Professional Documents
Culture Documents
EQUATIONS (PDE)
Structure
12.1 Introduction
Objectives
12.2 Formation of a PDE
12.2.1 Elimination Process
12.2.2 Oeanstrical Considerations
12.2.3 k u g h Subridivy Aspects
12.3 Preliminary Concepts
12.391 DtfmiUcnu
12.3.2 8pkm of,Equations
12.3-4 Methods of Solution
12.3.5 Classification of Second Order Equations
12.4 Lagrange Equation
12.41 -
Solution Its Geometrical Interpretation
12.4.2 Methods of Solution
12.43 General, Complete and Special Integrals
12-4.4 A Particular Equation
12.5 Higher Order Equations with Constant Coefficients
12.6 Summary
12.7 Answers to Exercises
12.1 INTRODUCTION
Physical systems are invariably defined by functions which depend on time and the
space coordinates. Temperature in the atmosphere, for example, changes from point
to point and is not the same at all times. Differential formulation of such systems
is essentially through partial differential equations. Ordinary differential equations
can, at most, turn out to be a kind of approximation.
After having learnt about fbrmulation, types, meaning and types of solutions and
the methods of obtaining solutions of ordinary differential equations, we would
explore the area of partial differential equations. We shall attempt to get an
overview of partial differential equations in this Unit.
Objectives
After studying this unit, you should be able to
distinguish between an ODE and a PDE,
formulate a physical problem in terms of a PDE dong with the initial
andfor boundary conditions,
distinguish various types of PDEs, and
solve special types of PDEs.
and u =1-a.
, Y
U = A ~ r -cospx;
~ t20, - a < x c a
The equation (12.2) is again a PDE. However, the graph of the function u does
not have an easy interpretation as was the case with the graph of the function
in Example 12.1. u (x, t) as before could be regarded a solution of (12.2).
Let us make the situation, a little more difficult. Let us choose u which
depends on an arbitrary function rather than a constant. For example, consider -
u, =f
r .
and u,,='- f + 1,
where Jdenotes the derivative off with respect to its argument (x - y).
Eliminant is obtained by adding the above two equations :
It is identical to the Equation (12.1). It also shows that the PDE (12.1) has a
host of solutions as represented by (12-3) with an arbitrary function f:
to obtain a d~fferentialequation.
12-2-2Geometrical Considerations
You may be wondering whether any geometrical meaning can be assigned to a
PDE and the corresponding family it represents. We will have to wait a little
longer even for a partial answer to this question. However, we shall examine a few
geometrical problems and obtain an equivalent formulation in terms of a
differential equation. We have noticed earlier that certain geometrical questions can
be reduced to a solution of an ordinary differential equation. It will, therefore, help
us to make a contrast between an ODE and a PDE.
Example 12-3
Find surfaces, all of whose normals intersect the axis of z.
Let z =f (x, y) represent one such surface. The direction ratios of a normal at a
point (x, y, z) of the surface are (S, ,S,, - 1). If (X, Y, Z) denotes a current
point on the normal through (x, y, z), then its equation is
X - x - y-Y
----- - z-z
L 4 -1
But this normal passes through a point on the axis of z. Hence a point
(0,0,Z) lies on it. Thus
I- -x _ - .-Y -z-z
fx-fv -1
Differential Equations
It implies that x f , =y f , ,
or equivalently xq - y p = 0. (12.6)
Here p and q stand for fX E zX,fy = z,, respectively. Now onwards we shall use
this notation. Solution of the PDE (12.6) will give us desired surfaces.
Example 12.4
Find surfaces whose normals, all intersect the circle 3 +y2 = 1 , z = 0.
Let z =f (x,y) represent one such surface. The equation of the nonnal at
(x,y, z ) of the surface is (Example 12.3)
Similarly, we obtain
-
If we look at the equations (12.6) (12.8), there does not appear to be an easy
way out to provide the geometrical contents of each of these.
Physical Problems
A rich variety of problems which can be formulated in terms of partial differential
equations are provided by the oscillations of a dynamical system with an infinite
number of degrees of freedom. We have already learnt in Unit 9 of the first course
as to how a dynamical system with finite degrees of freedom yields an ordinary
-
differential equation. Let us have a look at one specific problem vibrations of a
stretched string tied to two end points.
Vibrations in a String
Consider an elastic stretched string of length 1 fixed at its end points 0 (0,O) and
A (I, 0) in the ux-plane. Let u (x, t) denote deflection at any point x and at any time
t > 0 in the transverse direction of the string. The problem is to obtain the
differential equation which governs small transverse vibrations u (x, t ) and to state
4
explicitly the boundary and initial conditions satisfied by the function u (x, t).
