See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/250729009

Factorization properties of Chebyshev polynomials

Article in Computers & Mathematics with Applications · October 2005

DOI: 10.1016/j.camwa.2005.07.003

CITATIONS READS

32 439

3 authors, including:

Vilmar Trevisan Paul S. Wang

Universidade Federal do Rio Grande do Sul Kent State University
145 PUBLICATIONS 1,210 CITATIONS 149 PUBLICATIONS 1,703 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Paul S. Wang on 17 August 2020.

The user has requested enhancement of the downloaded file.

 An Intemational Journal
Available online at www.sciencedirect.com computers
&
~ , , . ~ ~_~o,..cT. mathematics
with applications
ELSEVIER Computers and Mathematics with Applications 50 (2005) 1231-1240
www.elsevier.com/locate/camwa

Factorization Properties of
Chebyshev Polynomials
M. O. RAYES
Southern Methodist University
School of Engineering a n d Applied Science
D e p a r t m e n t of C o m p u t e r Science a n d Engineering
Dallas, T X 75275, U.S.A.
mrayes©engr, smu. edu

V. TREVISAN
U F R G S - I n s t i t u t o de M a t e m ~ t i c a
91509-900 P o r t o Alegre, RS, Brazil
trevisan©ma~, ufrgs, br

P. S. WANG
D e p a r t m e n t of C o m p u t e r Science
Kent S t a t e University
Kent, OH 44240, U.S.A.
pwang0mcs, k e n t . edu

(Received June 2004; revised and accepted July 2005)

A b s t r a c t - - S e v e r a l factorization properties of Chebyshev are reported here. Studying the eu-

clidean division of two Chebyshev, we observe that the remainder .is either zero or (up to a sign)
another Chebyshev polynomial. This lead to explicit computation of the greatest common divisor of
two Chebyshev. We also obtain conditions for determining when a Chebyshev polynomial is divisi-
ble by another. Observing the modular representation over prime fields, we find two infinite sets of
fields Zp where a given Chebyshev polynomial factors completely into linear factors. We discuss how
to obtain the factors. (~) 2005 Elsevier Ltd. All rights reserved.

Keywords--Chebyshev polynomials, Polynomial factorization, Greatest common divisor, Modu-

lar roots, Primitive root.

1. I N T R O D U C T I O N
Chebyshev are of great importance in many areas of mathematics, particularly approximation
theory. Numerous articles and books have been written about this topic. Analytical properties
of Chebyshev are well understood, but algebraic properties less so. Some examples of algebraic
properties of Chebyshev studied may be seen in the references [1-3]. Other examples of algebraic
properties for Chebyshev include the work of Hsiao [4], who gave a complete faetorization of
Chebyshev of the first kind T~(x), determining which roots should be grouped together to yield

We would like to acknowledge the careful reading of two anonymous referees who caught mistakes and made this
paper more readable.

0898-1221/05/$ - see front matter (~) 2005 Elsevier Ltd. All rights reserved. Typeset by .4A,IS-TEX
doi:10.1016/j.camwa.2005.07.003
1232 M.O. RAYES et al.

irreducible factors with integer coefficients. Extending this result, Rivlin [5] adapts Hsiao's proof
for the Chebyshev of the second kind U~(x).
Reported here are several decomposition properties of Chebyshev including factorization and
divisibility. Conditions for determining when a Chebyshev polynomial is divisible by another are
developed. It is also shown that the remainder produced by Euclidean division of two Chebyshev
is again a Chebyshev polynomial, up to a sign. This fact leads to a direct computation of the
greatest common divisor of two Chebyshev. Presented also is the factorization of Chebyshev over
finite fields. Given any Chebyshev polynomial of degree n, two infinite sets of primes p are found
such that the polynomial can be factored into n linear factors over Zp. Procedures for finding
the modular roots are also discussed.
Let's begin with some basic definitions and properties of Chebyshev.

