A TEXT BOoK OF MECIIANICSs 149

Bsmple 3.2.2. Test for an eztreun the functional I y(z)) = (ay+-2y°/)dr:

subject to v(0) =1,y(1) = 2.
Solution. Herc f(T, y,y) ay+ y' - 21'/, sa functional, which is of the form
M+N y, where M = ry + nnd N= -212. The Euler's cquation (3.4) for this
problem is given by
OM ON
=0’ 1y+ 2a = 0.
Oy
Clearly, this extremal can not satisfy the boundary conditions y(0) = 1,y(1) =2. Thus,
an extremum cannot be achieved in the class of continuous functions.

Property 3.2.2. f depends on y' only : Here fis of the form f = f(v). Using
d
y=fayt yvt fyýyy', the Euler-Lagrange equation (3.3) becomes
(3.5)

for fy = fzy = Syy =0. Hence

fyy(y) =0’y= 0or fyy(u) =0.
If y=0, then y = c|t+C2 is a twO-parameter family of straight lines. If the equation
f d ) =0 has one or more real roots yk, then y = kËa +C, and we have a one
parameter family of straight lines contained in the two-parameter family y= C1z + C2.
Therefore, in the case f = f(y), the extremals are arbitrary straight lines y = C1T+c2.
Example 3.2.3. (Euclidean geodesic): Find the shortest distance connecting be
tueen two points, say (31, 1) and (z2, 2) on a surface S.(For a non-flat surface, the
curves of eztremal lengths are called geodesics).
Solution. Suppose that the general curve joining these two points A(*1,y1) and
B(T2, 2) is given by y = y(z) (Fig.3.2) in the ry-plane. By Phthagoras' theorem
the length of an arc ds along the curve APQB in a plane is ds= Vdr? + dy?. This is
a typical variational problem with a constraint as here we are required to minimize the
arclength Ijoining two points (1, yi) and (*2, V2) given by the functional,
B
dz.
3, Functionals Depend on Higher Order Derivatives
Let us now consider the extremum of a functional of the form
I= (3.10)
 TEXT BoOK OF MECIIANICS 157
A

Usingthe boundary condition y(z1) =n(2) =0, integration by parts gives

) de =
dz.

)de =

dr =
da2 )n(z) dz,
and so on. Thus from (3.12), we find that on the extremizing curve,

Since = n() is arbitrary deformation of the path, therefore

d
fy- t+(-1)" daznJytnys0, (3.13)

which is known as Euler-Poisson equation and the integral curves of this equations are
called extremals of the variational problem. Clearly the solution of this differential
equation of order 2n involves 2n arbitrary constants, which are determined by using
boundary conditions.

ample 3.3.1. Determine the eztremal of the functional I|u(-)] = Gtpudz

subject to y(-) =(-) = 0=v() =y().
Solution. Here f(z, y,y,y) = * + py, as a function. The Euler's equation (3.13)
for this problem is given by

d dy dy
da ,y= dr2

p-0+ y)=0; 2,p constants

u) +p=0’y=
24u
 158
VARINTIONAL
be obtained from the prescribed
where the constants c, C2, 3,CA cnm
tions y(-l) = v'(-)=0=y() = (). Using
the boundary conditions, we houndarhave,y conty
- - 6
24
P + 2+ cs =0
2

P+ +2 +cal +c4 =0
24u 6
+g = 0.
6
Solving these equations, we get, cË = 0, c3 =0, C2 =pl/12u, c4 =-pl /24n. Thus the
solution satisfying the prescribed boundary conditions is,
y=

This vibrational problem arises in finding the axis of the flexibly bent cylindrical beam
clamped at the ends. If the beam is homogeneous, then p and are constants.

Example 3.3.2. Find the curve which ertremizes the functional I= (

Jo
z)dz under the conditions y(0) = 0,(0) = 1, y(/4) = y(n/4) = 1/V2.
Solution. Here f(z, y, t , )= y-y + , as a function. The Euler's equation
(3.13) for this problem is given by.
d dy
y= do
y
da2
y-y =0’y(a) =ce*+ cge+ cg cOs r + c4 sin z,
where c, C2, C3 and ca are constants determined by using the conditions y(0) = 0.y(0) =
1, y(a/4)=y(/4)=1//2. Using the conditions, we get

1 1 1

Cye"/4 -m/4 1 1

Solving these equations, we get, cj = 2 = y = 0, and c4 =1. Hence the required curve
0s y(z) = sin z and

(sin a - sin r + a)da =

192
 TEXT BO0K Or MKONANICS
A

Ifthe functlonal I hAs nlocal maxtnmm or lnlnum nt u(), then u(r) nAtatlonnry
uncthon or 1, nnd for nll y wo must hnvo '(0) 0. To avnlunte this conlition for
thefumctlonnl given In utegrnl from abovo, wo necl to lrst cletermlne (n). By lirect
caleulatlon

d
o)= da
L (tr) +am(r);ve) +nk) dr
+

Oyk I(æ) dr +

k=] LOyk n(z) dr; as a(z1)= k(2) = 0.