It is part of the given data that the two end points 0 and A of the string are fixed.
Deflection u (x, t) must therefore, vanish at these two points for all t 2 0.Thus
The right hand side is the resultant of the tensions TI and T2 in the vertical
direction and p Ax is the approximate mass of the portion P Q of the string. Make
use of the equation (12.10) in the equation (12.11), to get
But
t a n 8 + A 8 -tan8
tan (8 + b e ) -tan8 =
I-A8tan8
Use has been made of the fact that A 8 is small, tan A 8 = A 8. If the quantities are
retained only upto their first order in smallness, the equation (12-12), in the limit
when Q +P, reduces to
- T a2u
P a ? '
a4 a a2u
where use has been made of the fact that tan 8 = - ; sec28 - = -.
ax ax a 2
Denote c2 = T/p, to obtain
The equation (12.13) is the governing equation for the deflection u (x, t) and is
named as one dimensional wave equation or simply the wave equation. It is called
one dimensional as only one space dimension, that is, x-dimension is involved in
it. The reasons for calling it a wave equation, will be discussed at a later stage. It .
must be noted that the dimensions of c are that of velocity and is called the wave
velocity. The wave velocity (c =
string and its stretching tension.
m) depends on the material density of the
Let the shape of the string and the velocity of each of its material point be
prescribed initially. It amounts to say that the functions u (x, t) and u, (x, t) are
given at some instant of time, say t = 0, for every x E (0,C). We can write it in the
following form :
These are referred to as the initial conditions. Thus the transverse small vibrations
in a fixed-end stretched string are governed by the DE (12-13), the boundary
Partial Differenttal
conditions (12.9) and the initial conditions (12.14). These three together are termed Equations (PDE)
as an example of an 'Initial Boundary Value Problem' written briefly as IBVP.
A host of different IBVP's can be created by changing the boundary and the initial
conditions of the vibrating string. Let us have a look at some other situations.
i) If initially, the material points of the string are at rest and the shape is
prescribed with fixed ends, then the boundary conditions (12.9) continue to
hold. However, the initial conditions get changed to
u, (x, 0) = 0
ii) Let the end at x = 1, of the string be free. Thus there can not be any force
acting at this end in the vertical direction. Since the only force acting' on any
part of the string is a component of the tension, the string must be horizontal
at.the end where it is free. Hence
The deflection u (x, t) may not vanish at a free end while the slope u, (x, t)
*
may not vanish at a fixed end. The condition of free end can be appraximated
by tying the string to a ring which rides on a vertical rod with little friction.
iii) Let one end say at x = 1 be supported elastically. Thus the end is neither free
nor fixed but is a combination of the two :
for some given a > 0. The equations (12.9), (12-16) and (12.17) provide us
with a qualitatively different type of boundary conditions.
Vibrating Membrane
The treatment of the foregone pages can be extended to cover small transverse
vibrations u (x, y, t) in a two dimensional perfectly elastic membrane which is
confined in a two dimensional domain D. Let S denote boundary of D and n be an
outwardly drawn unit normal vector at any point of S. Let the functions f (x, y) and
g (x, y) specify the shape and the velocity respectively at a point (x, y) of the
membrane at an initial instance say t = 0. The IBVP corresponding to different
physical situations can be stated as follows :
Differential Equation
Initial conditions :
Boundary conditions :
i) If the boundary S is clamped, then
for a > 0.
It must be noted that the boundary of the membrane can be partly clamped and
partly fiee. In fact, the beatingl of a d m or a plate will all be covered under the
above formulation. In the differential equation (12-18), dL refers to the wlocian.
For example, in cartesian coordinates (x, y),
a2 a=
'V = -
&+J'
Heating of a Finite Rod
We mhall consider the heat flaw due to conduction. Energy balance relation a d
Fourier's law of heat conduction art crucial to understand flow of heat f'rorn point
to point. Let V be a domain in a body with n as an outwardly drawn unit nonnal
vector at a point P of the boundary S of V. Let C, T, p, denote specific heat at
constant volume, temperature and density functions respectively. Let q denote heat
flux per unit area per unit time from outside to inside the volume V. The principle
of energy states that the rate of increase in the quantity of heat in V must be equal
to the rate of flux of heat into V. Thus
where the constant of proportionality k is the conductivity of the material. Ume the
relation (12.24) in the equation (12-23), to obtain
It must be noted that Gauss' theorem has been- used to write the last step. Since
above relation holds for arbitrary V , one gets
The equation (12.25) is the basic heat conduction equation which governs tran&r
of heat from point to point through conduction.
