In this chapter, we extend the results from the previous four chapters to the neglected case where the base field has characteristic 2. Throughout this chapter, let F be a field with algebraic closure \(F{}^{al }\).

1 Separability

To get warmed up, we give a different notation (symbol) for quaternion algebras that holds in any characteristic and which is convenient for many purposes.

Definition 6.1.1

Let A be a commutative, finite-dimensional algebra over F. We say A is separable if

$$\begin{aligned} A \otimes _F F{}^{al }\simeq F{}^{al }\times \dots \times F{}^{al }; \end{aligned}$$

otherwise, we say A isinseparable.

Example 6.1.2

If \(A \simeq F[x]/(f(x))\) with \(f(x) \in F[x]\), then A is separable if and only if f has distinct roots in \(F{}^{al }\).

6.1.3

If \({{\,\mathrm{char}\,}}F \ne 2\), and K is a quadratic F-algebra, then after completing the square, we see that the following are equivalent:

  1. (i)

    K is separable;

  2. (ii)

    \(K \simeq F[x]/(x^2-a)\) with \(a \ne 0\);

  3. (iii)

    K is reduced (K has no nonzero nilpotent elements);

  4. (iv)

    K is a field or \(K \simeq F \times F\).

6.1.4

If \({{\,\mathrm{char}\,}}F = 2\), then a quadratic F-algebra K is separable if and only if

$$\begin{aligned} K \simeq F[x]/(x^2+x+a) \end{aligned}$$

for some \(a \in F\). A quadratic algebra of the form \(K=F[x]/(x^2+a)\) with \(a \in F\) is inseparable.

Now we introduce the more general notation.

6.1.5

Let K be a separable quadratic F-algebra, and let \(b \in F^\times \). We denote by

$$\begin{aligned} \displaystyle {\biggl (\frac{K,b}{F}\biggr )} :=K \oplus Kj \end{aligned}$$

the F-algebra with basis 1, j as a left K-vector space and with the multiplication rules \(j^2=b\) and \(j\alpha =\overline{\alpha }j\) for \(\alpha \in K\), where \(\overline{\phantom {x}}\) is the standard involution on K (the nontrivial element of \({{\,\mathrm{Gal}\,}}(K\,|\,F)\) if K is a field). We will also write \(({K,b} \mid {F})\) for formatting.

From 6.1.3, if \({{\,\mathrm{char}\,}}F \ne 2\) then writing \(K \simeq F[x]/(x^2-a)\) we see that

$$\begin{aligned} \displaystyle {\biggl (\frac{K,b}{F}\biggr )} \simeq \displaystyle {\biggl (\frac{a,b}{F}\biggr )} \end{aligned}$$

is a quaternion algebra over F. The point is that we cannot complete the square in characteristic 2, so the more general notation gives a characteristic-independent way to define quaternion algebras. In using this symbol, we are breaking the symmetry between the standard generators \(i,j\), but otherwise have not changed anything about the definition.

2 Quaternion algebras

Throughout the rest of this chapter, we suppose that \({{\,\mathrm{char}\,}}F = 2\). (We will occasionally remind the reader of this supposition, but it is meant to hold throughout.)

Definition 6.2.1

An algebra B over F (with \({{\,\mathrm{char}\,}}F=2\)) is aquaternion algebra if there exists an F-basis \(1,i,j,k\) for B such that

$$\begin{aligned} i^2+i=a,\ j^2=b,\ \text { and }\ k=ij=j(i+1) \end{aligned}$$
(6.2.2)

with \(a \in F\) and \(b \in F^\times \).

Just as when \({{\,\mathrm{char}\,}}F \ne 2\), we find that the multiplication table for a quaternion algebra B is determined by the rules (6.2.2), e.g.

$$\begin{aligned} jk=j(ij)=(ij+j)j=bi+b=kj+b. \end{aligned}$$

We denote by \(\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) or \([{a,b} \mid {F})\) the F-algebra with basis \(1,i,j,ij\) subject to the multiplication rules (6.2.2). The algebra \(\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) is not symmetric in ab (explaining the choice of notation), but it is still functorial in the field F.

If we let \(K=F[i] \simeq F[x]/(x^2+x+a)\), then

$$ \displaystyle {\biggl [\frac{a,b}{F}\biggr )} \simeq \displaystyle {\biggl (\frac{K,b}{F}\biggr )} $$

and our notation extends that of Section 6.1.

