MTH745U/P (2013–14) Further Topics in Algebra (Fields and Galois Theory) 18th November 2013 Solutions 7 Question 1. Any quadratic extension of F is isomorphic to a quotient of the form F [X]/(aX 2 +bX +c) where a, b, c ∈ F , a 6= 0 and aX 2 +bX +c is irreducible over F (which happens if and only if it has no roots in F ). Note that aX 2 + bX + c will factor into linear factors in either F or F [X]/(aX 2 + bX + c). Let K be a splitting field over F for aX 2 + bX + c (with a 6= 0 and 2 6= 0). Then the roots of F in K are α1 and α2 where √ −b ± b2 − 4ac . α1 , α2 = 2a For i = 1, 2 it is clear√that F (αi ) and F (βi ) are identical subfields of K, where i −b βi := b + 2aαi = ± b2 − 4ac (and so αi = β2a ). But β2 = −β1 , and so √ 2 F (β2 ) = F (β1 ). Therefore K = F ( d) where d = b − 4ac (since K is a splitting field for aX 2 + bX + c), and K has degree 2 over F if and only if d is not a square in F (otherwise K = F has degree 1 over F ). √ The solutions of X 2 + X + c = 0 for the case 2 6= 0 are X = 21 (−1 ± 1 − 4c). In order to prove no such formula can exist in characteristic 2, we can consider a specific case, so we consider the equation X2 + X + 1 = √ 0 over F2 . The roots √ of this lie in F4 but not in F2 . But x ∈ F2 (and in fact x = x) for all x ∈ F2 . So no combination of algebraic operations and taking square roots applied to a collection of elements of F2 can produce an element not in F2 . √ √ Now we let 2 = 0. The two solutions of X 2 + √ X + c = 0 are X√= ∗ c and ∗ c√ + 1, ∗ ∗ ∗ and either of these can serve √ as the√value of c (note that ( c√+ 1) + 1 = c). This is akin to either + d and − d being able to serve as d when √ 2 6= 0. (In some cases, but certainly not universely, we can specify that d denotes a√ particular square root of d, for instance when d ∈ R+ 0 we can specify that 2 d ∈ R+ .) Now we consider the equation aX + bX + c = 0 where 0 p a, b, c ∈ F 2 with a 6= 0. If b = 0 then the only root of aX + bX + c = 0 is X = a/c, which √ has multiplicity 2, and occurs either in F , or in the quadratic extension F ( d) for d = ac . If b 6= 0 then we let Y = ab X (and so X = ab Y ) and observe that 0 = aX 2 + bX + c = b2 2 Y a + 1 b2 Y a +c= b2 (Y 2 a +Y + ac ) b2 p p So Y = ∗ ac , up to the uncertainty of adding 1. Therefore X = ab ∗ ac , up to the b2 b2 b uncertainty of adding a ∈ F . Thus X either lies in F , or X lies in the quadratic √ extension F ( ∗ d) where d = ac . b2 √ √ √ √ ab = a b and a+b = Note that if F is a field of characteristic 2 then √ √ √ √ √ a + b for√all a, b ∈ F . Also ∗ a + b = ∗ a + ∗ b up to the ambiguity inherent when using ∗ s. That is, if one adds a root of X 2 + X + a to a root of X 2 + X + b one obtains a root of X 2 + X + a + b, and either root of X 2 √ +X +a+ b√ can be √ ∗ ∗ obtained in this manner. There is no sensible way to relate ab and a ∗ b. p −1 Question 2. It is easy to see that for p prime ζp has minimal polynomial XX−1 = X p−1 + · · · + X 2 + X + 1, which