in Frontiers in Physical Acoustics (D. Sette,
Ed.), pp. 1–52, North-Holland, Amsterdam, 1986.
7. P.N. SIONÓID and A.T. CATES, The generalized Burgers and Zabolotskaya-Khokhlov
equations: transformations, exact solutions and qualitative properties, Proc. R. Soc.
Lond. A 447:253–270 (1994).
8. A.T. CATES, A point transformation between forms of the generalized Burgers equation,
Phys. Lett. A 137:113–114 (1989).
9. B. MAYIL VAGANAN and S. PADMASEKARAN, Large time asymptotic behaviors for periodic
solutions of generalized Burgers equations with spherical symmetry or linear damping,
Stud. Appl. Math. 124:1–18 (2010).
10. P.L. SACHDEV, Nonlinear Diffusive Waves, Cambridge University Press, New York, 1987.
11. P.L. SACHDEV, K. R. C. NAIR, and V.G. TIKEKAR, Generalized Burgers equations and
Euler-Painlevé transcendents. III, J. Math. Phys. 29:2397–2404 (1988).
12. J. DOYLE and M.J. ENGLEFIELD, Similarity solutions of a generalized Burgers equation,
IMA J. Appl. Math. 44:145–153 (1990).
13. B. MAYIL VAGANAN and M. SENTHIL KUMARAN, Exact linearization and invariant
solutions of a generalized Burgers equation with variable viscosity, Int. J. Appl. Math.
Stat. 14:97–105 (2009).
14. J.F. SCOTT, Uniform asymptotics for spherical and cylindrical nonlinear acoustic waves
generated by a sinusoidal source, Proc. R. Soc. London Ser. A 375:211–230 (1981).
15. J.F. SCOTT, The long time asymptotics of solutions to the generalized Burgers equation,
Proc. R. Soc. London Ser. A 373:443–456 (1981).
16. D.F. PARKER, An approximation for nonlinear acoustics of moderate amplitude, Acoust.
Lett. 4:239–244 (1981).

INDIAN INSTITUTE OF TECHNOLOGY MADRAS

(Received June 22, 2010)