2. C H E B Y S H E V POLYNOMIALS
The Chebyshev of the first kind Tn(X) may be defined by the following recurrence relation. Set
To(x) = 1 and Tl(x) = x, then

T~(x)=2xT~_I(x) -T~_~(x), n=2,3, .... (i)

Alternatively, they may be defined as

T~(x) = c o s n ( a r c c o s x ) , (2)
where 0 < arccosx _< ~r. The roots of Tn(x) are real, distinct, within the interval [0,1], and given
by the closed formula,
(2k - 1) 7r
~k = c o s k = 1, .. n. (3)

It is easy to see also that the roots ~k are symmetric with respect to the line x = 0. In other
words, if ~ is a root of T~(x), then so is - ~ . For factorization purposes, the decomposition
properties,

T.,,n(x) = T~(Tn(x)), .~,n >_ O, (4)

1
T,~(x)T~(x) : ~ (Tm+~(x) + TH~_, I(x)), m, n >_ O, (5)

are useful. They can be proven using trigonometric identities [5, p. 5]. We can also define T-n(x)
as follows,
T_~ (x) = cos -n(arccos x) : cos n(arecos x) : T~ (x). (6)

The Chebyshev of the second kind are defined by setting Uo(x) = 1, Ul(x) = 2x, and the
recurrence relation,
Un(x) --- 2xUn_l(x) - Un-2(x), n = 2, 3 . . . . (7)

They may also be defined by

~ + 1 T.+l(x ) = sin ((n + 1) arccosx)

V~(x) sin(a~ocosx) (S)
It is easy to see that U~(x) is an integral polynomial of degree n. Its roots are all real, distinct,
symmetric with respect to the line x = 0 and are given by the expression,

kTr
r]k = COS n -}- 1 ' k = 1,...,n. (9)
 Chebyshev Polynomials 1233

Useful decomposition properties for the U polynomials include the following [6, p. 97],

Umn_l(x) = m, n > 0, (10)

=
1 + m, n > 0. (11)

To extend the definition of Chebyshev of the second kind for negative n, we notice that for n > 1,

1 1 1 J X
U_,~(x) - -- n- T4- 1' -,~+1 - n - 1T'_(n_l)(X)- n - 1Ti~_l)()=-Un_2(x). (12)

For convenience, we define U_l(x) = 0.

There are many fascinating properties of the Chebyshev and the reader is encouraged to refer
to the excellent books by Rivlin [5] and Suyder [6].

3. D I V I S I O N PROPERTIES
The division properties of Chebyshev Tn(x) and U,~(x) are characterized. Criteria to determine
when a Chebyshev polynomial is divisible by another are given. We also prove that Chebyshev
axe (essentially) closed under division. Specifically, we show that the remainder of the division
of two Chebyshev is, up to a sign, another Chebyshev polynomial. These results lead to the
computation of the greatest common divisor (gcd) of two Chebyshev.

3.1. D i v i s o r s of Tn(x)

The following property may be proven by applying the decomposition property (4).
PROPERTY 1. Let n > 1 be an integer. If h is any odd divisor of n, then Tn/h(X) is a divisor of

Let Tin(x) and Tn(x) be two Chebyshev of the first kind. Performing the Euclidean division,
we obtain integral quotient and remainder polynomials q(x) and r(x) satisfying

T~(x)=q(x)Tn(x) +r(x), deg(r) < deg(Tn). (13)

The q(x) and r(x) can be determined using the following result.
PROPERTY 2. Let m _> n be two positive integers. The polynomials q(x) and r(x) satisfying the
Euclidean division (13) are given by

1
q(x) : 2 ~-~(--1)k-t'lTm_(2k_l)n(X),
k=l

r(x) = (-1)lTIm_2~nl(X),

where l >_ 1 is the unique integer satisfying Im - 2ln I < n, if there is such an 1. Otherwise,

l--1
q(x) ~---2 Z ( - - 1 ) k-t-1 Tm_(2k_l)n(X ) 4- ( - - 1 ) / - 1 ,
k=l
r(x)=0,

where l satisfies m = (2l - 1)n.

