Scribd
Upload
72 views2 pages

Tridiagonal Matrix Algorithm: A. Salih

The document discusses the tridiagonal matrix algorithm (TDMA) for solving systems of tridiagonal equations. A tridiagonal system is one with nonzero coefficients only on the main diagonal, lower diagonal, and upper diagonal. TDMA applies Gaussian elimination to transform the coefficient matrix into an upper triangular matrix. This allows the system to be solved through forward substitution to find constants, followed by back substitution to solve for the unknowns. A Fortran program implementation of TDMA is also presented.

Uploaded by

lomesh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
72 views2 pages

Tridiagonal Matrix Algorithm: A. Salih

The document discusses the tridiagonal matrix algorithm (TDMA) for solving systems of tridiagonal equations. A tridiagonal system is one with nonzero coefficients only on the main diagonal, lower diagonal, and upper diagonal. TDMA applies Gaussian elimination to transform the coefficient matrix into an upper triangular matrix. This allows the system to be solved through forward substitution to find constants, followed by back substitution to solve for the unknowns. A Fortran program implementation of TDMA is also presented.

Uploaded by

lomesh
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Tridiagonal Matrix Algorithm

A. Salih
Department of Aerospace Engineering
Indian Institute of Space Science and Technology, Thiruvananthapuram
October 2010

A system of simultaneous algebraic equations with nonzero coefficients only on the main
diagonal, the lower diagonal, and the upper diagonal is called a tridiagonal system of equations.
Consider a tridiagonal system of N equations with N unknowns, u1 , u2 , u3 , uN as given below:

b1 c1 u1 d1
a2 b2 c2 u2 d2


a3 b3 c3
u3
d3

... ..
= ..
. . (1)
.. ..

...
. .

aN1 bN1 cN1 uN1 dN1
aN bN uN dN
A standard method for solving a system of linear, algebraic equations is gaussian elimination.
Thomas algorithm, also called TriDiagonal Matrix Algorithm (TDMA) is essentially the result
of applying gaussian elimination to the tridiagonal system of equations.
The ith equation in the system may be written as

ai ui1 + bi ui + ci ui+1 = di (2)

where a1 = 0 and cN = 0. Looking at the system of equations, we see that ith unknown can be
expressed as a function of (i + 1)th unknown. That is

ui = Pi ui+1 + Qi (3)
ui1 = Pi1 ui + Qi1 (4)

where Pi and Qi are constants. Note that if all the equations in the system are expressed in this
fashion, the coefficient matrix of the system would transform to a an upper triangular matrix.
To determine the constants Pi and Qi , we plug equation (4) in (2) to yield

ai Pi1 ui + ai Qi1 + bi ui + ci ui+1 = di

or
(bi + ai Pi1 ) ui + ci ui+1 = di ai Qi1
or
ci di ai Qi1
ui = ui+1 + (5)
bi + ai Pi1 bi + ai Pi1

1
Comparing equations (3) and (5), we obtain
ci di ai Qi1
Pi = Qi = (6)
bi + ai Pi1 bi + ai Pi1
These are the recurring relations for the constants P and Q. It shows that Pi can be calculated
if Pi1 is known. To start the computation, we use the fact that a1 = 0. Now, P1 and Q1 can
be easily calculated because terms involving P0 and Q0 vanish. Therefore,
c1 d1
P1 = Q1 = (7)
b1 b1
Once the values of P1 and Q1 are known, we can use the recurring expressions for Pi and Qi for
all values of i.
Now, to start the back substitution, we use the fact that cN = 0. As a consequence, from
equation (6), we have PN = 0, which results in uN = QN . Once the value of uN is known we
use equation (3) to obtain uN1 , uN2 , u1 .

A Fortran implementation
The following Fortran code will solve a general tridiagonal system. Note that n is the number
of unknowns.

program TDMA
implicit doubleprecision(a-h,o-z)
parameter (nd = 100)
doubleprecision A(nd), B(nd), C(nd), D(nd), X(nd), P(0:nd), Q(0:nd)
c
A(1) = 0
C(n) = 0
c
c forward elimination
do i = 1, n
denom = B(i) + A(i)P(i1)
P(i) = C(i) /denom
Q(i) = (D(i) A(i)Q(i1)) /denom
enddo
c
c back substitution
do i = n, 1, 1
X(i) = P(i)X(i+1) + Q(i)
enddo
stop
end

You might also like