Example 6.2.3

The ring \({{\,\mathrm{M}\,}}_2(F)\) of \(2\times 2\)-matrices with coefficients in F is again a quaternion algebra over F, via the isomorphism

Lemma 6.2.4

An F-algebra B is a quaternion algebra if and only if there exist F-algebra generators \(i,j\in B\) satisfying

$$\begin{aligned} i^2+i=a,\ j^2 = b, \ \text { and }\ ij=j(i+1). \end{aligned}$$
(6.2.5)

Proof. Proven the same way as Lemma 2.2.5. \(\square \)

6.2.6

Let \(B=[{a,b} \mid {F})\) be a quaternion algebra over F. Then B has a (unique) standard involution \(\overline{\phantom {x}}:B \rightarrow B\) given by

$$\begin{aligned} \alpha =t+xi+yj+zij\mapsto \overline{\alpha }=x+\alpha =(t+x)+xi+yj+zij\end{aligned}$$

since

$$\begin{aligned} \begin{aligned} \alpha \overline{\alpha }&= (t+xi+yj+zij)((t+x)+xi+yj+zij) \\&= t^2+tx+ax^2+by^2+byz+abz^2 \in F. \end{aligned} \end{aligned}$$
(6.2.7)

Consequently, one has a reduced trace and reduced norm on B as in Chapter 3.

We now state a version of Theorem 3.5.1 in characteristic 2; the proof is similar and is left as an exercise.

Theorem 6.2.8

Let B be a division F-algebra with a standard involution that is not the identity. Then either B is a separable quadratic field extension of F or B is a quaternion algebra over F.

Proof. Exercise 6.9. (This theorem is also implied by Theorem 6.4.1.) \(\square \)

3 \(*\) Quadratic forms

We now turn to the theory of quadratic forms over F with \({{\,\mathrm{char}\,}}F=2\). The basic definitions from section 4.2 apply. For further reference, Grove [Grov2002, Chapters 12–14] treats quadratic forms in characteristic 2, and the book by Elman–Karpenko–Merkurjev [EKM2008, Chapters I–II] discusses bilinear forms and quadratic forms in all characteristics.

Let \(Q:V \rightarrow F\) be a quadratic form with \(\dim _F V=n<\infty \) and associated bilinear form T. Then \(T(x,x)=2Q(x)=0\) for all \(x \in V\), so one cannot recover the quadratic form from the symmetric (equivalently, alternating) bilinear form.

6.3.1

We begin with the definition of the discriminant. When n is even, we simply define \({{\,\mathrm{disc}\,}}(Q)=\det (T) \in F/F^{\times 2}\)—this is equivalent to Definition 4.3.3 when \({{\,\mathrm{char}\,}}F \ne 2\), having absorbed the square power of 2.

When n is odd, the symmetric matrix T always has determinant 0 (Exercise 6.8); we need to “divide this by 2”. So instead we work with a generic quadratic form, as follows. Consider the quadratic form

$$\begin{aligned} Q^{univ }(x_1,\dots ,x_n) :=\sum _{1 \le i \le j \le n} a_{ij} x_i x_j = a_{11} x_1^2 + a_{12} x_1x_2 + \dots + a_{nn} x_n^2 \end{aligned}$$
(6.3.2)

over the field \(F^{univ } :=\mathbb Q (a_{ij})_{i,j=1,\dots ,n}\) (now of characteristic zero!) with \(a_{ij}\) transcendental elements. We compute its universal determinant

$$\begin{aligned} \det ([T^{univ }])=\det \begin{pmatrix} 2a_{11} &{} a_{12} &{} \cdots &{} a_{1n} \\ a_{12} &{} 2a_{22} &{} \cdots &{} a_{2n} \\ \vdots &{} \vdots &{} \ddots &{} \vdots \\ a_{1n} &{} a_{2n} &{} \cdots &{} 2a_{nn} \end{pmatrix} \in 2\mathbb Z [a_{ij}]_{i,j} \end{aligned}$$
(6.3.3)

as a polynomial with integer coefficients. We claim all of these coefficients are even: indeed, reducing modulo 2 and computing the determinant over \(\mathbb F _2(a_{ij})_{i,j}\), we recall that the determinant of an alternating matrix of odd size is zero (over any field). Therefore, we may let

$$\begin{aligned} \delta (a_{11},\dots ,a_{nn}) :=\det (T^{univ })/2 \in \mathbb Z [a_{ij}]_{i,j} \end{aligned}$$
(6.3.4)

be the universal (half-)discriminant. We then define

$$\begin{aligned} {{\,\mathrm{disc}\,}}(Q) :=\delta (Q(e_1),T(e_1,e_2),\dots ,Q(e_n)) \in F/F^{\times 2} \end{aligned}$$

by specialization. Repeating the argument in 4.3.2, if \(t_{ij} \in F\) and \(e_i' :=\sum _j t_{ij} e_i\) then

$$\begin{aligned} \delta (Q(e_1'),T(e_1',e_2'),\dots ,Q(e_n')) = \delta (Q(e_1),T(e_1,e_2),\dots ,Q(e_n))\det (t_{ij})^2 \end{aligned}$$

(verified universally!) so \({{\,\mathrm{disc}\,}}(Q)\) is well-defined. Moreover, this definition agrees with Definition 4.3.3 when \({{\,\mathrm{char}\,}}F \ne 2\).