Since '(a) =0 for e =0, therefore

d8f
k=] Loyk à Telr) dz =0.
Since nk = k() are perfectly arbitrary and independent of one another, the terms
within the square bracket are separately zero, i.e.,

Of
Oyk G0,
dr k=1,2,.",n (3.15)

oD the extremizing curve. Equation (3.16) is known as Euler-Lagrange equations each

of which must be satisfied for an extreme value.

Example 3.4.1. Findz andy as functions of t, so that=+u)-mgu

I dt
hay have stationary value, given that t(to) = 0, ylo) = nd r()= 9()= ||
 VARIATIONAL Pp
162
contajns
V.,)=(+)-mgy,as a tunction, two dependen
Solution. Here f(, (3.16) for this problem are given by
equations
Euler's
variables z and y. The
-=0
fo-=0andsy
dt
d and a(mý)+mg= 0
al )=0
and v()=-9t +dËt+ de.
a(t) = qtt cg
determined fromthe given conditionsslto) =
constants,
where, c1, 2, dË, d, are arbitrary
T0, Ylto) = yo and z(t1) = 1,y(ti) = y1.
AExample 3.4.2. Find the ertremals of the functional 2y:)
dz,
z(/2)= -1.
satisfying v(0) =0, y(n/2) = 1, z(0)= 0,
of Euler's equations (3.16)
Solution. Let f(z, y, z,y, z)= y++ 2yz. The system
are given by
d
=0and fy -fy =0
d
22- -(2y) =0 and 2y - da (2z) = 0
da
z-y=0 and y-" =0’y==y.
The general solution of this differential equation is given by
ulz) = c1e" + c2e+ cz cOS T + c4 Sin z
z(z) = = C1e + c2e- C3 COS T- c4 sin z.

where c,C2, C3 and c4 are constants determined by using the conditions u(0) = 0.
u(a/2) = 1,z(0) = 0,z(/2) =-1. Using the initial conditions, we get
C1 +2t c3 =0; 1 + 2-cg =0
cei +ei +c4 =-1; (e~ + cgei-cq = 1.
Solving these equations, we get, c1 = 2 = 0, c4 = -1,C3 = 0. Thus, the curves y =
- sin z; 2= sin T extremizes the functional I.

3.4.1 Euler-Ostrogradsky Equation

We now consider the problen1 of determining the
ple integrals leading to one of morepartial extrema of functionals involving multt
diflerential equations. For example, consider
 VARIATIONAL 'RINCIPLE
166

same way ng wo solve on ordinary cxtremum of afunction

IWe solve this problcm in the augmented functional as
A and u nnd forming the
by introducing two Lagrangians

multinlior.
functional as before and the Lagrange's
these
Now, we find the extremum ofequationsforthis new
functional and the given constraints.
are found from the Euler's
stationary path y = v(z) for the functionalI
Example 3.4.4. Find the
(0)=1; v(1)=1,() =l
boundarycondition y(0) =0 da da
dz subject to the introducp
second order derivative from the functional I, we
Solution. To eliminate the
two variables, namely, y1 =y and y2 = da'
dy as it is supposed to be a function of I.y
problem reduces to finding the externmum of the functional
only.Thus the
I= + daa

conditions y1(0) = 0,y2(0) = 1; y(1) = 1, y2(1) = 1 and the

with the boundary
dy1 =0. The constraint is incorporated into the functional I.
constraint condition y2
de the augmented functional as
using Lagrange's multiplierAand we formulate
I* = dy2

where, y=
dyi dy2 Here the integrand is
da da
f(a, y1, 2,i, s4) =1+ (4)+A(2-).
To fnd the optimal path of this functional, the Euler's equations for y1 is given by
af d of d
a=0’-)=0+A= constant.
To find the optimal path of this functional, the Euler's equations for y> is given by
d 0f
Ôy2
constant = A(say)
A
y2(z) = + Br +C.
 BoOK OF MECIHANIC:S 167
TEXT

thecondition
Ising
ya(0) =1,V2(1) =1 we get
C= 1; 2
+ B+C=lC1,+2B =0.

constraint condition y2 - 0, we have

I'singthe
dyi
= 2() = + Br +C
A B
u(z)=+,+Ca + D.
l'singthe boundary conditions yi(0) = 0,y1(1) =1, we get
A B
D=0; + +1=1’ D=0, A +3B =0.
6 2
Therefore., the solution is A = B=0,C = 1 and D 0. Hence the optimal path is
giren by
v(z) =y1(z) = ; 0<<1
and the corresponding value of the functional is

r-[+()) dr =
Jo
da = 1.

xample 3.4.5. Prove that the sphere is the solid fgure of the Tevolution which, for
a given surface area, has marimum volume.
Solution. The surface area is given by,

