You are on page 1of 11

CHAPTER II

POLYNOMIAL EQUATIONS
Finding roots of polynomials is the simplest problem which may be addressed by
perturbation theory. Nevertheless, this type of problem may give us insight into proper
formulation of the perturbation problem, singular and regular cases, uniform and
nonuniform solutions, rescaling coordinates and rescaling parameters. We will consider
the following examples adapted from Simmonds and Mann[10], pp. 3-17.
Regular Expansion.
EXAMPLE 2.1: Consider
(2.1)

D # #D !" !

with roots D" a%b and D# a%b. These roots may be found exactly by the quadratic formula

and a perturbation method is not required. But our methods will also work for
polynomials of higher degree which are not solvable by the quadratic formula. In this
case we will "cheat" and use the quadratic formula to verify our results. The following
discussion outlines the general method of attack in solving a problem using perturbation
techniques.
First, we notice that the constant .01 is relatively small when compared with the
other constants in the equation. If we replace it by zero, the equation is easily factored
with roots
D" ! !
D# ! #
Second, we create a family of problems intermediate between the easy, factorable
problem and the original problem by replacing !" with %. Equation (2.1) now has the
form
(2.2)

D # #D % !

and it represents a family of equations, one for each value of %. When % ! we have the
factorable or so called reduced problem and when % !" we have the target problem.
Third, we find an approximate solution to (2.2) assuming each root has a regular
perturbation expansion of the form

16

17

D% +! +" % +# %# S%$

(2.3)
then

D # a%b +!# #+! +" % #+! +# +"# %# S %$

substituting these expressions into a2.2b we get

+!# #+! +" % #+! +# +"# %# #+! #+" % #+# %# % S %$ !

combining terms and using the fundamental theorem of perturbation theory


+!# #+! !

+!

a#+! +" #+" "b% !

+"

"
#

"#

+#

"
)

")

a#+! +# +"# #+# b%# !


therefore,
D" a% b

"
"
% %# S %$
#
)

"
"
D# a%b # % %# S %$
#
)
We see that these are the roots of the reduced equation when % !.
Fourth, whenever possible, say something about the error of these
approximations. In this case we may compare our approximation when % !" to the
actual roots of the target equation.
Exact Roots:
D" !!&!"#&'
D# "**%*)(%
Approximate Roots:

D" a!"b !!&!"#&

D# a!"b "**%*)(&
with error of !!!!!". If the approximation is not good enough with six decimal digits
of accuracy then one may compute more terms of the root expansions.

18
This short example reveals an important aspect of regular perturbation problems.
That is, these methods can only be applied when the target problem is close to a solvable
problem (the reduced problem). In fact, we see that the target problem is replaced by a
sequence of problems of which the reduced problem is the first. Since the solution of the
reduced problem gives us the first term of the series solution it must be solvable for the
target problem to be solvable by the regular perturbation method.
Singular Expansion.
EXAMPLE 2.2: Find the roots of the singular problem

a2.4b

%D # #D " !.

Substituting a2.3b we get

%+!# #+! +" % #+! +# +"# %# #+! +" % +# %# " !

collecting terms and using the fundamental theorem


#+! " !

+!

"
#

a+!# #+" b% !

+"

"
)

+#

"
"'

a#+! +" #+# b%# !


so, we discover only one root!
D" a% b

" "
"
% %# S %$
# )
"'

What are the characteristics of the missing root? We try to find the second root by using
the quadratic formula. Applied to a2.4b, the quadratic formula gives
a2.5b

D" a% b

" " %
%

D# a% b

" " %
.
%

Taking l%l " we can expand " % in a power series in %. Using the binomial
expansion
" % " " % " %#
#
)

and substituting into a2.5b we find

19
D" a% b

" "
% S %2
# )

D# a% b

# " "
% S %#
%
# )

So, the missing root D# a%b has the characteristic that it approaches infinity as % p !,
indicating a singularity in the model. The singularity arises when % !. We say that the
expansion for D# a%b does not hold uniformly for all %, but the expansion for D" a%b is
uniform. The form of D# a%b suggests the change of variable

a2.6b

=a%b %D a%b

This change (rescaling coordinates) converts a2.4b to


=# #= % !

where the problem has been transformed from singular to regular behavior. The general
procedure of singular perturbation theory is to extract the singular behavior of a solution
and by a change of variable and/or parameter to reduce the singular problem to a regular
one. We have already solved this reduced problem. The solutions of example 2.2 were
=" a%b

"
"
% %# S%$
#
)

"
"
=# a%b # % %# S %$
#
)