1234 M. O, RAYE$ et al.

PROOF. Replucing rn by rn - n in equation (5), and using the extended definition (6), we have

Tin(x) = 2Tn(~)T~_~,(x) - Tm-2n(X), integers m, n. (14)

Let l be the smallest positive integer satisfying Irn - 2In I <_ n. Applying the decomposition
formula (14) 1 - 1 times, we deduce

= +... +

If Irn - 2In I < n, then deg(T~(x)) < deg(Tm-(2t-2)n(x)) a n d so we apply property (14) once
more, proving the first case. On the other hand, if m = ( 2 / - 1)n, then rn - { 2 / - 2)n = n. It
follows t h a t r(x) = 0 and the second case is proved. |
From the above property, we see that the remainder of two Chebyshev of the first kind is either
zero or another Chebyshev of the first kind (up to a sign). We may also deduce from Property 2
that if T,~(x) is a divisor of T,~(x) then n is a divisor of m and m / n is odd. This statement may
be seen as the converse of Property 1. The following theorem summarizes the results.

THEOREM 1. For integers 0 < n < m, Tn(x) is a divisor of Tin(x) if and only i f m -= ( 2 / - 1)n
for some integer l >_ 1. Otherwise, the remainder of the Euclidean division of Tin(x) by T,~(x) is
given by r(x) = (-1)ZTl,~_2,~l(x), where 1 is the unique integer satisfying }m - 2nil < n.

THEOREM 2. Let m, n be positive integers and g = gcd(m, n). Let n = gnl and m = gml. If nl
and m l ere odd, then
gcd(T~(x), T~(x)) = Tgcd(,~,~)(x) = Tg(x),

otherwise
gcd(T,~(x), T~(x)) = 1.

PROOF. By Theorem 1~ all polynomiM remainder sequence are Chebyshev, and so the

gcd(T (x),

is a Chebyshev polynomial. Hence, we need only to consider the common factors Th(x) of Tn(x)
and T,~(x). Let h be a common factor of m and n, say n = hnl and m = hml. By Theorem 1,
we see that the only common factors of T~(x) and T,~(x) are those Th(x) whose cofactors nl
and m l are odd. As Tg(x) is the highest degree polynomial satisfying these conditions, the result
is proved.
In particular, we can state the following.

COROLLARY 1. I[rn and n are odd, then gcd(T,~(x), Tn(x)) = Tgod(m.n)(x).

COROLLARY 2. / f r n or n / s a power d t w o and m ~ n, then ged(Tm(x),r,,(x)) = 1.

3.2. D i v i s o r s o f U,,(x)
By applying the decomposition property (10), we obtain the following.
PROPERTY 3. U,~(x) is a divisor of Um if there exists an integer I > 0, such t h a t rn = In + t - 1.
pRoo , urn(x) = (x) = (x)) |
To determine the Euclidean division of U,~ by U~, we use the extended definition for negative
indices of Chebyshev and apply equation (11) with m replaced by n + 1 and n replaced by n - m
and obtain
Um(X) = 2Tm_n(x)Un(x) - U2n--m(X), integers m, n. (15)
 Chebyshev Polynomials 1235

Because U - l ( x ) = 0, the above works for 2n - m -- - 1 . For m -- n, the formula still holds
and can be written as U,~(x) -~ (2Tm_n(X) 1)Un(x). Also, notice that 2n - m < n and if
- -

2n - m _> - 1 , we have the remainder and quotient determined.

If, on the other hand, 2n - m < - 2 , we may apply the extended definition for U2~-m(x).
Summarizing, we have

Urn(x) =
{ 2Tm_~(x)U~(x) - U2. . . .

2Tm_n(x)Un(x) + Um-2~-2,
if n < m < 2n + 1,

if 2n + 2 <_ m < 3n + 2.
(16)

If m > 3n + 2, we apply again the formula given by equation (15). In general, we have the
following.
P R O P E R T Y 4. Let m > n be two positive integers. Let