We must recmphsrsiseAat the normal gradient of temperature is associated with
the flux of heat. Let ur conider one dimenaionnl heat conduction in a rod of
length 2 whore one end it kept at conitant temperature and the other end is
insulated. Lst us assume the rod to lie dong the axis of x with one end at the
origin 0 and the other end at the point A. Let the end at 0 be at constant Pnrtlnl Diflerentld
Equations (PDE)
temperature To and the end at A be insulated. If T (x, t) denotes the temperature at
any time and at any point x of the rod, then the determination of T (x, t) is
governed by
Insulatian of the end at A of the rod means absence of flux of heat at A. This
yields second of the boundary conditions. Initial temperature of the rod is given by
*
the fbnction f (x). Physical conditions will guide us to a different kind of boundary
conditions. A sample of a few are given in the following :
a) Prescribed temperature at two ends
b) An m o u n t Qoof bent per unit time per unit m a ir supplied at one end the
other end ir h p t at zero tampemluae
T(1, t ) SO.
C) The two end6 of tho rod are inmulated
BT
-
h
(0, t) = 0 =%(l, t).
I
I
Let us formulate the problem of one dimensional melting of an ice block at 0°C
undtr the influence of constant heat supply Qoat one end. Let X ( t ) denote the melt
line and u (x, t) denote the temperature in the molten ice which is not removed as
mdting pmceedn. The two h c t i o n s u (x, t) and X (t) defining the temperature of the
molten ice and the position of melt line respectively are unknowns. m e boundary
value problem for the molten ice is governed by the following
The first of the boundary conditions ensures constant flux of heat at the end x = 0.
The second of the boundary conditions implies that the temperature of the melt line
is zero. Initial condition for the molten ice has no meaning as the melt line
coincides with the end x = 0 of the slab at t = 0.Thus,
there is one more physical aspect which must be taken into consideration at the
melt line. Consider the'melt lines at time t and t + A 1, so that
Thus a thickness A x =x (t) A t of the slab is enclosed between the melt lines X ( t )
and X ( t + A t). The amount of heat per unit time from the molten ice crossing the
melt line X ( t ) is equal to -4 [ X ( t ) ,t ] . This entire amount of heat is utilised to
melt the slab of thickness Ax. This also implies that there is no flux of heat across
the melt line X (t + A t). Thus
where p is the density of ice and L denotes the latent heat of melting. .Hence we
obtain
- k u , [ X ( t ) , t ] = ~ ~ ~ ( tt>) O
, (12.36)
-
The set of equations (12.33) - (12.36) provides the complete formulation for
determining u (x, t ) and X ( t ) . The equations (12.33) and (12.34) gives the solution
u (x, t ) in terms of the unknown function X ( t ) . Such a solution u (x, t ) is used in
the equation (12.36) which along with the equation (12.35) provides us an initial
value problem for determining the unknown function X (t).
Dirichlet and Neumann Problems
One comes across, quite often, problems which require solution of Laplace's
equation in two or three dimensions with appropriate boundary conditions. Let us,
for example, determine steady state temperature distribution u ( x , y ) in a two
dimensional domain D such that the temperature is prescribed at the boundary S of
D. If the transfer of heat is only through conduction, then the determination of u is
govemed by
where
au denotes the normal derivative of u. You will often come across these type
-
an
of boundary value problems in your engineering programme. There are special
names which are attached to these different types of boundary value problems. The
BVP govemed by the DE (12.37) and the condition (12.38) is called the Dirichlet
Problem. Neumann BVP is described by the equation (12.37) and the condition
(12.39). Sometimes a mixture of the normal derivative and that of the unknown
function u is prescribed at the boundary. The boundary condition
a > 0 along with the DE (12.37) is said to describe a mixed boundary value problem.
is exact and
Thus f can be determined from the following PDE which is nothing but an
expanded version of the above equation. :
This poses an interesting question that for solving an ODE, we must solve a PDE
of the above kind. We have already come across a few particular cases of the
above equation. For instance, P = x and Q =y reduces the above equation to the
equation (12.6). It shows that PDE may arise elsewhere in Mathematics to answer
some subsidiary questions which might crop up. It did happen in the case of
finding out a solution of an ODE through integrating factors.
- - - - -
12-3.1 Definitions
'
then (12-41) is called a partial differential equation. Here the unknown function
u (x, y, z) is the dependent variable and x, y, z are the independent variables. The
point (x, y, z) belongs to such a domain D in three dimensional space that u is
sufficiently smooth h c t i o n for every (x, y, z) in D. A function u (x, y, z) which
satisfies (12.41) identically in the independent variables when u and its partial
derivatives are substituted in the equation (12.41) is called a solution of the PDE
(12.41). we shall look for the totality of solutions and characterise individual
solutions by extra restrictions which might be imposed on u other than the equation
(12.41).
