§3P. Polynomial Rings

§3P. Polynomial Rings

Arun Ram
Department of Mathematics and Statistics
University of Melbourne
Parkville, VIC 3010 Australia
aram@unimelb.edu.au

Last updates: 08 March 2011

Polynomial rings

Let R be a commutative ring. Then R[x] is a commutative ring.

Proof.
To show:
    1. If fx, gx∈ Rx then fx+ gx= gx+ fx .
    2. If fx, gx, hx∈ Rx then fx+ gx + hx= fx+ gx+ hx .
    3. There exists 0∈Rx such that if fx∈Rx then 0+fx= fx+0= fx .
    4. If fx∈Rx then there exists -fx∈Rx such that fx+ -fx=0 .
    5. If fx, gx, hx∈ Rx then fxgx = fx gx hx .
    6. There exists 1∈Rx such that 1⋅fx= fx= fx⋅1 for all fx∈Rx .
    7. If fx, gx, hx∈ Rx then fx gx+ hx = fxgx+ fxhx and gx+ hx fx= gxfx+ hxfx .
    8. If fx, gx∈ Rx then fx gx= gx hx .
  1. Let fx, gx∈ Rx such that fx= r0+ r1x+ r2x2 +⋯ and gx= s0+ s1x+ s2x2 +⋯ .
    Then fx+ gx = r0+s0+ r1+s1x+ r2+s2 x2+⋯ and gx+ fx = s0+r0+ s1+r1x+ s2+r2 x2+⋯ Since addition in R is a commutative operation, ri+si= si+ri   for all i. So fx+ gx= gx+ fx .
  2. Let fx, gx, hx∈ Rx and fx= r0+ r1x+ r2x2+⋯ , gx= s0+ s1x+ s2x2+⋯ , and hx= t0+ t1x+ t2x2+⋯ . Then fx+ gx + hx = r0+s0+ t0 + r1+s1+ t1 x+ r2+s2+ t2 x2+⋯ and fx+ gx+ hx = r0+ s0+t0 + r1+ s1+t1 x+ r2+ s2+t2 x2+⋯. Since addition in R is an associative operation, ri+si+ ti= ri+ si+ti  for all i. So fx+gx +hx= fx+ gx+hx .
  3. Let 0 denote the zero polynomial 0= 0+0x+0x2+⋯. Let fx∈Rx and let fx= r0+r1x +r2x2+⋯. Then 0+fx= 0+r0+ 0+r1x+ 0+r2x2 +⋯. Since 0+ri=ri for all i, 0+fx =fx. Since addition of polynomials is commutative by a), 0+fx= fx+0 =fx.
  4. Let fx∈ Rx such that fx= r0+ r1x+ r2x2+⋯ .
    Then let -fx= -r0+ -r1x+ -r2x2+⋯ .
    Since -ri∈R for all i, -fx∈ Rx .
    Then fx+ -fx= r0-r0+ r1-r1x+ r2-r2x2 +⋯. So fx+ -fx= 0+0x+0x2+⋯=0. Since addition of polynomials is commutative by a), fx+ -fx= -fx+fx =0 .
  5. Let fx, gx, hx∈ Rx and let fx= r0+r1x +r2x2+⋯, gx= s0+s1x +s2x2+⋯,
    and hx= t0+t1x +t2x2+⋯.
    Then fxgx =c0+c1x+ c2x2+⋯, where ck= ∑ i+j=k risj, and fx gx hx= d0+d1x+ d2x2+⋯ where dn= ∑ k+l=n cktl. So dn= ∑ k+l=n ∑ i+j=k risj tl= ∑ i+j+l=n risjtl, by the distributive law in R.
    Also gxhx =e0+e1x+ e2x2+⋯, where eq= ∑ a+b=q satb, and fx gx hx = d0′+d1′x + d2′ x2+⋯, where dn′ = ∑ p+q=n rpeq. Then dn′ = ∑ p+q=n rp ∑ a+b=q satb = ∑ p+a+b=n rpsatb, by the distributive law in R.
    So dndn′.
    So fxgx hx= fx gxhx .
  6. Let fx,gx ∈Rx and let fx= r0+ r1x+ r2x2+⋯ and gx= s0+ s1x+ s2x2+⋯.
    Then fxgx =c0+ c1x+ c2x2+⋯, where cm= ∑ i+j=k risj, and gxfx = c0′ + c1′ x+ c2′ x2+⋯, where ck′ = ∑ i+j=k sjri. Since R is a commutative ring, ck= ∑ i+j=k rirj= ∑ i+j=k sjri = ck′ , for all k. So fxgx = gxfx .
  7. Let 1∈Rx be the polynomial given by 1= 1+0x+0x2+⋯. Let fx∈ Rx and fx= r0+ r1x+ r2x2+⋯.
    Then 1⋅fx = c0+ c1x+ c2x2+⋯ where ck= ∑ i+j=k airj, and a1=1  and  ai=0  for all  i>0. So ck= a0rk+ 0+0+⋯+0= rk
    So 1⋅fx= fx.
    Since multiplication in Rx is commutative by h), 1⋅fx= fx⋅1 =fx.
  8. Let fx, gx, hx∈ Rx and suppose fx= r0+ r1x+ r2x2+⋯ ,
    gx= s0+ s1x+ s2x2+⋯ , and hx= t0+ t1x +t2x2+⋯.
    Then fx gx+ hx = c0+ c1x+ c2x2+⋯ where ck= ∑ i+j=k ri sj+tj. Also fxgx +fxhx = c0′ + c1′ x + c2′ x2+⋯ where ck′ = ∑ m+m=k rmsn+ ∑ m+n=k rmtn= ∑ m+n=k rmsn+ rmtn . So ck= ck′ for all k.
    Thus fx gx+ hx = fx gx + fx hx. Sine multiplication in Rx is commutative by h), gx+ hx fx = fx gx+ hx = fx gx + fx hx = gx fx + hx fx.
  9. So Rx is a commutative ring.
□