Restoring the original variable using a2.6b we obtain


D" a% b

" "
% S %2
# )

D# a% b

# " "
% S %#
%
# )

where we have reproduced the roots of the singular problem by a change of variable
without recourse to the quadratic formula. This is satisfactory since our new method will
work for polynomials of degree # We also notice that it was necessary, by a change
of variable, to include negative powers of the parameter in the assumed form of our
perturbation expansion, a2.3b. D# a%b is not a regular expansion because it is not expressed
in positive integer powers of the expansion parameter as it must be to be of the form
a2.3b. As a result it does not approach a finite value as % p ! The change of variable

20
=a%b %D a%b worked for the specific case of a2.4b but will not work in general. In the
general case we will proceed by making the change of variable =a%b %P D a%b and allow
the specifics of the problem to determine a value for P which will transform singular
behavior to regular. This is a specific case of the method of undetermined gauges.
Undetermined Gauges.
EXAMPLE 2.3: Find an expansion for the roots of

a2.7b

D # #% D % ! .

Using expansion a2.3b in a2.7b we have

+!# #+! +" % #%a+! +" %b % S %# !

which gives us
+!# !

a#+! +" #+! "b% !

+!

"

a contradiction. Thus, we find no roots, indicating that this equation has no roots of form
a2.3b. Consider the change of variable
D a%b %P =a%b

where

=a%b +! +" % +# %#

with +! ! Then a2.7b becomes


a2.8b

%2P =# a%b #%P" =a%b %" !.

We solve for P by equating all possible pairs of gauge function exponents. Consider the
three possibilities that result.
i

#P P "

P"

ii

P""

P!

iii

#P "

"
#

Case i would leave % in the first and second terms and the singular behavior remains.
Case ii D a%b =a%b and the singular behavior remains.

21
Case iii let P "# , then a2.8b becomes
a2.9b

%=# a%b #%$# =a%b % !

=# a%b #%"# =a%b " !.

Make a change of parameter (rescaling parameter)


" %"#

to get rid of the fractional exponent. Then a2.9b becomes


a2.10b

=# a" b #" =a" b " !.

The regular expansion for = in parameter "

=a" b ,! ," " ,# " # S " $

is substituted into a2.10b to get

,!# #,! ," " #,! ,# ,"# " # #,! " #," " # " S " $ !.

Collecting terms and using the fundamental theorem we find


,!# " !

a#,! ," #,! b" !

a#,! ,# ,"# #," b" # !

,!

"

"

,"

"

"

,#

"
#

"#

Therefore,
"
=" a" b " " " # S " $
#
"
=# a" b " " " # S " $ .
#
Restore original parameter %

"
=" a%b " %"# % S %$#
#
"
"#
=# a%b " % % S %$#
#

and then restore original variable z to obtain the final result.

22
"
D" a%b %"# % %$# S %#
#
"
D# a%b %"# % %$# S %# .
#

a2.11b

We now see that in the case of singular problems it is necessary to include fractional
powers of % as well as negative powers in our perturbation expansions. These unknown
gauge exponents are determined after substitution into the target problem. Since the
regular series expansion will not work, the method of undetermined gauges is classified
as a singular perturbation method. As a check on these results we see that the roots of the
reduced equation are accurately represented by a2.11b when % p !
We will now state and prove a theorem giving us a procedure to obtain an
approximation for any polynomial perturbation. We will then illustrate the method with
an example.
THEOREM 2.1: Each root of
a2.12b

T% aD b %!! +! ,! % -! %# %!" +" ," % -" %# D


%!8 +8 ,8 % -8 %# D 8 !

is of the form
a2.13b

D a% b % P A a % b

where Aa%b is a continuous, analytic function of % for % in some neighborhood of !,


Aa!b !, and !4 4 0 8 Without loss of generality, !4 is nonnegative and
at least one of them is zero. Clearly we may also assume +5 ! unless
+5 ,5 !
Proof: First we find the proper values of P which cause theorem 2.1 to be
satisfied. Substituting a2.13b into a2.12b we have
T% %P Aa%b U% aAb V% aAb !

where

a2.14b

U% aAb %!! +! %!" P +" A %!8 8P +8 A8 .

Note that a2.14b collects all the terms with the lowest power of % for each power of A.
Then if %P Aa%b is a root of a2.12b
T% %P Aa%b !

for all %

23
a2.15b

lim U% aAa%bb !

%!