The PDE becomes an ODE if the number of independent variables is one. In a
PDE, the number of independent variables could be any number greater than or
equal to two. In the equation (12~41)~
we have chosen this number to be three.
Qe order of the highest derivative occuring in a differential equation is called the
order of the differential equation.
I
Thus, for any arbitrary function f, u =f ( x +y) is a solution of the given DE.
The PDE is of first order and the unknown u depends on two independent
variables x and y. Hence the general solution must involve one arbitrary
function and must depend only on one independent variable. The function f
depends on x + y . Hence f (x +y) provides the general solution of the given DE.
Exercise 2
VerifTi that u (x, t) =f (x + ct) + g (x - ct) is the general solution of c2u,= u,.
Partial Differentid
Equations (PDE)
The two equations (12.46) and (12.47) are both of order three and are derivable
from the same two first order equations. We are having two equations to determine
the same unknown u(x, y). It is an over determined system and fails, in general, to
provide a solution. We can have a look at still a simpler example of an over
determined system.
Let us assume that a solution is such that its dependence on x and y can be
separated in the form
+
where and yr are unknown fimctions. If 4 and y can be determined, then we do
obtain a solution. On substitution, we obtain,
2 2
$,= 1 - Y,
Left hand side is a function of x alone while right hand side depends only on y. If
assumed fonn of solution is possible, the two sides must reduce to a constant,
Thus
$2x = 1 - , g.v = h 2
where X is an unknown constant. Solving for +(x) and yr(y), we obtain
+(x) = k 3rx + a
v i y ~ = ~ ~ y + - p
Thus we do get a solution in the form u(x,y) = +(x> + YO.But the question
whether it provides the totality of solutions remains unanswered. To get a feel for
this unanswered question, let us apply this method to another equation :
I as a solution of the given DE where a,f3 and 6 are arbitrarf' constants. Now
t consider
Thus f,=& and hence f is another solution sf the given PDE (12.48). It is obvious
that we cannot make a choice of the constants a:f3 and 6 in the solution f12.50) to
obtain the solution (12.51). It, theaeforc, explains that this method fkils to yield
totality of solutions of the equation (12-48). It must also be noted (It is not being
proved, you are bang asked to assume it) that the equation (12.48) does not
possess a singular solution.
Separation of the variables of a solution could be of a different type. Let us
assume that
uix, _V) - $(x) YO)
provides a solution of (12.48). Here the separation of variables is in the product
form. On substituting 2roduct of u in the equation (12-48), we get
The two sides being firnctions of x or y alone, must reduce to a constant. Thus
Differential Equations
where 3L is an unknown constant. The two ordinary differentnal equations
and
+=c,ehX+c2e - a x
so that U(X,Y ) = e* Y [ ~ ~ ~ X + ~ ~ ~ X ]
where the constant c has been absorbed in c, and c,. Moreover a , P and h are
unknown constants to be determined from initial and boundary conditions.
Quite often the choice of the parameter h is governed by physical requirements of
a problem. For example if u(x,y) is treated as temperature governed by differential
equatior, (12-48) with y denoting the time, then one expects u(x,y) to vanish as y
approaches infinity. This physical requiren~entrestricts h to a negative real number.
Let h= - v2, so that, one obtains
2
u, (x,y) = a e-' sin (m + p)
and each different value of v provides a solution. The suffix v ic written with u in
the spirit that each value of v provides a solution u, (x, y). It amounts to state that
the equation (12.47) has an infinity of solutions.
Solution by Superposition
It 1s interesting to observe that if u, (x, y) and 15(x, y) a= two solutions of the
equation (12.48) then aul + pu2 i s also a solution for any arbitrary constants a and
p, It follows from the fact that the differential equation (12.48) is a linear equation,
so that
X
To evaluate the integral on the right, put v2y = ? and -- a. We obtain
6-
a Partld DIfferentid
z Eqs.(lonr (PDE)
Noticethat J(O)=I~'sinpdt=&sinp
-a
a
and S(a)=
-a
I r~'cos(at+~)dt
--
-
e- t
~ c2 o s ( a r + p ) [
-a
- 7
-a
- t2
aysin(at+p)dr
a
=- 5 J(u)
We have used the integration by parts to simplify the right hand side. J(a) is thus,
governed by an ODE of first order and an initial condition which can easily be
solved to give
2
--
J(a) = 6 sin p e
Thus
2
6 -- X
u(x, y) = -sin j3 e 4~
G
which coincides with the function given by (12.51). It does give an indication that
totality of solutions are possibly obtainable for the DE (12.48) with the help of
product form of separation of variables.