we proved irreducible on Coursework 4, Question 6. Now ζ9 cubes to ω (a primitive cube root of 1), and so is a root of X 6 +X 3 +1, which is in fact irreducible over Q, which we essentially shall prove here. Our argument is that Q(ζ9 ) contains subfields of degrees 2 (namely Q(ω)) and 3 (namely Q(γ9 ), and proved here) over Q, so that Q(ζ9 ) cannot have degree less than 6 over Q by the Tower Law. We take heed of the hint (omitted from the paper copies of the sheet) that the minimal polynomial of γn over Q is cubic for each n ∈ {7, 9, 13}. Now for n = 7 or 9, we have γn = ζ + ζ −1 , where ζ = ζn . Therefore, we have: 1 = 1, γn = ζ + ζ −1 , γn2 = ζ 2 + ζ −2 + 2, γn3 = ζ 3 + ζ −3 + 3(ζ + ζ −1 ). We now use “symmetrised” forms of the minimal polynomials of ζ7 and ζ9 given above, namely we observe that ζ 3 + ζ 2 + ζ + 1 + ζ −1 + ζ −2 + ζ −3 = 0 if ζ = ζ7 and that ζ 3 + 1 + ζ −3 = 0 if ζ = ζ9 . (Technically, these are Laurent polynomials in ζ rather than ordinary ones.) Thus if ζ = ζ9 we find that ζ 3 + ζ −3 = −1, and so γ93 = ζ 3 + ζ −3 + 3(ζ + ζ −1 ) = −1 + 3(ζ + ζ −1 ) = 3γ9 − 1. Therefore, γ9 is a root of X 3 − 3X + 1. Similarly, we find that γ7 is a root of X 3 + X 2 − 2X − 1. The last case is a bit harder. Now we let ζ = ζ13 . We have: 2 γ13 1 = 1, γ13 = ζ + ζ −1 + ζ 5 + ζ −5 , = ζ 2 + ζ −2 + ζ 10 + ζ −10 + 4 + 2ζ 6 + 2ζ −6 + 2ζ 4 + 2ζ −4 3 You cay try to calculate γ13 if you want to. Instead we use algebraic conjugacy (not yet done in lectures) to help us. For 1 6 j 6 12 the element ζ j is a 13 −1 root of g(X) = XX−1 , which we know is irreducible, and thus Q(ζ) ∼ = Q(ζ j ) ∼ = Q[X]/(g(X)) via an isomorphism φj sending ζ to ζ j and fixing Q. In fact F := Q(ζ) = Q(ζ j ) (why?), and so φj is an automorphism of F . For 1 6 j 6 12, we ∗j ∗1 ∗5 ∗8 ∗12 define γ13 := γ13 φj = ζ j +ζ −j +ζ 5j +ζ −5j . We have γ13 = γ13 = γ13 = γ13 = γ13 , together with: ∗2 ∗3 ∗10 ∗11 γ13 = γ13 = γ13 = γ13 = ζ 2 + ζ −2 + ζ 3 + ζ −3 ∗4 ∗6 ∗7 ∗9 and γ13 = γ13 = γ13 = γ13 = ζ 4 + ζ −4 + ζ 6 + ζ −6 . 2 2 ∗2 ∗2 ∗4 ∗2 Calculating the products γ13 (see above), and γ13 γ13 = γ13 + 2γ13 + γ13 = γ13 − 1, and applying various of the maps φj , we get the following relations. ∗4 ∗2 2 ∗2 ∗4 2 ∗2 ∗4 (γ13 )2 = 3 + γ13 − γ13 , (γ13 ) = 3 + γ13 − γ13 , (γ13 ) = 3 + γ13 − γ13 , ∗2 ∗2 ∗4 ∗2 ∗4 ∗4 γ13 γ13 = γ13 − 1, γ13 γ13 = γ13 − 1 and γ13 γ13 = γ13 − 1. ∗2 ∗4 We also note the relations ζ 13 = 1 and γ13 + γ13 + γ13 = −1 (which follows from P 6 i ζ = 0), and we have made some use of them above, such as to note that i=−6 13 2 ∗2 ∗4 ∗4 (γ13 ) = 4 + γ13 + 2γ13 = 3 + γ13 − γ13 . Therefore, we obtain 3 2 ∗4 2 2 2 γ13 = γ13 γ13 = 3γ13 + γ13 γ13 − γ13 = 3γ13 + (γ13 − 1) − γ13 = −γ13 + 4γ13 − 1. Therefore γ13 is a root of X 3 + X 2 − 4X + 1. In fact, the minimal polynomials of γ7 , γ9 and γ13 are X 3 + X 2 − 2X − 1, X 3 − 3X + 1 and X 3 + X 2 − 4X + 1 respectively. In all cases, my claimed minimal polynomial is in Z[X] and has constant term 1 or −1. Therefore, since