Module 1

Polynomials in one Variable.

Any linear algebra can be used as a ring of scalars; we are mostly concerned with a ring Kin general but the polynomial algebra K is also needed when the subject develops.We provide a discussion here. Recall that for any ring K two algebras K of polynomials as polynomial arise: one in which we write polynomial as a() = n an + . . . + a1 + a0 , ai K, an and the other in which we write polynomial asa() = an n + . . . a1 theta + a0 , ai K, an = 0, As modules, one of them is a right module and the other is a left modules and the development run in parallel.As rings they can be regarded as isomorphic via f (an n ) = n an the element an K is called the leading coecient of a()

(i) We take a ring K[] of polynomial writing its elements as a = a() = a0 + . . . + an n , an = 0 and recall that it obey the following relations: (a) If a() = a0 + a1 + . . . + an n , an = 0 b() = b0 + b1 + . . . + bm m , bm = 0 and c() = (a())(b()) = c0 + c1 + . . . + = (ab)0 + (ab)1 + . . . then while this multiplication in K[] is in general noncommutative [cn = (ab)n =
n=p+q

ap bq =

n=p+q bp aq

= (ba)n in general]we have deg [b()a()] = deg [a()b()]

[= deg (a()) + deg (b()) = n + mif either an or bn is not a zero divisor in Ki.e. if we do not get an bm = 0K although an = 0, bm = 0 note that this may happen if K is a matrix ring M atn (L) for some ring L] (b) deg [a() + b()] max{dega(), degb()} = max{n, m}

(c) dega() = 0if f a() = a0 K, we say that a = a() is a constant.

(d) We set deg (0) = for the zero polynomial oK[].

c Copyright Reserved IIT Delhi

(e) K[] in cen K[]}.

2 [i.e.a() = a() for all a()K []] and thus each monomial {k /kN is

(ii) (a) Proposition If () has an invertible leading coecient and dega() deg(), we can nd some b() with deg [a() ()b()] < dega().
1 n Proof : Say a() = a0 + . . . + an+m n+m , () = 0 + . . . + m m then with b() = m am+n , 1 n+m 1 n+m1 we have a() ()b() = a0 + . . . + an+m n+m m m1 m an+m m an+m 1 n n+m . . . 0 is 0. m an+m which has degree < n + m = dega( ) since the coecient of

(b) We note that the same argument works to prove that we can nd c() with deg [a()c()()] < dega()

The left division algorithm : Given a()K[], and ()K[] with an invertible leading coecient, there exists exactly one q ()K [] such that a() = () + r(),degr() < deg() and then r() is also uniquely determined we say q is the quotient and r is the remainder. Proof : If there are two polynomials q (), q () satisfying the requirement so that a() = ()q () +

r() = ()q () + r (), we have ()[q () q ()] = r() r () (1)

Assume deg [q () q ()] = n, deg () = m so that LHS of (1) has degree n + m [ () has invertible
1 coecient m and thus if ] is non zero we do not have m = 0 since = m (m )

But deg (r()) < deg(), deg (r ()) < deg() and deg [r () degr()] max{degr (), degr()} < deg() and so we have n + m < m which forces n = and thus q () = q () which then forces r () = r() and consequently uniqueness is established for q () and then for r() = a() ()q (). For existence, we note that the set S = {deg [a() ()b()]|b() k []} will have a least element; let q () correspond to that i.e., let q () be such that deg [a() ()q)]. We record r() = a() ()q (); thus degr() is the least element of S . If degr() deg() then [(iii)(a) on preceding page 11] we know there is some b() with deg [r() ()b()] < degr() so that with q1 = q () + b(), we have deg [a() ()q1 ()] < degr() which contradicts the choice of q () that had assumed deg (a() ()q1 ()) as the least element of S. Therefore, we must have deg (r()) < deg (()) and we have found our q () with a() = ()q () + r(), degr() < deg() as required. [this argument works for dega() deg();if dega() < deg() putq () = 0, r() = a(). Further,note
c Copyright Reserved IIT Delhi

3 that if dega() = 0, then deg() = 0 is forced since deg() = is not permissible with invertible
1 leading coecient; then a() = a0 , () = 0 and 0 is invertible, take q = 0 a0 ].

