You are on page 1of 13

Mathematical Methods

(MATH F241)
BITS Pilani Dr. Amit Setia (Assistant Professor)
Department of Mathematics
K K Birla Goa Campus
Rayleigh-Ritz Method to find
Approximate solution

I  y  x     F  x, y, y '  dx 1
x2
Let min
y  x C  x1 , x2 
1 x1

under the boundary conditions


y  x1   y1 & y  x2   y2

Let y  x  be the solution to the variational problem.


Let B  0 , 1 , 2 ,..., n ,.... be the basis for the infinite
dimensional vector space C1  x1 , x2 .
Let y  x  be an approximate solution of y given by
n
y  x    cii  x  , where c0  1
i 0
BITS Pilani, K K Birla Goa Campus
The basis functions should be chosen in such a way that
the boundary conditions are satisfied by y  x 
i.e. y  x1   y1 , y  x2   y2
For this,
i) 0  x  must satisfy the boundary conditions
ii) 1  x  , 2  x  ,..., n  x  vanish at both end points of the interval.
 1 becomes
  

Let min I  y  x     F  x,  cii  x  ,  cii '  x   dx 2
x2

y x  x1
 i 0 i 0 
under the boundary conditions
y  x1   y1 & y  x2   y2
Our aim is to find the approximate
 n n

min I  y  x     F  x,  cii  x  ,  cii '  x   dx  3
x2

y x  x1
 i 0 i 0  BITS Pilani, K K Birla Goa Campus
1  x  , 2  x  ,..., n  x  are known basis functions,
the only unknowns are c1 ,......, cn
  3 
 n n

min I  c1 ,......, cn    F  x,  cii  x  ,  cii '  x   dx  4
x2

c1 ,......,cn x1
 i 0 i 0 
Using classical calculus,
I
 0, i  1, 2,......, n
ci
 we are having n equations in n knowns.

BITS Pilani, K K Birla Goa Campus


Example

Apply the Rayleigh-Ritz method to the BVP


y ''  x   y  x    x
with end point condition as
y  0   0, y 1  0

BITS Pilani, K K Birla Goa Campus


solution
First we need to convert the BVP into its
corresponding variational form
i.e. Multiply the equation by the variation of
the dependent variable  y  x  &integrate over 0 to 1

  y '' y  x  y dx  0
1

    y '  y'  y y  x y  dx   y '  y x 0  0


1 1
0

Now since  y vanishes at x  0 and x  1


    y '  y'  y y  x y  dx  0
1

 y '2 y 2
1 
     xy  dx  0
0
 2 2 
 
1
   y '2  y 2  2 xy dx  0
0
BITS Pilani, K K Birla Goa Campus
By Rayleigh-Ritz method,
let 0  x   0
1  x   x 1  x 
2  x   x 2 1  x 
.........................
n  x   x n 1  x 
 y  x   0  x   c11  x   .......  cnn  x 
let n  1  y  x   0  x   c11  x   c1 x 1  x 
 y '  x   c1 x  1  c1 1  x   c1 1  2 x 

 1     c1 1  2 x     c1 x 1  x    2 x  c1 x 1  x    dx  0
 1 2 2

0 
 11 1 
  c12  c1   0
 30 6 
BITS Pilani, K K Birla Goa Campus
 11 2 1 
 c1  c1   0
 30 6 
11 1
  2c1   c1   c1  0
30 6
 11 1
  c1    c1  0
 15 6
5
 c1 
22
5
 y  x  x 1  x 
22
The exact solution is
e x  e x
y  x  x 
e  e 1
BITS Pilani, K K Birla Goa Campus
5
 y  x  x 1  x 
22
The exact solution is
e x  e x
y  x  x 
e  e 1

BITS Pilani, K K Birla Goa Campus


Problem
Verify that the extreme values of the functional
I  y  x  
K  y  x   
J  y  x  
are the eigenvalues of the coresponding
Strum-Liouville problem


where I  y  x     py '2  qy 2 dx 
x2

x1

J  y  x     ry 2 dx  k
x2

x1

with y  x1   0 and y  x2   0

BITS Pilani, K K Birla Goa Campus


Solution
we showed that the extremals of the functional
I  y  x  
K  y  x    1
J  y  x  


where I  y  x     py '2  qy 2 dx   2
x2

x1

under the constraint


J  y  x     ry 2 dx  k  3
x2

x1

with y  x1   y1 and y  x2   y2  4
are the solutions of the Strum-Liouville's equations
 py ' ' qy   ry  0 5
with y  x1   y1 and y  x2   y2
where p, q, r are functions of x BITS Pilani, K K Birla Goa Campus
i.e. extremals of 1 are eigenfunctions

Let k be the eigenvalues of the Strum-Liouville Problem


 py ' ' qy   ry  0
with y  x1   0 and y  x2   0
and y  x   k  x  be its corresponding eigenfunction.

  pk ' ' qk  k rk  0 6


From 1 ,  2  ,  3 , we get

I  y  x     py '  qy  dx
x2
2 2

K  y  x    7
x1
  J  y  x   x2

2
ry dx
x1

BITS Pilani, K K Birla Goa Campus


  py '  dx    py ' y ' dx
x2 x2
2
Now
x1 x1

  py ' y  x  x    py ' ' y dx


x2 x2

1 x1

 py ' ' y dx
x2
 
x1

  7  becomes

  py ' ' qy  ydx


x2

K  y  x   

x1
x2  3
 ry dx
2
x1

Replacing y  x   k  x  in  3 , we get

  p ' ' q 
x2 x2

K  y  x   

x1 k k k dx

 x1
k rk 2 dx
 k
x2 x2
 r dx  rk 2 dx
2
x1 k x1

K k  x    k , k  1, 2,3,........
BITS Pilani, K K Birla Goa Campus

You might also like