Example 6.3.5

For example, \({{\,\mathrm{disc}\,}}(\langle a \rangle )=a\) for \(a \in F\), and if

$$\begin{aligned} Q(x,y,z)=ax^2+by^2+cz^2+uyz+vxz+wxy \end{aligned}$$

with \(a,b,c,u,v,w \in F\), then

$$\begin{aligned} {{\,\mathrm{disc}\,}}(Q)=4abc+uvw-au^2-bv^2-cw^2 \end{aligned}$$

in all characteristics.

Definition 6.3.6

We say Q is nondegenerate if \({{\,\mathrm{disc}\,}}(Q) \ne 0\).

Next, even though not every quadratic form over F can be diagonalized, so we will also make use of one extra form: for \(a,b \in F\), we write [ab] for the quadratic form \(ax^2+axy+by^2\) on \(F^2\).

Lemma 6.3.7

There exists a basis of V such that

$$\begin{aligned} Q \simeq [a_1,b_1] \boxplus \dots \boxplus [a_m,b_m] \boxplus \langle c_1, \dots , c_r \rangle \end{aligned}$$
(6.3.8)

with \(a_i,b_i,c_j \in F\).

Proof. Exercise 6.11. \(\square \)

We say that a quadratic form Q isnormalized if Q is presented with a basis as in (6.3.8).

Example 6.3.9

The quadratic forms \(Q(x,y,z)=x^2+yz+z^2\) and \(Q(x,y,z)=x^2+y^2+z^2\) are normalized over \(\mathbb F _2\), but the quadratic form \(Q(x,y,z)=xz+yz+z^2\) is not.

Example 6.3.10

For a normalized quadratic form as in (6.3.8),

$$\begin{aligned} {{\,\mathrm{disc}\,}}(Q)&={{\,\mathrm{disc}\,}}([a_1,b_1] \boxplus \cdots \boxplus [a_m,b_m]){{\,\mathrm{disc}\,}}(\langle c_1, \dots , c_r \rangle ) \\&= (a_1 \cdots a_m)^2 {{\,\mathrm{disc}\,}}(\langle c_1,\dots ,c_r \rangle ). \end{aligned}$$

In \(F/F^{\times 2}\), we have

Therefore, Q is nondegenerate if and only if \(a_1\cdots a_m c_1 \cdots c_r \ne 0\) and \(r \le 1\).

Example 6.3.11

Let \(B=\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) be a quaternion algebra. Then \(1,i,j,ij\) is a normalized basis for B, and by (6.2.7),

$$\begin{aligned} {{\,\mathrm{nrd}\,}}\simeq [1,a] \boxplus [b,ab], \end{aligned}$$

so \({{\,\mathrm{disc}\,}}({{\,\mathrm{nrd}\,}})=b^2\) so \({{\,\mathrm{nrd}\,}}\) is nondegenerate.

4 \(*\) Characterizing quaternion algebras

We now consider the characterization of quaternion algebras as those equipped with a nondegenerate standard involution (revisiting Main Theorem 4.4.1, but now with \({{\,\mathrm{char}\,}}F = 2\)).

Theorem 6.4.1

Let B be an F-algebra (with \({{\,\mathrm{char}\,}}F=2\)). Then B has a nondegenerate standard involution if and only if one of the following holds:

  1. (i)

    \(B=F\);

  2. (ii)

    \(B=K\) is a separable quadratic F-algebra; or

  3. (iii)

    B is a quaternion algebra over F.

Proof. If \(B=F\), then the standard involution is the identity, and \({{\,\mathrm{nrd}\,}}\) is nondegenerate on F because the reduced (half-)discriminant of the quadratic form \({{\,\mathrm{nrd}\,}}(x)=x^2\) is 1.

If \(\dim _F B=2\), then \(B=K\) has a unique standard involution (Lemma 3.4.2). By 6.1.4, we see that the involution is nondegenerate if and only if K is separable.

So suppose \(\dim _F B > 2\). Since B has a nondegenerate standard involution, there exists an element \(i \in B\) such that \(T(i,1)={{\,\mathrm{trd}\,}}(i) \ne 0\). We have \(i \not \in F\) since \({{\,\mathrm{trd}\,}}(F)=\{0\}\). Rescaling we may suppose \({{\,\mathrm{trd}\,}}(i)=1\), whence \(i^2=i+a\) for some \(a \in F\), and \({{\,\mathrm{nrd}\,}}|_{F+Fi} = [1,a]\). (We have started the proof of Lemma 6.3.7, and 1, i is part of a normalized basis, in this special case.)