Product form of separation of variables method will turn out to be extremely
important. It does not mean that this method can be used in all situations. Its
applicability is quite often decided by the structure of the differential equation.
Example 12.7
Separating variables, find solution of
u,+uy=2(x+y)u
i.e. -
l@-b=2.-L*=k say.
0&
' w dy
where 3L is a constant. Hence
Z=(X+h)+
and *
dy
= (2y - X) w.
Example 12.8
I
I
Here the Laplace equation is given in polar coordinates.
Let u(r,8)= p r ) q@), so that the DE reduces to
jL
+a (r) = r- a
We notice that g2- 4AC = 4,O and - 4 for wave, heat and Laplace equation
respectively. Thus these equations are hyperbolic, parabolic and elliptic respectively.
It may be pointed out that a PDE could be of one type in part of the domain and of
another type in the other part. Solution of such PDEs is more difficult to obtain. For
example the trans-sonic flow is governed by a PDE which changes its character
from elliptic to hyperbolic through parabolic type. In aero-dynamic flights where
speds cross the velocity of sound, transonic flow is of utmost impomwe.
You may be wondering what the terminology of conic sections-hyperbolic, elliptic
and parabolic has got to do with the classification of a BDE or with the nature of
its solution. It indeed has got to do a lot. But a concrete explanation of all these
questions is beyond the scope of this course. We may however be content with a
mere analogical reasoning. If A, B, C and D are assumed as mere constants and
urn u, and up are replaced by 2,xy and 3 respectively in the equation (12.52),
then we obtain
az az .
where P,&, R are prescribed h c t i o n s of x, y, z md p = - * q =- -- , IS called
& @
Lagrange equation. It is required to determine a surface z = u(x, y) which satisfies
identically the equation (12.53). Such a surface is called an integral surface of the
Differential Equations
equation. This equation is quasilinear first order equation as it happens to be linear
in highest order denvatives p and q and could still be non-linear in the unknown
function u. We shall learn in this section all about the equation (12.53) and the
methods of obtaining its solution.
the point A are orthogonal to each other. Hence the vector (P, Q, R) lies in the
tangent plane at A and normal to any integral surface is orthogonal to this
direction. Alternatively one might say that the tangent planes at A for various
integral surfaces forms a coaxial planes passing through a single direction (P, Q, R).
This curve is called c.~uacteristiccurve or Monge axis. Thus the family of curves
through A lying on integral surface is given by the DE :
Definition
For instance the functions u = x +y and v = sin (x +y) are not independent on any
domain while u = x +y and v = x2 +y are independent. Linear dependence of
fimctions is a particular case of dependence of functions. If two functions are
dependent, these may not be linearly dependent. However the converse holds. In a
likewise manner, the definition can be extended to functions of more than two
variables. There is a simple criterion to test dependence of u(x,y) and v(x, y ) on a
domain D. We state the result without proof.
Partial Differential
Theorem Equations (PDE)
Two functions u(x,y ) and v(x,y ) are independent in a domain D if and only if
It shows that, in order, to obtain the general integral of (12-53),one must obtain
two linearly independent solutions of the subsidiary equation (12.54). It must be
noted in the above proof that the coefficients of p, q and the right hand side of the
equation (12.55) can not identically vanish as the functions u and v are independent.
a) - 2(xcBr+ydy+zdz)
---
and
Y 2 .cy2 +2
The first equation resulted from second and third tiaction whiXc the secor~d
one is &om fourth and second fraction. two independent solut~onsare,
therefore,
L
-- . J . -
- = const. and ----- = constant.
Y Y
Hence the general integral is given by
Exercise 4
Find the integral surface for the equation
I
A second iatcgd louad with the help of tbe first
Most often, it is easy to obtain one indepsndent solution of the ge equation.
The following cxampks illustrate that the knowledge of one integral helps to
obtain the second independent solution.
Esercise 5
Find the general integral of
p+3q= 5z+tan(y-3x)
Example 12.10
Find the gcneral integral of
xp+yq==z
and obtain the particular surface through the curve
yz= l , x + y = 1 .
Subsidiary equations are
Hence $I
Y
(- )
x Y
3 - = 0 for arbitrary hlnction ( denotes the general solution.
Alternatively, the general solution is given by
[ - $, i,=)
1 is an arbitrary point on the curve. It also lies on the integral
~urfacefor some yr. Thus
For a given point A (x, y, z) on the integral surface, knowledge of the normal vector
determines tangent plane uniquely. The normal vector (p, q, - 1) is obtained from
first order PDE (12.56) or in particular from (12.53) as the case may be. Since p
and q both can not be determined from the single equation, one obtains, at a point
of the integral surfkce, one parameter family of tangent planes. The envelope of
these planes determines the integral surface.