~= Trt--n

Then,
l
Urn(x) = 2U~(x)~_Tm_(2k+l)n_2k(Z ) -- U2<z÷l)+21_m(X ).
k=0
W h e n (re - n ) / ( 2 n + 2) is an integer, that is, re = (2l + 1)n + 2l, the above equation can be
rewritten as
l

and we have zero remainder. If re = (2l + 2)n + 2l + 1, we again have zero remainder because
U_l(x) = 0. In all other eases, the first term of the equations given in P r o p e r t y 4 determines
the quotient of the Euclidean division of U~ by U,~, while the second term gives the (nonzero)
remainder.
Using the extended definition (12), we have proved the following.
THEOREM 3. Let m >_ n be two positive integers. U,~(x) is a multiple of Us(x) if and only if
re -- (1 + 1)n + l for some integer 1 >>_O. Otherwise, the remainder of the Euclidean division of
urn(z) by V~(~) is give, by r(~) = --U2(z+I)n+2*-m(~), ~here
l= re-n

EXAMPLE. Consider re = a3 and n = 4. We have

l= L~lre-n =2,

determining that
u33(x) = 2V4(x)(T~9(x) + T~9(x) + Tp(x) ) - U_5(z).
This sets the remainder as -U_~(x), which is, by equation (12), equal to Ua(x).
THEOREM 4. Let m and n be two nonnegative integers, g = gcd(re + 1,n + 1). Then,

gcd(U~(x), V~(~)) = g~_l(~).

PROOF. Using Theorem 3, we see t h a t Ug_l(x) divides both Urn(x) and U~(x). Since Theorem 3
implies t h a t the gcd is a Chebyshev polynomial of the second kind, let us now suppose that
Uh(x) is a c o m m o n factor of Urn(x) and U,~(x). By Theorem 3, h satisfies m = (h + 1)/1 + h and
n -- (h + 1)/2 + h, for some integers 11 >_ 0 and 12 >_ 0. These two equations m a y be rewritten as
m + l -- ( h + l ) ( / 1 + 1) and n + l -- ( h + 1)(/2 q- 1). B y t h e definition o f g = g c d ( m ÷ 1,nq- 1), we
notice t h a t h + 1 <_ g, which proves that Ug-1 (x) is the Chebyshev polynomial of highest degree
dividing both Urn(x) and U~(x). II
1236 M . O . RAYES et al.

COROLLARY 3. I f m -~ 1 Or n ~- 1 is p r i m e and n ?t m , then gcd(U~(x), Urn(x)) = 1.

4. M O D U L A R FACTORIZATION
We now consider the factorization of Chebyshev over finite fields Zp. Specifically, we show the
existence of primes p for which the T~(x) (or Un(X)) factors into linear factors in Zp. Let ~k be
the roots of T~(x) defined in equation (3), for k = 1 , . . . , n, for some some fixed n. Notice that
5k = cos(2~/4n)(2k - 1), or

~k : (ei(21r/4n))2k--1-~ (e-i(27r/4n))2k-i W 2k-1 ~- W-2k+ 1

2 2 '

where w = e ~2"/4~ is a primitive complex (4n) th root of unity. Consider the field Q(w), the
rationals adjoined by w. We know by definition that

Q(w) = {(ao/bo) + (al/bl)W ÷ . . . + (as_l/bs_l)wS-l : aj, bj C Z } ,

where s = [Q(w) : Q] is the degree of the extension field Q(w) over Q. It is well known that
s = ¢(4n). As a remark, we observe that Learner's result [7] shows that [Q(w) : Q(w+l/w)] = 2.
Let p be an odd prime. We define

¢~(w) = {(ao/bo) + (al/bl)~ + . . . + (a~-l/bs_1)~-i : aj,b e Z, p ~ j } .

It is easy to see that Q~p(W) is a ring. Moreover, all the powers of w, including negative ones,
belong to Q~p(W). Let GF(q) be a finite field of characteristic p with q elements (q is some power
of p). Let us assume that GF(q) has a primitive (4n) th root of unity 8. Defining the natural ring
homomorphism,
• :g >Zp,

by kO(a) = a mod p, we can extend ~ to the polynomial ring Q~p(W)[X] onto GE(q)[x] in the
following way,
O(a/b) = O(a)/qt(b),

• (~) = o,

• (x) = x.