all the cases are cubic, we only have to test (essentially by Gauß’s Lemma) for 1 and −1 being potential roots to prove their irreducibility over Q. Naturally, none of these polynomials has 1 or −1 as a root. ∗j Note that since γ13 = γ13 φj for all j with 1 6 j 6 12, the minimal polynomial ∗j ∗2 of γ13 is the same as that of γ13 . Thus X 3 + X 2 − 4X + 1 has roots γ13 , γ13 and ∗4 γ13 . Similar considerations apply to the other two cases. Question 3. If aX 2 + bX + c2 is an irreducible quadratic (over Q), one of whose roots is x0 , then the other is − ab + x0 , so that the two roots sum to − ab . So the splitting field of an irreducible quadratic has degree at most 2 over the ground field. We showed in lectures that the splitting field of an irreducible cubic has degree 3 or 6 over the ground field. Note that if aX 3 + bX 2 + cX + d (a 6= 0) is an irreducible polynomial in F [X] (F a field) having roots x0 , x1 , x2 in a splitting field then x0 + x1 + x2 = − ab , and so the splitting field is generated over F by any two of x0 , x1 , x2 . Not all the polynomials in the question were irreducible, though most were. Irreducibility of claimed irreducible factors by one or more of: Coursework 4, Question 6; Eisenstein’s Irreducibility Criterion; Gauß’s Lemma and the fact that reducible polynomials of degrees 2 or 3 must have a linear factor, and thus monic reducible polynomials in Z[X] of degrees 2 and 3 have an integer root dividing the constant term. √ √ (a) X 2 −2 has splitting field Q( 2); X 2 +7 has splitting field Q( 2); X 2 +X √ +1 and X 3 −1 = (X −1)(X 2 +X +1) both have splitting field Q(ω) = Q( √ −3); 3 2 2 and X − 2X + 1 = (X − 1)(X + X − 1) has splitting field Q( 5). So all the splitting fields have degree 2 over Q. 3 (b) X 4 + X 3 + X 2 + X + 1 has roots ζ5 , ζ52 , ζ53 , ζ54 and splitting field Q(ζ5 ), of degree 4 over Q. X 6 +X 5 +X 4 +X 3 +X 2 +X +1 has roots ζ7 , ζ72 , ζ73 , ζ74 , ζ75 , ζ76 and splitting field Q(ζ7 ), of degree 6 over Q. (c) The splitting fields of each of the polynomials f7 , f9 and f13 (square discriminants 49, 81, 169 respectively) each have degree 3 over Q, and in order to show this it suffices to show we can write all roots of the fi as Q-polynomials in one of the roots. We can readily verify that the roots of f7 are ζ7 +ζ7−1 = γ7 , ζ72 +ζ7−2 = γ72 −2 and ζ74 + ζ7−4 = ζ73 + ζ7−3 = −γ72 − γ7 + 2. In the case of f9 , the roots are ζ9 + ζ9−1 = γ9 , ζ92 + ζ9−2 = γ92 − 2 and ζ94 + ζ9−4 = −γ92 − γ9 + 1. (Here, ζ94 + ζ9−4 6= ζ93 + ζ9−3 = −1.) ∗2 ∗4 In the case of f13 , the roots are γ13 , γ13 and γ13 (noted in the the previous question). In fact, from the relations given in the previous question, we ∗4 2 ∗2 2 easily see that γ13 = γ13 + γ13 − 3, and so γ13 = −γ13 − 2γ13 + 2. (d) In this case, the splitting fields of the cubic polynomials contain a quadratic subfield, and thus have degree 6 (not 3) over Q. Also, note that if a and b are coprime (that is gcd(a, b) = 1) and if F is