Let R be an integral domain. Then R[x] is an integral domain.

Proof.
To show: If ax,bx∈ Rx and axbx=0 then either ax=0 or bx=0 .
  1. Let ax= a0+ a1x+ a2x2+⋯ and let bx= b0+ b1x+ b2x2+⋯.
    Let cx= axbx = c0+ c1x+ c2x2+⋯.
    Assume ax≠0.
    Then al≠0 for some l≥0.
    Let k be the smallest k≥0 such that ak≠0.
    To show: bx=0.
    1. To show: bN=0 for all N≥0.
      1. Proof by induction on N.
        Base case: N=0.
        1. We know ck=0 since cx= ax bx=0.
          So ∑ i+j=k aibj=0. Since ai=0 for all i<k, 0= ∑ i+j=k aibj= akb0. Since R is an integral domain and ak,b0∈R and ak≠0, we have b0=0.
        Induction assumption: assume bn=0 for all n<N.
        1. We know ck+N=0 since cx= ax bx=0 .
          So ∑ i+j=k+N ai bj=0. Since ai=0 for all i<k and bn=0 for all n<N, we have 0= ∑ i+j=k+N ai bj =akbN. Since R is an integral domain, and ak,bN∈R and ak≠0, we have bN=0.
        So bN=0 for all N≥0.
      So bx=0.
    So Rx is an integral domain.
□

Let F be a field. The ring F[x] is a Euclidean domain with size function given by deg: F[x] → ℕ f(x) ↦ deg(f(x)).