(v) In this preceding, q is the left quotient and r is the left remainder on left division by ();we say () is a left divisor of a if f r=0. we similarly have The right division algorithm: Given a() K [], and ()K[] with invertible leading coecient, there is exactly one q ()K[] such that a() = q ()() + r() with degr() < deg() and then r() is also uniquely determined. Further, we say q ()is the right quotient, r() is the right remainder on right division by (). [The proof will use (iii) b on pase 11 preceding ].

2 3 1 0 1 2 0 Example 0.1. If a3 = ,a0 = a1 = , 1 2 0 1 3 0 0 2 1 1 1 ,0 = where a() = a0 + a1 + a2 2 + a3 3 1 1 0 2 () = 0 + 1 . The right quotient 1 3 7 14 q2 = ,q1 = 0 0 2 5

1 and 1 = 2

and remainder are q () = q0 + q1 + q2 2 , r() = r0 with 43 81 43 206 ,q0 = and r0 = while the left 12 24 11 62 94 169 quotient is q () = q0 + q1 + q2 2 and the left remainder is r0 = on left and right 115 206 division of a() by () respectively.[Please verify this].

Proposition For a ring K and , K, the following are equivalent:

1. There is an invertible yK such that = 1

2. There are invertibles p(), q ()K[] with = p()( )q ()

Proof :

(1) (2) Since cenK[] and the constants , 1 K[], we nd = 1 1 =

1 1 = ( ) 1 which prove (2). (2) (1)Let invertibles p(), q ()K[] be given with = p()( )q ().Since has leading coecient 1 = 0 the division algorithm assume the existence of uniquely given a()and u()such that
c Copyright Reserved IIT Delhi

4 (p()1 )K[] can be written as (p()1 ) = ( )a() + u(), degu() < deg ( ) = 1 which forces u = u()K. then u( ) = [p()1 ( )a()]( ) = (p())1 ( ) ( )a()( ) = (p())p()( )q () ( )a()( ) = ( )[q () a()( )] comparing the highest degree terms on both sides, we get q () a()( )u The equation is thus u( ) = ( )u which has forces u = u. Dividing p() by ( ) [which has leading coecent 1 = 0 ] we get p() = ( )b() + r(), with degr() < deg ( ) = 1 forcing r() = rK. We have 1 = p()[p()]1 = p()[( )a() + u] = p()( )a() + p()u = ( )[q ()]1 a() + [( )p( + r)]u = ( )[q ()a() + b()u] + ru so that 1 ru = ( )[q ()a() + b()u] comparing the coecient if on both sides,we get RHS=0 hence 1 ru = 0 which forces ru = 1. Further,we have 1 = [p()]1 p() = [( )a() + u]p() = [( )a() + u][( )b() + r] = ( )a()( )b() + ( )a()r + u( )b() + ur so that 1 ur = ( )a()( )b() + ( )a()r + u( )b()Comparing the coecients of on both sides, we get RHS=0 which forces 1 ur = 0. Therefore, u is invertible with inverse r. Euclidean domain : We say a ring is an entire ring if f = 0 either = 0 or = 0; a commutative entire ring is also called integral domain.[Some text books use integral domain for entire ring also ]. An integral domain K is called a Euclidean domain if f there is an Euclidean functionK|{0} g N satisfying E1 if divides [= 0] then g () g () E2 For every pair of elements , of K, = 0 there exists elements , K with = + with = 0 or g ( ) < g ().[thus in a domain, it is called Euclidean algorithm]. In the preceding section, we proved that if F is a eld, F[] is a Euclidean domain [g (a()) = dega()] The ring of integers Z is also an integral domain [ This is the reason for integral domain ] with g (a) = |a|, aZ. Indeed, if b = ac = 0 then |c| 1 and hence |b| = |c||a| |a|. Further, for any two integers a, b, a = 0 the division algorithm in Z ensures b = |a|q + r = a(q ) + r with r = 0 or 0 < r < |a|. These two examples of Euclidean domains are the ones we shall use in this course .However, there are other Euclidean domains: let d = 1 be a square free integer [in the sense that its prime factorization has no square] and let = {
1+ d 2