generated by subfields having degrees a and b over Q, then F has degree ab over Q. The cubic polynomials have non-square discriminants −108 = 62 (−3), −31 and 148 = 22 · 37 respectively (see the hint). So the splitting fields√all have degree √ 6 in these cases. The splitting field of X 3 − 4X + 2 is Q(α, 148) = Q(α, 37) where α is any root of X 3 − 4X + 2. This splitting field is a subfield of R, unlike √ the splitting fields of X 3 − 2 and X 3 + X +√ 1. The roots of X 7 − 3 are ζ7i 7 3 for 0 6 i 6 6, √ so its splitting field is Q(ζ7 , 7 3). It is generated by subfields Q(ζ7 ) √and Q( 7 3) having degrees (respectively) 6 and 7 over Q, and so Q(ζ7 , 7 3) has degree 42 over Q. So the splitting fields have respectively degrees 6, 42, 6 and 6 over Q. Question 4. For these particular primes we can proceed by trial and error, and even calculate cubes of all elements in Fp if necessary. For p = 31 we find that there are three elements cubing to 2, namely 4, 7 and 20 = −11. So over F31 , we find that X 3 − 2 = (X − 4)(X − 7)(X + 11) has all roots in F31 and thus splitting field F31 . (The roots have the form 1 · 4, 5 · 4 and 25 · 4, where 1, 5, 25 are the cube roots of 1 in F31 .) Now for the case p = 7 we find that the cubes in Fp of 0, 1, 2, 3, 4, 5, 6 are 0, 1, 1, 6, 1, 6, 6 respectively, none of which is 2. So X 3 − 2 has no root in F7 , and is therefore irreducible over F7 (as it has no linear factor and degree at most 4 3). Let α(6= 0) be a root of X 3 − 2 in F7 [X]/(X 3 − 2) ∼ = F343 , a field of degree 3 3 over F7 . Then the other two roots of X − 2 are 2α and 4α (note that 2 and 4 are primitive cube roots of 1 in F7 ). Therefore F7 (α) ∼ = F343 is a splitting field for X 3 − 2 over F7 . Now we come to the case p = 11, and cubing the elements 0, . . . , 10 in F11 gives us 0, 1, 8, 5, 9, 4, 7, 2, 6, 3, 10. So F11 has a unique element (7) that cubes to 2, and we can show that this is not a repeated root, for if it were, we would have X 3 − 2 = (X − 7)3 , which is not the case. (There is also only one cube root of 1 in F11 .) So over F11 we have X 3 − 2 = (X + 4)(X 2 − 4X + 5), where X 2 − 4X + 5 2 is irreducible over F11 as its discrimant is (−4)2 − 4 · 5 = −4 = √ 2 (−1) which 3 is a non-square in F11 . So the roots of X − 2, namely 7, 2 ± −1, lie in and generate an extension of degree 2 over F11 . This extension is the splitting √ field, and is F121 . (Note that elements of Fp2 can be written in the form a + b −1 for a, b ∈ Fp if and only if p ≡ 3 (mod 4). You should √ consider what Question 1 says about writing elements of Fp2 in the form a + b d for a, b, d ∈ Fp .) Let us consider what happens for general p. So let fp = f = X 3 − 2 ∈ Fp [X]. We have Df = 3X 2 , and so if the characteristic is not 2 or 3 we have −2 = f − 13 (Df ), and so 1 = − 12 f + 16 (Df ), and thus f does not have a repeated roots except possibly when p = 2 or 3 (since f and Df have a g.c.d. of 1). In fact, for p = 2 and 3, there is a