S= 2ryds = 2ry/1 + y2 dz
Jo

and the volume of the solid so formed is V= Tyda. Here, we are to maximize
Jo
we are to consider an
Vsubject tofixed S. Takingf= Ty and g = 2ry/1 +y,
auxiliary function H(,Y,/) as
H(,y,t) = f(z,y, /) + Ag(z, V, u/)
Ty+ 27ydy1 +y; A= Lagrangian parameter.
Hence,
iNoW, H has to satisfy Euler's cquation, where H does not contain æ explicitly.
,oH 2mAy 2 C.
H = constant = c ’ T + V1+y2
 168 VARÍNTrtONAL, PAIMCtptE
paRs through O(0,0) nnd A(a, 0), for whlch c 0, ael s0,

2\

When r=0,y =0 then k= 2\. Thus,

(o- 24)?+y²= (21)2

which is a circle with centre (2A, 0) and radius 2A. Hence the figure formed ty the
revolution of thc given arc is a sphere.
Ekample 3.4.6. Find the curve of length l which passes through the points (0.0) and
(, 0) and for which the arca betuween the curve and the z aTis is a TninimuTn.

Solution. Herc, is the fixcd perimeter of a plane curve passing through two given
points O(0,0) and A(l, 0). Let S be the area enclosed by that plane curve and
axis. The we are to maximize Su(=)] = | :y da, subject to the constraint Ly(z)) =

V1+y dz and the boundary condition y(0) = 0and y(1) = 0.Taking f= yand
g= V1+y, we are to consider an auxiliary function H(z,y,) as

H(z,y./) = S(z, y, y') + Ag(z, v, y')

= y+A/1+y; A= Lagrangian parameter.
Now, H has to satisfy Euler's equation, where H does not contain z explicitly. Hence,
OH d Ay'
=0’1
=c, integrating,c = constant

(z- c)' = dy ( - c)
1+y2 VA²- (*- c)?
y=-/A? - (u - e +e
(z +e)' + (u -e,)² = A,
which represents the equation of the circle.
 \5

Exanple 3.5.2. Findthe eatremum of the functional

I= | + +2yz| de
wth y(0) =0and z(0) = 0 and the point (2, 2, z2) noves over the fized plane = T2.

Solution. Let f(, y, z, z', , z') = 2+2+2yz, the Euler's equation gives z-y=0
and y" -2=0. These two equations gives an equation yl") y=0. The solutions are

y= CË cosh + C2 sinh z + Cg cos z + C sin z

= C cosh z + Ch sinh - C3 cos z Ca sin z.

The constants C1, C, Cy and Ca are determined by the given conditions y(0) =0 and
z(0) = 0, which give Cj = Cg = 0. Further, the condition at the moving boundary
point (T2, V2, Z2) can be derived from the condition that yz = 0, (since z2 is fixed) as
[Fyl= =0 and [F]=z, = 0
y(o2) = 0; and z(2) = 0.
Thus the above solutions lead to,

Cz cosh z2 + C4 cos T) = 0; Cz cosh 2 - C4 cos T, = 0.

If, cOs T, #0, then C2 = Ca = 0 and, therefore, an extremum is attained on y = 0, = 0.
But if cos T2 =0, then Gz =0and C4 remains arbitrary. In this case, the extrernal is
y= C4 sin z, 2 = -C¡ sin .
Example 3.5.3. Find the shorlest listance between the parabola y= ²and the struight
line z -y= 5.
 >0, for all.
Svy = 2 But for
attained on y=0 for a < T.
shows that a strong minimumis
is also attained on y =0.
3.6.5. Test for an eztremum the functional
AExample
(e+3) dz, y(0) =0, v(2) = 1.
I[v(z)] =Jo
U. /) = e +3. The Euler equation (3.3) gives
Solution. Here f(z,

()
d
=0>0-e)=0’ ey=0’=0.
y da
C1 T + C2, where c1 and c are
The solution of this ordinary differential equation is Y=
given functional is attained
arbitrary constants. This shows that, an extremal of the
only on the straight line. Since y(0) =0and y(2) = 1, we have
0= c2 and 1 = 2c1 ’ 1 1/2, c, =0.

Hence, the extremal satisfying the boundary condition is y = t/2 which is included in
the central field of extremals y = c1T.
Since fT,y, y) is three times differentiable with respect to y', for any y
S'y =e>0, for all y
it follows that on the straight line y = t/2a strong minimum is attained.
3.7 Exercise

[Multiple Choice Questions]

1. The functional l +(y+2/)+kyf +ylda, v(0) =0, u(1) =14 (0) =2
(1) =3 is path independent, ifk equals
(a) 1 (b) 2 (c) 3 (d) 4.

2. For Iy(e)] = f(a,3, U') da, I is called

(a) arc length (b) range (c) functional (d) None of these.
3. The possible values of a for which the variational problem,

Iu(e)](3y
Jo + 2r°y)dr, yla) =1
has extremals are
(a) -1,0 (b) 0, 1 (c) -1,1 (d) -1,0,1