Proof.
To show: If ax,bx∈ Fx and ax≠0 then there exist qx, rx∈Fx such that bx= axqx+ rx where either rx=0 or degrx< degax.
  1. Assume ax, bx∈ Fx and ax≠0.
    Case 1: bx=0.
    1. Then bx=ax ⋅0+0.
      So qx=0 and rx=0 satisfies the condition.
    Case 2: degbx < degax .
    1. Then, since bx= ax⋅0+ bx and degbx< degax, qx=0 and rx=bx satisfies the condition.
    Case 3: degbx≥ degax .
    1. Let ax= a0+ a1x+ a2x2+⋯+ asxs and bx= b0+ b1x+ b2x2+⋯+ btxt ,
      where as,bt∈ Fx , as≠0, bt≠0.
      Proof by induction on degbx.
      Base case: degbx=0 .
      1. Then degax=0 since degax≤ degbx .
        So bx=b0∈ Fx and ax=a0∈ Fx .
        So bx= b0a0 ⋅ ax+0 .
        So qx= b0 a0-1 and rx=0 satisfies the condition.
      Induction assumption: Assume that if b1x∈ Fx and degb1x <t then there exist q1x, r1x∈ Fx such that b1x= q1x ax+ r1x where either r1x=0 or degr1x <degax .
      Assume that degbx=t .
      1. Let b1x= bx- bt as-1 xt-s ax .
        Then degb1x< degbx . Note that the coefficient of xt in b1x is -bt+bt=0 .
      So degb1x<t =degbx .
      Thus, by the induction assumption, there exist q1x, r1x∈ Fx such that b1x= q1x+ r1x where either r1x=0 or degr1x< degax .
      Then bx = b1x- bt as-1 xs-t ax = q1x ax+ r1x - bt as-1 xs-t ax = q1x- bt as-1 xs-t ax +r1x.
    So, if qx= q1x- bt as-1 xs-t and rx= r1x then bx= qxax +rx and either rx=0 or degrx< degax
So Fx with the size function given by deg is a Euclidean domain. □

Let R,S be commutative rings and ϕ:R↦S be a ring homomorphism. Then the map ψ: R[x] → S[x] r0+r1x+r2x2 +⋯ ↦ φ(r0)+φ(r1)x+φ(r2)x2 +⋯ is a ring homomorphism.

Proof.
To show:
    1. If fx, gx∈ Rx then ψ fx+ gx = ψfx+ ψgx .
    2. If fx, gx∈ Rx then ψ fxgx = ψfx ψgx .
    3. ψ1R= 1S where 1R and 1S are the identities in R and S.
  1. Let fx,gx∈ Rx and let fx= r0+r1x+ r2x2+⋯ and gx= r0′ + r1′ x+ r2′ x2+⋯ .
    Then ψ fx+gx = ψ r0+ r0′ + r1+ r1′ x+ r2+ r2′ x2+⋯ = φ r0+ r0′ +φ r1+ r1′ x+φ r2+ r2′ x2+⋯ Since φ is a homomorphism, φ fx+gx = φr0+ φ r0′ + φr1+φ r1′ x+ φr2+ φ r2′ x2+⋯ = φr0+ φr1x+ φr2x2+⋯ + φ r0′ +φ r1′ x+φ r2′ x2+⋯ = ψ fx + ψ gx .
  2. Let fx,gx ∈Rx and let fx= r0+ r1x+ r2x2+⋯ and gx= r0′ + r1′ x+ r2′ x2+⋯ .
    Then ψ fxgx = ψ c0+ c1x+ c2x2+⋯ , where ck= ∑ i+j=k ri rj′ . So ψ fxgx = φc0+ φc1x+ φc2x2 +⋯.
    Since φ is a homomorphism, φck= φ ∑ i+j=k ri rj′ = ∑ i+j=k φ ri rj′ = ∑ i+j=k φri φ rj′ . So ψ fxgx = d0+d1x+ d2x2+⋯, where dk= ∑ i+j=k φri φ rj′ . So, by the distributive law in S, ψ fxgx = φr0+ φr1x+ φr2x2+⋯ φ r0′ + φ r1′ x+ φ r2′ x2+⋯ = φfx φgx.
  3. Let 1R be the identity in R. ψ1R = ψ 1R+0Rx +0Rx2+⋯ = φ1R+ φ0Rx+ φ0Rx2+⋯ Since φ is a homomorphism, φ1R=1S and φ0R=0S .
    So ψ1R= 1S+0Sx+ 0Sx2+⋯.
So ψ is a homomorphism. □

Let R be a commutative ring and let α∈R. Then the evaluation homomorphism evα:R[x]→R is a ring homomorphism.