if d1mod4 otherwise

Consider Q[ d] [we met this on pase 8 preceding ] which is a eld; for = p + q d we writep2 q 2 d as N (). Then for d = 1, 2, 3, 7, 11, 2, 3, 5, 6, 7, 13, 17, 21, 29, the function g () = |N ()| is a

c Copyright Reserved IIT Delhi

5 Euclidean function; we shall not prove this here. Proposition The gcd of any two elements , not both zero, of Euclidean domain E exists and can be expressed as + with , E. Proof : Suppose g is the Euclidean function, g () g ( ). Then by the property of g [the Euclidean

algorithm,also called division algorithm ] we have = 1 + 1 , g (1 ) < g ( ); = 1 2 + 2 , g (2 ) < g (1 ) ... ... ...

Thus g ( ) > g (1 ) > g (2 ) . . . is a decreasing sequence of nonnegative integers which must stop and after some time, we have n2 = n1 n + n , n1 = n n+1 + n+1 with n+1 = 0 = k for any k > n + 1 now 1 = 1 so 1 has the form + with = 1, = 1 . In general, if i1 = i1 + i1 and i2 = i2 + i2 i = i1 i + i2 = (i1 + i1 )i + i2 + i2 = [i2 i1 ] + [i2 i1 i ] also has this form + [with = i2 i1 , = i2 i1 i ]. Thus n = n + n for some n , n E. Now n = n .1+0 so n divides n and n1 = n n+1 +0 so n divides n1 . But we have n2 = n1 n + n = n n+1 n + n = n [n+1 n + 1] so n divides n2 . Similarly, n divides all the remaindersi .Let i = n pi , pi E. Then = 1 2 + 2 = n p1 2 + n p2 = n [p1 2 + 2 ] so that n divides and = 1 + 1 = n [p1 2 + 2 ]1 + n p1 = n [p1 2 1 + p1 ] so that n divides . Thus is a common divisor of and . If is any other divisor of and say = p, = q, p, qE then n = n + n = pn + qn = [pn + qn ] so that divides n .Thus n is the greatest common divisor of and , and is of the form + [with = n , = n ]. This proves that advertized result.But there is a further piece of information.Suppose = n a, = n b then writing n m = ,
2 2 we have n m = n an b = an b = n ab [because of commutativity ] i.e. m = a and m = b [ E

is a integral domain ,n = 0 ] so that m is a common multiple. If m is another common multiple, then m = c1 , m = c2 hence n m = (n + n )m = m n + m n = c2 n + c1 n = [c2 n + c1 n ] = n m[c2 n + c1 n ] But then n [m m(c2 n + c1 n )] = 0 and since n = 0, and E is an integral domain,we get m m(c2 n + c1 n ) = 0 which means m = m(c2 n + c1 n ) so that m is a divisor of m.Thus m is the least common multiple of , . To sum up: In a Euclidean domain, both the LCM and GCD exists. [There are rings in which the LCM exists and the GCD dose not and vice versa; this is not the place

c Copyright Reserved IIT Delhi

6 to go in more detail] Now suppose A is an ideal in a Euclidean domain E [For ideal see Modules, pase 14 ]. Pick aA with a = bq + r where g (b) has the least value for element of A.Then since a, bA, we have r = a bqA with g (r) < g (b) which is not possible since b has the least value g (b) for elements of A.Thus r = 0 and we have A = (a) = {a| E} .Such an ideal which is generated by a single element is called a principal ideal.To sum up: Every Euclidean domain E is a principal ideal domain in the sense that every ideal in E is a principal ideal.

c Copyright Reserved IIT Delhi