single root with multiplicity 3, for we have f (X) = f2 (X) = X 3 (case p = 2) and f (X) = f3 (X) = X 3 − 2 = X 3 + 1 = (X 3 + 1)3 (case p = 3). (It is no coincidence that 2 and 3 are the only prime factors of −108, which is the discriminant X 3 − 2.) For general p, and working over a splitting field for X 3 − 2, we find (much like the situation over C) that the roots are α, ωα and ω 2 α, where α3 = 2 and ω is a primitive cube root of 1 (take ω = 1 in characteristic 3). Now ω is a root of X 2 + X + 1, and if 3 6= 0 then X 2 + X + 1 does not have 1 as a root, and so (away ∼ from the case 3 = 0) ω lives in Fp if and only if (F× p , ·) = Cp−1 has an element of order 3. This happens if and only if p ≡ 1 (mod 3). Alternatively, for p 6= 2 or 3, we can√argue that ω lies in Fp if and only if −3 is a square in √ Fp , since 1 1 ω = 2 (−1 + −3) (which we cannot really distinguish from 2 (−1 − −3)). If p ≡ 1 (mod 3) then (since ω ∈ Fp ) all of the roots α, ωα, ω 2 α of X 3 − 2 lie in Fp , or none do. In the former case (such as p = 31), X 3 − 2 factors into linear factors over Fp , and so the splitting field is Fp . In the latter case (such as p = 7), the polynomial X 3 − 2 has no root in Fp , and so is irreducible over Fp . The splitting field of X 3 − 2 is then Fp (α, ωα, ω 2 α) = Fp (α) = Fp3 . In this case, α ∈ Fp if and only if 2 is a cube in (F× p , ·), which happens if and only if (p−1)/3 2 ≡ 1 (mod p). If p ≡ 2 (mod 3), we argue that the cubing map on Fp is injective, and hence surjective since Fp is finite. For if x3 = y 3 (with x, y ∈ Fp ), then 0 = x3 − y 3 = 5 (x − y)(x2 + xy + y 2 ) = (x − y)(x − ωy)(x − ω 2 y) [now working in Fp (ω)], thus showing that x = y, as x−ωy = 0 implies that x = y = 0 and x−ω 2 y = 0 implies ∼ that x = y = 0. Alternatively, we note that (F× p , ·) = Cp−1 has order coprime to 3. From the Euclidean Algorithm in Z, we find that 1 = (p − 1) − 3( p−2 )= 3 2p−1 2p−1 −2(p − 1) + ( 3 )3. We let m := 3 , and observe that m > 0 is an integer, and that xm cubes to x, and so is the unique cube root of x in Fp , as we have now shown that the cubing map is surjective. (If x = 0 we observe directly that xm = 0m = 0 cubes to x, while if x 6= 0, we get 3m = 2p − 1 = 2(p − 1) + 1 and so (xm )3 = x3m = (x(p−1) )2 x = x.) So we may pick α = 2m to be the unique cube root of 2 in Fp . If α 6= 0 (that is p 6= 2), then the splitting field of X 3 − 2 is Fp (α, ωα, ω 2 α) = Fp (ω) = Fp2 , while if α = 0 (that is p = 2), then the splitting field of X 3 − 2 is Fp (α, ωα, ω 2 α) = Fp . To summarise, the full results are as follows. • If p = 2 or 3 then X 3 − 2 has a single root with multiplicity 3 in Fp , and so the splitting field is Fp . • If 2 6= p ≡ 2 (mod 3) then X 3 − 2 has 3 distinct roots in its splitting field, which is always Fp2 . Precisely one of these roots lies in Fp • If p ≡ 1 (mod 3) then X 3 − 2 has splitting field Fp if 2(p−1)/3 ≡ 1 (mod p), and splitting field Fp3 if 2(p−1)/3 6≡ 1 (mod p). In