We see now that

(T~ (x)) = • (2 ~-1 (x - ¢ ~ ) . . - ( x - ~,))

0-2,~+1
(2F -1 (5 - • (2) /"
Since the quantities,

~(2) '
are well defined in GF(q), we see that q2(T~(x)) has all its roots in GF(q). Hence, we can find n
linear factors of T~ (x) modulo an odd prime p if either one of the following circumstances occur.
(i) The field Zp itself has a primitive (4n) th root of unity.
(ii) GF(q), a field with characteristic p, has a primitive (4n) th root of unity and all the
quantities ~23-1 _ ~-2j+1, j = 1 , . . . , n, belong to the ground field Zp.
 Chebyshev Polynomials 1237

T h e first situation is solved by the following.

LEMMA 1. L e t n and K be positive integers. I f p = 4 n K + 1 is prime, then Zp has a primitive

( 4 n ) th root of unity.
PROOF. A well known result states t h a t Zp has a primitive M th r o o t of u n i t y if and only if M
divides p - 1. |
For dealing with the second situation, we need a technical lemma.

LEMMA 2. L e t p be a prime. L e t c~ e G F ( p 2) be a root o f the irreducible polynomial f ( x ) =

x 2 + a x + b over Zp. For any c, d E Zp, we have

( c + d a ) p+I = e 2 - cda + d2b E Zp.

PROOF. As the a r i t h m e t i c is done modulo p, we have

p+l
(c nc da)p+l : j~.o ( P +
J l)cJ

= c p+l + (p + 1) c p da + (p + 1) c ( d a ) p + (da) v+l

= c 2 + c d a + cdc~ p -4- d2c~p+I .

T h e last equality is a consequence of F e r m a t ' s little theorem. Observing t h a t a p is t h e other

distinct root of f ( x ) , we see t h a t - a : a + ap, b = c~p + I a n d the result follows. |

LEMMA 3. L e t rt and K be positive integers. I f p = 4 n K - 1 is prime, then G F ( p 2) has a

p r i m i t i v e ( 4 n ) th root o f unity 0 and all the quantities 0 2j-a - 0 -2j+1, j = 1 , . . . , n, belong to the
ground field Zp.
PROOF. F r o m the fact t h a t 4n divides p2 _ 1 follows t h e existence of 0, a primitive (4rt) th r o o t
of u n i t y in GF(p2). Let f ( x ) = x 2 + a x + b be an irreducible p o l y n o m i a l in Zp[x] and let a be
a root of f ( x ) . Considering the a r i t h m e t i c of G F ( p 2) = Z p ( a ) , we denote 0 = c + d a , for some
e,d E Z v a n d c o m p u t e 0 -1 = (c - da - d a ) / ( c 2 - cda + d2b). It follows t h a t

e - ad
0_i_0--1 =C-4-
c 2 - c da + d 2b + a d-c2_cda+d2b .

To show t h a t 0 + 0 -1 6 Zp it suffices to show t h a t c 2 - cda + d2b --- 1. B y the technical L e m m a 2

above, we observe t h a t 0 p+l = c 2 - cda + d2b. As p + 1 = 4 n K a n d 0 is a primitive (4n) th root
of unity, it follows t h a t c 2 - cda + d2b -- 1. F r o m the i d e n t i t y

follows t h a t all t h e other quantities 0 2 j - 1 "~ 0 - ( 2 j - l ) belong to Zp. |

THEOREM 5. Let n > 2 be an integer. For ali the infinitely m a n y positive integers K for which
p -- 4 n K 4- 1 is a p r i m e number, Tn(x) has n roots in Zp.
PROOF. B y the results of L e m m a s 1 and 3, it remains to show t h a t t h e r e are infinitely m a n y
primes of t h e form p = 4 n K + 1 and p = 4 n K - 1. This follows from Dirichlet's t h e o r e m 1 for
(/, m) = (1, 4n) and for (l, m) = ( - 1 , 4 n ) , respectively. |

lIf 1 and m are integers with gcd(/, m) = 1, then there are infinitely many prime numbers p satisfying p -=
I (mod m).
1238 M.O. RAYESet al.