Proof.
To show:
    1. If fx,gx∈ Rx then evα fx+ gx = evαfx +evαgx .
    2. If fx, gx∈ Rx then evα fxgx = evα fx evα fx.
    3. evα 1R = 1R where 1R is the identity in R.
  1. Let fx, gx∈ Rx and let fx= r0+r1x+ r2x2+⋯ and gx= s0+s1x+ s2x2+⋯.
    Then evα fx+gx = evα r0+s0+ r1+s1x+ r2+s2x2 +⋯ = r0+s0+ r1+s1α+ r2+s2 α2+⋯. By the distributive law in R, evα fx+gx = r0+s0+ r1α+s1α+ r2α2+ s2α2+⋯ = r0+ r1α+ r2α2+⋯ + s0+ s1α+ r2α2+⋯ = evα fx+ evα gx.
  2. Let fx, gx∈ Rx and let fx= r0+r1x+ r2x2+⋯ and gx= s0+s1x+ s2x2+⋯.
    Then evα fxgx = evα c0+c1x+ c2x2+⋯ , where ck= ∑ i+j=k rirj. So evα fxgx = c0+c1α +c2α2+⋯ .
    Now compute evα fx evα gx. evα fx evα gx= r0+r1α+ r2α2+⋯ s0+s1α+ s2α2+⋯ . By the distributive law in R, evα fx evα gx = r0s0+ r1s0α+ r0s1α+ r0s2α2+ r1s1α2+ r2s0α2+⋯ = r0s0+ r1s0+ r0s1 α+ r0s2+ r1s1+ r2s0 α2+⋯ = c0+ c1α+ c2α2+⋯ where ck= ∑ i+j=k risj. So evα fxgx = evα fx evα gx.
  3. Let 1R be the identity in R and 0R be the zero in R.
    Then evα1R= evα 1R+0Rx+ 0Rx2+⋯ = 1R+0Rα+ 0Rα2+⋯ =1R.
So evα is a ring homomorphism. □

(Gauss' Lemma) Let R be a unique factorization domain. Let f(x),g(x)∈R[x] be primitive polynomials. Then f(x)g(x) is a primitive polynomial.

Proof.
Assume fx= r0+r1x +r2x2+⋯ and gx= s0+s1x+ s2x2+⋯ are primitive polynomials in Rx .
Proof by contradiction.
  1. Assume fxgx is not primitive.
    Then there exists an irreducible element p∈R that divides all the coefficients of fxgx.
    Since fx is primitive there must be at least one coefficient of fx which is not divisible by p.
    Since gx is primitive there must be at least one coefficient of gx which is not divisible by p.
    Let m be the smallest m such that rm is not divisible by p.
    Let n be the smallest n such that sn is not divisible by p.
    Suppose that fxgx =c0+c1x+ c2x2+⋯. Then, since p divides ri for all i<m, and p divides sj for all j<n, cm+n = rmsn+ rm-1sn+1+ rm+1sn-1+ ⋯+ r0sm+n+ rm+ns0 = rmsn+pc, where c is some element of R.
    Since cm+n is divisible by p it follows that rmsn= cm+n- pc is divisible by p.
    1. Suppose that d∈R such that rmsn=pd.
      Let rm= a1⋯ak, sn= b1⋯bl, and d= d1⋯dq, be factorizations of rm,sn and d into irreducible elements a1,…,ak, b1,…,bl, d1,…,dq ∈R.
      Then a1⋯ak b1⋯bl= pd1⋯dq. By the uniqueness of factorizations, either p is an associate to ai  for some 1≤i≤k. or p is an associate to bj  for some 1≤j≤l. So either pi=up or qj=up for some unit u∈R.
      Then, either a= upa1⋯ ai-1 ai+1 ⋯ak , or b= upd1⋯ dj-1 dj+1 ⋯dq .
    Thus, either rm or sn is divisible by ai or dj.
Contradiction.
So fxgx is a primitive polynomial in Rx. □

Let R be a unique factorization domain. Let F be the field of fractions of R and let f(x)∈F[x]. Then

  1. There exists an element c∈F and a primitive polynomial g(x)∈R[x] such that f(x)=cg(x).
  2. The factors c and g(x) are unique up to multiplication by a unit.
  3. f(x) is irreducible in F[x] if and only if g(x) is irreducible in R[x].