this case, all roots of X 3 − 2 generate the same field when appended to Fp . (Note that if p is a prime such that p ≡ 1 (mod 3) then it is reasonably easy to show that there are unique integers a, b > 0 such that p = a2 + 3b2 . It is much harder to show that those primes p ≡ 1 (mod 3) for which 2 is a cubic residue modulo p are exactly those that have the form p = a2 + 27b2 for some integers a, b > 0.) √ Question 5. / Q, √ and √ from Coursework 2, Question 3, we √ that √2 ∈ √ We know / Q( 2) and 5∈ /√ Q( 2, 3). Therefore repeated application know that 3 ∈ √ √ of the Tower Law gives |Q( 2, 3, 5) : Q| = 8. Time permitting, we shall prove a general result that will imply that any finite extension of Q is simple. Without √ √ √ that result, and working √ √from√first principles, we shall prove that Q( 2, 3, 5) = Q(α) where α = 2+ 3+ √5, though √ √ there is a lot of lattitude in the choice of α. It is clear that Q(α) 6 Q( 2, √3, √5), √ so we seek to prove this inclusion the other way round by showing that 2, 3, 5 ∈ Q(α). We can calculate that √ √ √ √ √ √ α2 = ( 2 + 3 + 5)2 = 10 + 2 6 + 2 10 + 2 15, 6 a result that does not seem to help much (we still square roots). Even √ three √ √ have 2 the fact that α is in the proper subfield Q( 6, 10, 15) of degree 4 over Q does not seem to help much. √ √ √ 2 However, we have shown that β := 6 + 10 + 15 = α −10 is in Q(α). Now 2 √ √ √ 1 2 we find that β = 31 + 10 6 + 6 10 + 4 15, and so γ := 2 (β 2 − 4β − 31) = √ √ √ √ 2 3 6 + 10 ∈ Q(α). thus√giving 15 ∈ Q(α). So √ γ = 64 + 12 15, √ √ So√we get 6+ √ 10 ∈ Q(α), and so 6, 10 ∈ Q(α) now we get β − 15 = √ √ (as √ we have already established that 3 6 + 10 ∈ Q(α)). From calculating 6α, 10α and √ √ √ √ √ 15α, we may √ observe√that √ 3α + √6α −1 √15α √ = 6 2, and so 2 ∈ Q(α). Thus 1 Q(α) contains 3 = 2 2 6 and 5 = 2 2 10. √ (If you follow through the above, you can obtain 15 as a degree √ 4 polynomial √ in β, and thus a degree 8 polynomial in α. The same applies to 6√and √ 10.√We √ √ √ then obtain 2 as a degree 9 polynomial in α, and 3, 5 and 30 = 2 15 as degree 17 polynomials. But the minimal polynomial of √ α has √ degree√8, and is X 8 − 40X 6 + 352X 4 − 960X 2 + 576, which has 8 roots ± 2 ± 3 ± 5. So in fact all the above square roots can be written as [now unique] polynomials in α of degree at most 7. These polynomials are: √ 1 (α7 − 28α5 − 56α3 + 960α), √2 = 576 1 (−α7 + 37α5 − 244α3 + 360α), √3 = 96 1 (−α7 + 37α5 − 244α3 + 360α), √5 = 96 1 (−α6 + 40α4 − 328α2 + 480), √ 6 = 96 1 (α6 − 36α4 + 216α2 − 256), √10 = 32 1 6 4 2 √15 = 481 (−α7 + 34α5 − 136α3 − 96) and 30 = 192 (α − 40α + 376α − 1440α). Note that I used a computer to help me calculate these polynomials. But I do not 7 recommend √ √ √calculating √ √ by √ hand √ either 1, α, . . . , α as a Q-linear combination of 1, 2, 3, √ 5, 6, 10, 15, 30 or (by7 hand) doing the linear algebra required to write 2, etc, in terms of 1, α, . . . , α .) Dr John N. Bray, 18th November 2013 7