EXAMPLE. Consider T6(x) = 32 Z 6 - - 48 x 4 + 18 x 2 - 1. Primes of the form p = 4 n K + 1, include

p = 73, for K = 3 and primes of the form p = 4 n K - 1 include p = 23, for K = 1. We have

T6(x) ~ 3 2 (x + 30) (x T 59) (x + 1 6 ) (x + 1 4 ) (x T 43) (x + 57) (mod 73) (17)

T6(x) -= 9 (x + 19) (x + 4) (x + 10) (x + 9) (x + 14) (x + 13) (rood 23) (18)

The modular properties of the polynomials U~ (x) are similar to those of the polynomials Tn (x).
Observing that
w k + w -k
r/k = 2 ' k = 1,...,n,

where w = e 2~ri/2(n+l) is a primitive complex (2n + 2) th root of unity, one can show the following.

THEOREM 6. Let n >_ 2 be an integer. For aII the infinitely many positive integers K for which
p = 2(n + 1)K ± 1 is a prime number, U~(x) has n roots in Z v.

5. F I N D I N G THE MODULAR ROOTS

The problem of finding the modular roots of the Chebyshev T~,(x) has been reduced to that
of finding a primitive (4n) th root of unity in GF(q), where either q --- p = 4 n K + 1 or q --- p2 =
(4nK - 1) 2, where p is some prime number and K is a natural number. We first consider the
case, q = p = 4 n K + 1.
A possible approach is to find a primitive element/3 E GF(q) and then take 0 = ~(q-1)/an.
The density of primitive elements in GF(q) ensures that a simple search procedure choosing, at
random, a small number of elements is an efficient probabilistic procedure. In fact, the expected
number of multiplications rood p is O((logp)/(log log p)2), meaning that the search procedure is
O(((log p)4 /(log log p))2) bit operations [8].
However, the existence of an efficient deterministic search procedure is much harder and has
been considered elsewhere. We summarize next the results relevant to this note.
Improving on results of Wang [9], Shoup [10] proved that if the extended Riemann hypothesis
(ERH) holds, then the least primitive root modp is in O(logp) 6. Improving on this, Bach [8]
shows that, assuming the ERH, the least primitive root m o d p is in O((logp)6/(log logp)3). Bach
also describes an algorithm to compute a set of size O((logp)4(loglogp) -3) which needs to be
searched for a primitive element (not necessarily the least). This type of result immediately gives
an efficient search procedure for primitive roots modulo p; it shows how to construct a small set
of elements of Zp, one of which generates all the nonzero elements mod p. In fact, this can be
done in O ((log p) ~/ (log log p) 3) bit operations [8].
As an example, consider T6(x) and p = 73 = 4 × 6 × 3 + 1. We are looking for the 24 th root of
unity in Z73. The first primitive element of Zp is/3 = 5. The corresponding primitive 24 th root
of unity is 0 = 572/24 = 53 = 52 (mod 73). The actual factorization of T6(x) (mod 73) is given
above by equation (17).
We consider now the case p = 4 n K - 1, where the extension field GF(p 2) = GF(q) has a
primitive (4n) th root of unity 9, which we wish to find. Shoup [10] shows that, assuming the
ERH, there is a deterministic polynomial time algorithm for primitive roots in GF(pU), with no
restrictions on p.
We consider here an alternative probabilistic procedure for the case where p = 4 n K - 1. It is
well known that - 1 is a square modulo p if and only i f p --- 1(rood 4). Hence, for p --- 4 n K - 1,
we see that the polynomial x 2 + 1 is irreducible in Zp. Therefore, we may consider GF(p 2) as
the Gaussian integers, with arithmetic done modulo p.
The result of Lemma 2 gives us a key for a search procedure. We find solutions c, d E Zv to the
equation c2 + d 2 = 1. Compute the order t of the element/3 = c + id E Zp (~ iZp. Notice that t
always divides p + 1, by Lemma 2. If 4n divides t, then we take 0 -- fit/an as ore" primitive (4n) th
root of unity. We repeat the search until 4n divides t . Since p + 1 divides p2 _ 1, we know that
 Chebyshev Polynomials 1239