Consider for example the differential equation p + q = 0. The tangent plane at a
point A (xo,yo, zo) of the integral surface is given by
where ( a , - a , - 1) for arbitrary a are the direction ratios of the normal. Since the
point (x,, yo,zo) is itself arbitrary, the tangent plane may be written as
('-
Eliminate a to get z = - ---- as a solution. In the second case
4
I
I It gives z = (x - y ) [In !y - x ) - 11 as a solution of the PDE p + q = 0. But none of
1 . the solutions obtained is a general solution.
i e
I Notice that z =f(.r- y ) where f is an arbitrary function provides us with the general
.c
solution of the equation p + q = 0. It is trivial to note that z = a (x - y ) + p as well
as z =f (x - y) for arbitrary function f and arbitrary constants a and P are both
solutions of p + q = 0. The first one happens to be a special case of the second
solution when f is chosen to be a linear function. But the second solution is the
envelope of the first where P is regarded to be an arbitrary function of a.The first
one for arbitrary a and P is called the complete integral.
Definition
I A two parameter family u = $ (x,y, a, b) of solutions of (12.56) depending on the
parameters a and b is called the complete integral of the equation (12.56).
If for an arbitrary function f; b =f(a), then the envelope of the family is obtained
by eliminating a from
and eda= o .
This eliminant which involves arbitrary hnction f is called general integral of
(12.56).
We shall notice that general integrals can directly be obtained easily for Lagrange
equation while it is not the case for nonlinear equations. It is the complete
integrals rather than general integrals which are more relevant in the case of
nonlinear equations. For a quasi linear equation (Lagrange equation), complete
integral is a particular case of the general integral wherein the arbitrary function
gets replaced by a linear function involving two arbitrary parameters.
Example 12-11
Obtain the general and the complete integral of z = p x + qy.
Subsidiary equations are
-
general solution. Complete integrd is s two parameter family of planes. For
an arbitrary function cp such that b op (d,complete integral reduces to one
parameter f'miily of planes
dF
The gcnewl integral is'obtained by elirninat~nga from P.= 0 and = 0. Thus
rA'a
Example 12.12
Verify that 42 + (2+ y2) = 0 and z = ax + by + a2 + b2 are solutions of
z = p x + qy +$ + q2 &here a and 8, are ahltmry constants.
px+qy+p2+#=-1-r+--=z.
s rZ s+y2
4 Thus the first one which is free horn
arbitrary constants is d s o a solution. However, it can not be obtained from
the second for any particular choice o f a and b.
Let us obtain the envelope sf the two parameter family of planes. It is the
eliminmt of a and b from
It c rncs out to bc 43 + (x2 +y2) = 0 . Thus the first soluti~nis the envelope of
the fimily of solutions represented by z :: nx -: hv + d? t. h 2. Keeping an
analo~ynltb the nrintrc of solutions sf an ODE,th~ssolution wall be caIled as
t!t. singular solution of the given PDE. You must note that to get a g e n s d
solu~ion,c-i~tt: has to obtain cnv;lopc uf one paranetcr family oftangent
planes fort) -.f (LJ) ~hcsu-1'1san abltmry function
11) the above Exanplc 1 2 . 1 2 thc PDE happened 30 be a nonlincas one and.
dler~fire,1 ) lnerlrod
~ 01 siibsrd~wequations and t v mdepcndcnt
~ ~ mtcgrals
does not work here It 1s lbr ibis reason, we onl) venfied i~hctl~er l gsvcn
;
t(? be e s y fix US
farnctlm is a xnlit~tniwclf S ~ T Cdiffkrential cquatlnn. it pr(>~cd
to get tbc geraerai sihunoa ~ l i ~ n e v thc
e r equatl&nhappencd k: bs a quasisil~near
onc Hut for 3 monlan~arequation? wc h a e to be sattsfncd uitb onl? the
cc,r~;plcb~
i~tt~j;r;il.
~ltca ~ is
c ab1tral-y p:urmeter Thus
-
equahons whcrein complete integral plays a dominant role. WC define a new
a combination of x and y. k t 2 z (11) and 224 = x2 + ny 2
variablc u thio~~gin
It yields
Above equation defines the complete integral. It also shows that for a complete
integral, arbitrary function 4 in the general integral has to be a linear function. It
must be noted that the choice of the combination of x andy for defining u depends
on the structure of a PDE.
Example 12-14
Determine the complete integral of p + q = p q .
One can solve for q in terms of p from the above PDE, so that q = ----
P
@ - 1).
