Scribd
Upload
144 views

On The Leverrier-Faddeev Algorithm

This document presents a novel proof of the recursive formulas used in the Leverrier-Faddeev algorithm to compute the coefficients of the characteristic polynomial of a matrix. It derives the formulas using Laplace transforms. Specifically, it expresses the characteristic polynomial multiplied by a rational function in two ways using Laplace transforms. Equating the coefficients of the resulting expressions derives the recursive formulas and establishes the algorithm. The approach provides an alternative derivation that is more accessible to engineering and applied science students compared to other proofs.

Uploaded by

Milos Radojevic
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
144 views

On The Leverrier-Faddeev Algorithm

This document presents a novel proof of the recursive formulas used in the Leverrier-Faddeev algorithm to compute the coefficients of the characteristic polynomial of a matrix. It derives the formulas using Laplace transforms. Specifically, it expresses the characteristic polynomial multiplied by a rational function in two ways using Laplace transforms. Equating the coefficients of the resulting expressions derives the recursive formulas and establishes the algorithm. The approach provides an alternative derivation that is more accessible to engineering and applied science students compared to other proofs.

Uploaded by

Milos Radojevic
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

On The Leverrier-Faddeev Algorithm

SHUI-HUNG HOU
Department of Applied Mathematics
The Hong Kong Polytechnic University
[email protected]

Abstract
A novel proof of the recursive formulas that compute the coecients
of the characteristic polynomial of a matrix, as used in the Leverrier-
Faddeev algorithm, is given by means of Laplace transform.

1 Introduction
Let
p(s) = a0 sn + a1sn,1 +    + an,1s + an (a0 = 1)
= (s , 1 )    (s , n ) (1)
be the characteristic polynomial of an n  n matrix A, and de ne matrices
B1 ;    ; Bn by
adj(sI , A) = B1 sn,1 +    + Bn,1 s + Bn ;
so that n,1 +
(sI , A),1 = B1 s    + Bn,1s + Bn : (2)
p(s)
A well-known algorithm, often attributed to Leverrier, Faddeev and oth-
ers [1], indicates that the matrices Bk and the coecients ak can be obtained
in a successive manner by means of the formulas
B1 = I; a1 = , 11 tr(AB1 );
B2 = AB1 + a1 I; a2 = , 12 tr(AB2 ); (3)
.. ..
. .
Bn = ABn,1 + an,1I; an = , n1 tr(ABn );
with 0 = ABn + an I terminating as a check of computation. Here tr stands
for the trace of a matrix.
The above algorithm has been covered in most of the books on linear
system theory [2]{[5]. However, the majority of them do not give a complete
proof with the exception of a few, and the proofs therein are usually rather
involved and sometimes incomplete [6, 7]. It is the purpose of this paper
to give an alternative derivation using Laplace transform. The approach is
more readily accessible to students of engineering and applied sciences.

2 Preliminaries
The derivation of the Leverrier-Faddeev algorithm (3) consists of establish-
ing
B1 = I; Bk = ABk,1 + ak,1I; k = 2;    ; n; (4)
0 = ABn + an I; (5)
and the coecient formula
a = , 1 tr(AB ); k = 1;    ; n:
k
k k (6)
It is easy to see that (4) and (5) follow immediately from comparing
coecients of like powers of s in the identity
(sI , A)(B1 sn,1 +    + Bn,1 s + Bn ) = p(s)I:
In order to derive the coecient formula (6) we observe that (4) implies
B1 = I; Bk = Ak,1 + a1 Ak,2 +    + ak,1I; k = 2;    ; n: (7)
Substituting (7) into (6) gives
a = , 1 tr(Ak + a Ak,1 +    + a
k
k 1 k,1 A); k = 1;    ; n: (8)
Hence it follows that the desired formula (6) will be established if we can
prove (8).
3 The Derivation
In order to derive (8) we consider the following rational function of s :
G(s) = s ,1 +    + s ,n : (9)
1 n
Its inverse Laplace transform is given explicitly by
g(t) = L,1 [G(s)]
= 1 e1 t +    + n en t :
Clearly the derivatives Dr,1g(t) have the initial values
Dr,1g(0) = r1 +    + rn = tr(Ar ) = r : (10)
In addition, we have the following two expressions for p(s)G(s), which
are useful in our derivation of (8).
Lemma 1
p(s)G(s) = sp0(s) , np(s) (11)
with p0 (s) denoting the derivative of p(s).
Proof This follows immediately from (1) and the de nition of G(s) in (9)
by observing that
p0(s) = sp,(s) +    + s p,(s) :
1 n
The proof of Lemma 1 is completed.
Next, the Laplace transform of the kth derivative Dk g is given by
L[Dk g] = sk G(s) , sk,1g(0) , sk,2g0 (0) ,    , sDk,2g(0) , Dk,1 g(0);
which together with r = Dr,1 g(0) from (10) imply
sk G(s) = sk,11 + sk,22 +    + sk,1 + k + L[Dk g]: (12)
This enables us to express p(s)G(s) in another way.
Lemma 2
p(s)G(s) = (a0 1 )sn,1 + (a02 + a11 )sn,2
+    + (a0 n +    + an,1 1 ): (13)
Proof From (12) we have
G(s) = L[g];
sG(s) = 1 + L[Dg];
s2G(s) = s1 + 2 + L[D2 g];

s G(s) = sn,21 +    + sn,2 + n,1 + L[Dn,1g];
n , 1

snG(s) = sn,11 + sn,22 +    + sn,1 + n + L[Dng]:


Thus,
p(s)G(s) = (a0 sn + a1 sn,1 +    + an)G(s)
= a0 sn G(s) + a1 sn,1 G(s) +    + an G(s)
= (a0 1 )sn,1 + (a0 2 + a1 1 )sn,2 +    + (a0 n +    + an,1 1 )
+ L[p(D)g]:
This produces the desired result (13) by taking into account the fact that
L[p(D)g] = 0, because
h i
p(D)g = (D , 1 )    (D , n) 1 e1 t +    + nen t
= 0:
The proof of Lemma 2 is completed.
We now can conclude the derivation of (8). Substituting (11) into (13)
we obtain
(a0 1 )sn,1 + (a0 2 + a1 1 )sn,2 +    + (a0 n +    + an,1 1 )
= sp0 (s) , np(s) (14)
n, 1 n ,
= ,1  a1 s , 2  a2 s ,    , n  an :
2

Equating coecients of like powers of s in (14) immediately produces


,k  ak = a0 k +    + ak,11 ; k = 1;    ; n;
as desired.
References
[1] D. K. Faddeev and V. N. Faddeeva, Computational Methods of Lin-
ear Algebra, Freeman, San Francisco, 1963.
[2] T. Kailath, Linear Systems, Prentice Hall, Inc., Englewood Cli s, NJ,
1980.
[3] C. T. Chen, Linear System Theory and Design, Holt Reinhart and
Winston, New York, 1984.
[4] B. Friedland, Control System Design, McGraw Hill, New York, 1986.
[5] M. Jamshid, M. Tarokh, and B. Shafai, Computer-Aided Analysis
and Design of Linear Control Systems, Prentice Hall, Inc., Englewood
Cli s, NJ, 1992.
[6] L. A. Zadeh and C. A. Desoer, Linear System Theory, The State
Space Approach, McGraw Hill, New York, 1963.
[7] R. A. DeCarlo, Linear Systems, Prentice Hall, Inc., Englewood Cli s,
NJ, 1989.

You might also like