Proof.
  1. Let fx= a0b0 + a1b1 x +⋯+ akbk xk ∈Fx .
    Then fx= 1b0b1⋯bk c0+c1x+⋯ ckxk where ci= ai b1⋯bi-1 bi+1⋯bk .
    Let d=gcd c0c1⋯ck .
    Then fx= d b0⋯bk c0′ + c1′ x+⋯+ ck′ xk where ci′ = cid .
    Note that ci′ ∈R since d divides ci.
    Furthermore c0′ + c1′ x+⋯+ ck′ xk =gx is primitive since gcd c0′ , c1′ ,…, ck′ =1 .
    So fx= cgx where c= d b0b1…bk ∈F and gx= c0′ + c1′ x+⋯+ ck′ xk ∈Rx is a primitive polynomial.
  2. Suppose fx= cgx and fx= CGx where c,C∈F and gx,Gx ∈Rx are primitive polynomials.
    Let gx= a0+ a1x+⋯+ akxk and let Gx= b0+ b1x+⋯+ bkxk .
    Suppose c= ab and C= AB where a,b,A,B∈R .
    Since fx= ab gx = AB Gx , aBgx= bAGx. So aBai= bAbi for all 1≤i≤k .
    Since gx is primitive, gcd aBa0, aBa1,…, aBak = aB .
    Since Gx is primitive, gcd bAb0, bAb1,…, bAbk = bA .
    Thus by Proposition 1.6 of §2.T, aB=ubA  for some unit u∈R. So ab = u AB .
    So c=uC where u∈R is a unit.
    So CGx=cgx =uCgx .
    By the cancelation law, Gx= ugx .
    So c=uC and Gx= ugx .
    So c and gx are unique up to multiplication by a unit.
  3. ⇒
    Assume that fx is irreducible in Fx.
    Proof by contradiction.
    1. Assume gx is not irreducible in Rx.
      Then there are g1x and g2x in Rx such that gx= g1x g2x .
      So fx= cgx= cg1x g1x .
      Since Rx⊆ Fx then g1x, g2x∈ Fx .
    Contradiction.
    So gx is irreducible in Rx.
    ⇐
    Assume gx is irreducible in Rx .
    Proof by contradiction.
    1. Assume fx is not irreducible in Fx.
      Then there are f1x and f2x in Fx such that fx= f1x f2x .
      So by a), there exist c1,c2∈F and primitive polynomials g1x, g2x∈ Rx such that f1x= c1g1x and f2x= c2g2x. Let c=c1c2.
      Then fx= c1c2 g1x g2x .
      By Gauss' lemma g1x g2x is a primitive polynomial in Rx. So, by b), gx= ug1x g2x , where u∈R.
      So gx is not irreducible in Rx.
    Contradiction.
    So fx is irreducible in Fx.
□

Let R be a unique factorization domain. Then R[x] is a unique factorization domain.