Now
1
=pdx +- dy.
P-1
I f p is chosen to be an arbitrary parameter a, then
a
Thus z=ax+---y+b
a-1
provides us with the complete integral.
It must be noted that a particular integral is not obtained by speciQing values to a
and b as is the procedure for an ODE. On the contrary, one has to assume b to be
a specific function of a &d the envelope of such one parameter family of planes
yields the desired result. For example let b = a , then
is the required particular integral. But it is almost impossible to get the general
integral.
It may be interesting to have a closer look at the geometrical meaning .)f an
integral surface of a nonlinear equation. For the Lagrange equation, the tangent '
planes at a point to all the integral surfaces formed a family of coaxial planes.
What is going to happen to this family in the case of a nonlinear equation ? In
order to provide a clear explanation, we concentrate only on a particular example
of a PDE, namely
Along with it, we shall consider its linear counterpart
Since (1, 1, 0) and (a, -a,- 1) are orthogonal vectors, the direction (1, 1, 0) lies
in each of the planes of the family.
Now for the integral surfaces represented by (12.57), (p, q, - 1) along with
p + q =pq is again a vector normal to a tangent plane. Hence the family of such
planes through (x,, yo,zo) is given by
a
a(~-~~)+--(y-~~)-(z-z~)=O (12.59)
a- 1
d
a+8
and (a+8)1+- -----m-n=0.
a- l + 0
Solving for I, m, n, one gets
that is,
This line generates a cone x - xo = z - z, through the point (x,, yo, zo). It is a
degenerate cone of semivertical angle of 90' (Plane can be regarded as a
degenerate cone). It shows that for a nonlinear equation, family of tangent planes
to- all the integral surfaces through a point envelopes a cone called Monge cone
whose generators are charactenstic curves. Integral surface is generated by these
Monge cones. Monge cones play the same role as lineal elements play in the
solution of a first order ODE.
For solving a nonexact ODE, one needs to know just one integrating factor. We
shall, in this context, be interested only to get a particular integral of the PDE. Let
us visualise the nature and complexity for finding integrating factors through
certain examples.
Example 12.15
Determine the integrating factor of the DE.
and
Y
The two independent solutions are - = constant and u$ -- constant. Hence the
X
1
general integral is given by u = -f ( y/x) where f is an arbitrary function. Por
3
different functionsf, a host of integrating factors can be written down. For
example - -
1
. 1;.
)P x XY
-I-are some of the integratiqj factors.
.Pnrtld DlRerentld
Equstlona (PDE)
"
Let
a
-- and
a
- be denoted by D and Dlrespectively, so that
8 u (x, Y )
may be
3y a?!q$'--r
written as Dr q -'u. The equation
where a,, a,, ..., a, are all constants and f (x,y) is a known function, is called nth
I
I order PDE with constant coefficients for determining the unknown u (x, y). In case
f= 0,the equation is said to be a homogeneous equation otherwise a
nonhomogeneous one.
It is desirable to make a few comments about the equation. Each term on the left
hand side of the equation is of order n. There could be lower order terms as well.
The presence of such terms gives rise to certain complications. We, therefore do
not consider such equations. For example u,,- c2 u,= 0 is a second order
homogeneous equation and is within the frame of the equation (12.61). However
u, - c2 u,= 0 is not covered under the present treatment. You may like to prove the
following two theorems which are Ocing stated without proof. You have already
come across such results while studying ODE
i Theorem
to get
a4
-(x,y)=x
2
+2xy+go(y)
@
where go is an arbitrary function. Further integration with respect to y yields
u (x,Y) = x 2 y + x ? + ~ ( Y ) + f ( x )
as the required general solution, where .f and g are arbitrary functions and
Example 12-17
Find the general solution of t = sin xy.
a2u
Integrate t = ---- = sin xy with respect to y to gct
au2
4
X" t a, 3,"- ' + --.+ a, = (3L - m,) ( A - m2) ... (A, - m,).
Substitute the expression for u (x, y) from (12.63) in the DE to get
Dn u + a l D"-'D, u I - U ~ D ; ' - ' ~u D
+ -; - - + D ; U
II , ( D L ~ D + 2 0 ~~ ; - 8 L $ ) u = O
~ 1D
i
1 is given by
I
' *
) ( Y +2~)+2fi
u (~,y,)'.fj( Y + ~ x+.yJ$ ( y + ~ ) .
The characteristic equation corresponding to PDE is
3L3-63L2+123L-8=(h-2)3=~.
I Thus 2 is a repeated root of multiplicity 3 of the characteristic equation. The '
DE can also be written as
(n- 2n1j3 =- 0.