If we are to satisfy Aa!b !, a2.15b implies that at least two of the exponents of the set
I e!! !" P !# #P !8 8Pf

must have identical, minimal values. We agree that if the coefficient of D < ! then
!< <T does not appear in set I . If we take a value of P for which there is only a
single minimal value in I say !5 5 P, then we may divide a2.12b by the gauge function
%!5 5P and from a2.14b we have
a2.16b

lim %!5 5 P T% %P Aa%b +5 A 5 a!b !.

%!

Since Aa!b ! +5 ! is implied. This is a contradiction because if +5 ! then its


gauge exponent would not be in the set I So there must be more than one exponent in
I with identical minimal values for each proper value of P. We also see that if the
exponent value is not minimal then the resulting limit a2.16b would be undefined.
We proceed by finding all the proper values of P and their associated minimal
exponents to form the set
a2.17b

eaP" , /" b aP# , /# b aP4 , /4 bf.

a2.18b

X%4 aA %b %/4 T% %P4 A X!4 aAb L%4 aA %b

We use the elements of this set to rewrite a2.12b in the form

where X!4 aAb is the part that has no %'s and limL%4 aA %b !. By multiplying T% by %/4
%!

and changing the variables from D to A we have extracted the dominant part of T% as a
polynomial X!4 which is independent of %. The solutions of X!4 aAb ! for each 4 will
give values of A which, when substituted into a2.13b, will give is the 8 roots of a2.12b.
In general, the non-zero roots of the polynomials X%4 aAb ! need not be regular.
The !'s in a2.12b and the associated proper values and exponents aP4 , /4 b may be noninteger rationals. Thus, to obtain a regular expansion, new parameters must be
introduced. Let
a2.19b

% " ;4

where ;4 is the least common denominator of the set of exponents e!! !8 8P4 f
Then from a2.18b we form
V"4 aAb X4 aA " ;4 b

24

where

X%4 aAb X4 aA %b

The roots of X%4 aAb ! are identical to those of V"4 aAb !, but the non-zero roots of
the latter will have regular expansions in " of the form
Aa" b ,! ," " ,R " R S " R " .
EXAMPLE 2.4: Construct expansions for the roots of
a2.20b

T% aD b " % %# $%# D %$ a"' %bD % %# % % %$ D '

Step 1.
Set D %P A in a2.20b and determine the set of exponents I .

T% aD b " % %P" # $%# A %$%P a"' %bA% %#'P % % %$ A'


I e! " P $ %P # 'Pf

Step 2.
Determine the pairs (P4 /4 ) of proper values with associated exponents and
polynomials. Proper values and associated exponents may be determined by trial and
error or with the aid of a graph. A computer program illustrating the algorithm is
included as Appendix I. Figure 1 shows a plot of the exponent lines. The intersections
of minimal value may be found at P $% and P #&

25
2+6P

-3+4P
1+P
P
-5/2

3/4

-13

Figure 1.
Proper Values

$
(%
#
%
! X%" aA %b " % % # $% A a"' %bA
%
%"$# % % %$ A'
&
"$
#$#
#
%
"$ X%# aA %b % a" %b % # $% A a"' %bA
#
% % %$ A'
Step 3.
For each 4 determine ;4 , where ;4 is the least common denominator of the set of
exponents of X%4 . Make the change of parameter % " ;4 and list the associated
polynomials V"4 aAb.
a2.21b
a2.22b

% " % V"" aAb " " % " ( # $" ) A "' " % A%


" #' % " % " "# A'

% " # V"# aAb " #' " " # " #$ # $" % A "' " # A%
% " # " ' A'

Step 4.
The roots of X%4 aAb are identical to those of V"4 aAb but the non-zero roots of the latter
will have a regular expansion in of the form

26
Aa" b ,! ," " ,R " R S " R " .

a2.23b

Substitute a2.23b into V"4 aAb !, collect like powers of " , use the fundamental theorem
and obtain a sequence of equations in ,! ," ,R . Solve in order for the unknowns
,! ," ,R . From a2.2"b we get
" "',!% !
" '%,!$ ,% ,!% !

"
"
%
,!
/a5"b31#
"'
#
,%
From a2.22b we get

5 " # $ %

"&
,!
'%

"',!% %,!' !
'%,!$ ,# ,!% #%,!& ,# ,!' !

,! % #a"b5 5 & '


,%

$
,!
$#

Step 5.
Write down the roots D" a%b D' a%b via the change of variable D %P A.
D5 a% b

" a5"b31#
"&
(%
/
" % S %
#
'%

D5 a%b #a"b5 %&# "

$
% S %#
$#

5 " # $ %

5 & '

You might also like