Algorithm Ml:toots(n, K)
INPUT: Integers, n, K
OUTPUT: Roots of T~(x) rood 4nK - 1
MRoots-1 p =- 4 n K - 1
MRoots-2 Choose randomly c and d E GF(p)
MRoots-3 Compute a --=c2 -t- d2 rood p
MRoots-4 If a ¢ 1 then GO TO Mroots-2
MRoots-5 Compute the order t of a E GF(p 2)
MRoots-6 If 4n divides t then 0 = (c + id) t/4n else go back to MRoots-2
MRoots-7 for k = 1 to n do
0 2 k - 1 ~- 0--2k+ 1
output (k -- mod p
2
Figure 1. A randomized algorithm for roots of Tn(x).

there exist elements of order p + 1 in Zp (~ i ~ p and this search will terminate. This algorithm is
shown in Figure 1.
As the number of primitive elements is ¢(q - 1), it follows that average number of trials the
algorithm MRoots performs is bounded by (q - 1)/¢(q - 1). It has the advantage of testing a
prior necessary condition, and such arithmetic is done modulo p, not in G F ( p 2 ) .
Similar algorithms may be implemented for finding the linear factors of Un, the Chebyshev of
the second kind. It suffices to replace Step MRoot-1 with p = 2(n + 1)K - 1 and the condition "if
4n divides t" in step MRoots-6 by "if 2n + 2 divides t". Step MRoots-7 also should compute r/k
instead of (k.
As an example, we take U3(x) and p = 23 = 2(3 + 1)3 - 1. Solutions (c, d) to c 2 + d 2 = 1
( m o d p) include (4, 10), (8, 11), (9, 9), (10, 19), (11, 15). The respective orders of the corresponding
elements are 24, 12, 8, 24, 3 and we may take 0 = (4 + 10i) 24/s = 14 + 9i as the primitive 8 TM root
of unity. The corresponding roots are 14, 0, 9.

6. C O N C L U S I O N
In this paper, several algebraic properties of Chebyshev polynomials of the first and second
kind have been presented. Also, tests for deciding when a Chebyshev polynomial is divisible
by another have been presented. Further, it has been shown t h a t the remainder produced by
Euclidean division of two Chebyshev polynomials is, up to a sign, another Chebyshev polynomial,
leading to the determination of the greatest common divisor of two Chebyshev.
In addition, this paper has discussed the problem of factorizing Chebyshev polynomials over
finite fields. It has been shown that, given any Chebyshev polynomials, two infinite sets of
primes p can be found such that Z~ contains all the roots of the polynomial. Finally, procedures
for computing the modular roots have been discussed.

REFERENCES
1. T. Bang, Congruence properties of Tchebycheff polynomials, Mathematica Scandinavica 2, 327-333, (1954).
2. L. Carlitz, Quadratic residues and Tchebycheff polynomials, Portugaliae Mathematica 18 (4), 193-198,
(1959).
3. R.A. Rankin, Chebyshev polynomials and the modular group of level p, Mathematica Scandinavica 2~ 315-
332, (1954).
4. H.J. Hsiao, On factorization of Chebyshev's polynomials of the first kind, Bulletin of the Institute of Math-
ematics Academia Sinica 12 (1), 89-94, (1984).
5. T.J. Rivlin, The Chebyshev Polynomials--From Approximation Theory to Algebra and Number Theory,
Wiley-Interscience, John Wiley, (1990).
6. M.A. Snyder, Chebyshev Methods in Numerical Approximation~ Prentice-Hall, New Jersey, (1966).
7. D.H. Lehmer, A note on trigonometric algebraic numbers, American Mathematical Monthly 40, 165-166,
(1933).
8. E. Bach, Comments on search procedures for primitive roots, Mathematics of Computation 66, 1719-1729,
(1997).
 1240 M . O . RAYES et al.

9. Y. Wang, On the least primitive root of a prime, (in Chinese), Acta Math. Sinica 9, 432-441, (1959); On the
least primitive root of a prime, (English translation), Sci. Sinica 10, 1-14, (1961).
10. V. Shoup, Searching for primitive roots in finite fields, Mathematics of Computation 58, 369-380, (1992).

View publication stats