Let (D -- u = I*.SO that PDE gets reduced to
Dlfrerrntid Equnnons
whose general solution is v =A ( y + 2r).Now denote (D- 20,) u = w to get
The first two equations yield y + 2r = constant = a (say). The third and the
first equation, then give
dw =f, (a) k,
that is, w - xfi ( y + 2x) = constant.
The general integral, therefore, comes out to be
w=xfi(Y+w+fi(Y+w.
You may similarly obtain u (x, y) after making use of the expression of w.
u(xy~)=f3(~+~)+xf,(~+~)+~fi(~+~)
where fi,f, and f, are arbitrary functions.
Example 12-19
Solve for the general solution
The characteristic.equation
k3- 4k2 +4X= 0
has the roots 0, 2, 2. The general solution, therefore, can be written as
u(~,Y)=f(Y)+g(Y+~)+xe(Y+~)
where g and 8 are arbitrary functions.
Particular Integral of a Nonhomogeneous Equation
We shall now develop a general method of finding a particular integral of the
equation (12.61). Since the left hand side car, be factorized, particular solution can
be obtained step by step after getting a solution of
The first two equations yield y + m, x = constant = c (say) as one of the independent
solution TIic second independent solution can be obtained from the first and the
thlrd equation,
dw =f (x, c - mix) dx.
is given by
Yxercise 52
- - - -
12.6 SUMMARY
A physical problem, depending on its nature, when formulated in terms of a
differential equation may yield an ODE or a PDE. The nature, complexities, types,
meaning and types of solutions along with a comparison between a PDE and an
ODE in all these aspects play a crucial role in understanding a problem.
Methods of solutions along with boundary and initial conditions do have
similarities as well as dissimilarities among ODES and PDEs. First order equations
and in particular Lagrange equation provide richness and base for the higher order
equations.
mncr#trd ~~ In this unit, you have learnt
formulation of a few real problems into BDEs,
types of PDEs, nature and method of obtaining a solution,
Lagrange equation and its solution, and
higher order equations with constant coefficients.
Hence u t t = c2 Uxx
Exercise 3
a~ az
Notice that z (x, y ) = O implies that p = - = 0 and q = - = 0 and hence is a
ax 4,
solution of the DE. Differentiating partially with respect to x and y
respectively, we obtain from the second relation
P=2&[1+f]
q =2 4 s
Thus p-q=2G.
It shows that the second relation gives us the general solution. However, the
solution z = 0 is not obtainable from the general solution for any choice off:
The solution z = 0 will be called a singular/special solution.
If we make a particular choice of the arbitrary function f say a linear function
then the general solution yields a particular solution
& = a + ( l + b)x+ by.
Exercise 4
Subsidiary equations are
dw d~ dz
Hence ---
lln(5z+tana.)=p.
3 5
On substitution for a , one gets
Exercise 6
Subsidiary equations are
a(x+1)+3a2y-z=0.
Let
(a+8)(x+ 1)+3 ( U + ~ ) ~ ~ - Z = O
that is,
The family of generators described by the equation (A) for different values of
a denote characteristic curves through (- 1 , 0 , 0 ) .
Exercise 8
In 'this equation M =y2 +xy, N = - x2 so that the PDE and subsidiary equations
can respectively bc written as
and -
4 -
dx - --- du
2 -y'+xj u (2y + 3x1
Each factor in the last equations is equal to
u
Hence -- = constant is one independent solution.
d
One may not bother to determine the general solution as it, straight away,
gives xy2 as an integrating factor.
Exercise 9
Notice that M = 3 2 y + 2xy +y3 and N = 2 +y2. The PDE and the subsidiary
equations ase, theacfore, given by
d(33.y +?)-
-3 d =
~-
a3 +.v3
Hence &-!e = constant and ( 3 3y +$) 'e = constant are the two independent
solutions of the subsidiary equztions. General integral of the PDE is given by
u = e3"f [ ( 3 2y +?) e3E] where ,f is an arbitrary function. It also shows that e3"
is one of the integrating factors.
Exercise 10
The PDE can alternatively be put as
4 +g ( 2 4x1
u (x,Y) =f( Y + 2 ~ + uo (x,Y)
where f and g are arbitrary functions and
where c is to be replaced by y + 2x. Thus
= fir" + 3x3.
Hence u (1, y ) = 3x2 ( x + 2 y ) +f ( y + 2 x ) + g ( 2 y + x ) .
Exercise 12
The equation can be written as
( D - L ) , ) ~ Ux=+ 2~y
If uo (1, y ) is the particular solution, then
where
= e 2y + x .
x + 2,:
Hence u (x, y ) =f ( x +y> + xg (x +y ) + e is the required general solution.