Proof.
Assume gx∈Rx and let gx= a0+ a1x++⋯+ akxk .
To show:
    1. gx has a unique factorization into irreducible factors in Rx.
    2. The factorization of gx is unique up to multiplication by units in Rx and rearrangement of factors.
  1. By Theorem 1.3 and Theorems 1.1 of §2T and 1.3 of §2T, Fx is a UFD, so gx has a factorization in Fx, gx= f1x f2x⋯ frx, where fix∈ Fx  are irreducible in  Fx. Then, by Proposition 1.7 a), there exist elements c1,…,cr∈ F
    and primitive polynomials g1x,…, grx such that fix= cigix, for each  1≤i≤k. Since the factors fix are irreducible in Fx, it follows from Proposition 1.7 c), that the polynomials gix are irreducible in Rx.
    Since the gix are primitive, by Gauss's lemma, the product ggx⋯ grx is primitive.
    So gx= cg1x g2x⋯ grx, where c=c1c2 ⋯cr∈F .
    We also know that gx= gcda0⋯ak g′x where g′x is a primitive polynomial in Rx.
    Thus, by Proposition 1.7 b), c=ugcda0⋯ak where u∈R is a unit.
    It follows that c∈R.
    Since R is a UFD, c has a factorization c=d1⋯ds, where the elements dj are irreducible elements in R.
    So gx= d1⋯ds⋯ g1x⋯ grx, is a factorization of gx into irreducibles in Rx.
  2. Suppose that gx= d1′ d2′ ⋯ dl′ g1′ ⋯ gm′ , is another factorization of gx into irredicibles in Rx.
    By Proposition 1.7 c), each of the factors gi′ x is irreducible in Fx.
    So gx= d1′ d2′ ⋯ dl′ g1′ ⋯ gm′ , and gx= d1⋯ds g1x⋯ grx are both factorizations of gx in Fx.
    By Theorem 1.3 and Theorems 1.1 of §2T and 1.3 of §2T, Fx is a UFD,
    so r=m and there is a permutation σ such that gσi′ = αigix for some αi which are units in F.
    Proposition 1.7 b), gives that each αi is a unit in R.
    Let u= α1α2⋯αr .
    Then gx = d1⋯ds⋅ g1x⋯ grx = d1′ d2′ ⋯ dl′ g1′ x g2′ x ⋯ gm′ x = u d1′ d2′ ⋯ dl′ g1x g2x⋯ gmx. Theny Proposition 1.7 b) implies there is a unit v∈R such that d1⋯dl= vu d1′ ⋯ dl′ . Since Rx is a UFD, s=l and there is a permutation τ such that dτi =ui di′ , wehere the ui are units in R.
    So there is a rearrangement of the factors di′ and gj′ such that, up to multipication by units in R, they are the same as the factors di and gjx .
    So the factorization of gx in Rx is unique.
So Rx is a UFD. □

Let R be a UFD. For each irreducible element p∈R, let πp:R→R/p be the canonical surjection (Part 1, Ex. 2.1.5). Let π ˆ p :Rx→ Rx/ px be the corresponding homomorphism between polynomial rings (Prop 3.1.6). Let fx∈Rx. Then fx is not primitive if and only if there exists an irreducible element p∈R such that π ˆ p fx=0 .

Proof.
  • ⇒ Assume fx= c0+c1x+⋯ ckxk is not primitive.
    Then there exists p∈R irreducible such that p divides c0, p divides c1,…, p divides ck.
    So c0,c1,…,ck ∈p .
    So πpc0= πpc1=⋯= πpck=0 .
    So π ˆ p fx= πpc0+ πpc1x+⋯+ πpckxk =0 .
  • ⇐ Assume that fx= c0+c1x+⋯+ ckxk and that there exists an irreducible element p∈R such that π ˆ p fx=0 .
    Then π ˆ p fx=0 .
    Then πpc0= πpc1=⋯= πpck=0 .
    So c0,c1,…,ck ∈p .
    So p divides c0, p divides c1,…, p divides ck.
    So fx is not primitive.
□

(Gauss' Lemma) Let R be a UFD. Let fx,gx∈Rx be primitive polynomials. Then fxgx is a primitive polynomial.

Proof.
We shall prove the contrapositive:
If fxgx is not primitve then either fx is not primitive or gx is not primitive.
Assume fxgx is not primitve.
Then by Lemma 1.9, there exists an irreducible elemnet p∈R such that π ˆ p fxgx=0, where π ˆ p :Rx→ Rx/p x is the homomorphism between polynomial rings induced by the canonical surjection πp:R→R/p. Since π ˆ p is a homomorphism, π ˆ p fxgx = π ˆ p fx π ˆ p gx=0. Thus, by Lemma 1.4 of §2T, p is a prime ideal.
Thus by Proposition 1.4 of §1T, R/p is an integral domain.
So either π ˆ p fx=0 or π ˆ p gx=0. Thus, by Lemma 1.9, either fx is not primitive or gx is not primitive. □

References

[CM] H. S. M. Coxeter and W. O. J. Moser, Generators and relations for discrete groups, Fourth edition. Ergebnisse der Mathematik und ihrer Grenzgebiete [Results in Mathematics and Related Areas], 14. Springer-Verlag, Berlin-New York, 1980. MR0562913 (81a:20001)

[GW1] F. Goodman and H. Wenzl, The Temperly-Lieb algebra at roots of unity, Pacific J. Math. 161 (1993), 307-334. MR1242201 (95c:16020)

page history