By nondegeneracy, there exists \(j \in \{1,i\}^\perp \) such that \({{\,\mathrm{nrd}\,}}(j)=b \ne 0\). Thus \({{\,\mathrm{trd}\,}}(j)=0\) so \(\overline{j}=j\) and \(j^2=b \in F^\times \). Furthermore,

$$\begin{aligned} 0={{\,\mathrm{trd}\,}}(ij)=ij+j\overline{i}=ij+j(i+1) \end{aligned}$$

so \(ij=j(i+1)\). Therefore \(i,j\) generate an F-subalgebra \(A \simeq [{a,b} \mid {F})\).

The conclusion of the proof follows exactly as in (4.4.3): if \(k \in \{1,i,j,ij\}^\perp \) then \(k(ij)=k(ji)\), a contradiction. \(\square \)

Corollary 6.4.2

Let B be a quaternion algebra over F, and suppose that \(K \subseteq B\) is a commutative separable F-subalgebra. Then \(\dim _F K \le 2\). Moreover, if \(K \ne F\), then the centralizer of \(K^\times \) in \(B^\times \) is again \(K^\times \).

Next, we characterize isomorphism classes of quaternion algebras in characteristic 2 in the language of quadratic forms.

6.4.3

Let B be a quaternion algebra over F. We again define

$$\begin{aligned} B^0 :=\{\alpha \in B : {{\,\mathrm{trd}\,}}(\alpha )=0\}=\{1\}^\perp . \end{aligned}$$
(6.4.4)

But now \(B^0=F \oplus Fj \oplus Fk\) and in this basis

$$\begin{aligned} {{\,\mathrm{nrd}\,}}(x+yj+zij)=x^2+by^2+byz+abz^2 \end{aligned}$$
(6.4.5)

so \({{\,\mathrm{nrd}\,}}|_{B^0} \simeq \langle 1 \rangle \boxplus [b,ab]\). The discriminant is therefore

$$\begin{aligned} {{\,\mathrm{disc}\,}}({{\,\mathrm{nrd}\,}}|_{B^0})=b^2 = 1 \in F^{\times }/F^{\times 2}. \end{aligned}$$
(6.4.6)

Theorem 6.4.7

Let F be a field with \({{\,\mathrm{char}\,}}F=2\). Then the functor \(B \mapsto {{\,\mathrm{nrd}\,}}|_{B^0}\) yields an equivalence of categories between

Quaternion algebras over F,

under F-algebra isomorphisms

and

Ternary quadratic forms over F with discriminant \(1 \in F^{\times }/F^{\times 2}\),

under isometries.

Proof. We argue as in Theorem 5.6.8 but with \({{\,\mathrm{char}\,}}F = 2\). The argument here is easier, because all sign issues go away and there is no orientation to chase: by Exercise 6.12, there is a unique \(\zeta \in {{\,\mathrm{Clf}\,}}^1 Q \smallsetminus F\) such that \(\zeta ^2=1\). The inclusion \(\iota :B^0 \hookrightarrow B\) induces a surjective F-algebra homomorphism \({{\,\mathrm{Clf}\,}}^0({{\,\mathrm{nrd}\,}}|_{B^0}) \rightarrow B\), so by dimensions it is an isomorphism; this gives one natural transformation. In the other direction, the map \(m_\zeta :V \rightarrow B^0\) by \(v \mapsto v\zeta \) is again an isometry by (5.6.10), giving the other.

Here is a second direct proof. By 6.4.3, the quadratic form \({{\,\mathrm{nrd}\,}}|_{B^0}\) has discriminant 1. To show the functor is essentially surjective, let \(Q:V \rightarrow F\) be a ternary quadratic form with discriminant \(1 \in F^{\times }/F^{\times 2}\). Then \(Q \simeq \langle u \rangle \boxplus [b,c]\) for some \(u,b,c \in F\). We have \({{\,\mathrm{disc}\,}}(Q)=ub^2=1 \in F^{\times 2}\) so \(b \in F^\times \) and \(u \in F^{\times 2}\). Rescaling the first variable, we obtain \(Q \sim \langle 1 \rangle \boxplus [b, c]\). Thus by 6.4.3, Q arises up to isometry from the quaternion algebra \(\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) with \(a=cb^{-1}\).

For morphisms, we argue as in the proof of Proposition 5.2.4 but with \({{\,\mathrm{char}\,}}F=2\). In one direction, an F-algebra isomorphism induces an isometry by uniqueness of the standard involution. Conversely, let \(f:B^0 \rightarrow (B')^0\) be an isometry. Let \(B \simeq \displaystyle {\biggl [\frac{a,b}{F}\biggr )}\). Extend f to an F-linear map \(B \rightarrow B'\) by mapping \(i \mapsto b^{-1} f(ij)f(j)\). The map f preserves 1: it maps F to F by Exercise 6.15, since \(F=(B^0)^\perp =((B')^0)^\perp \), and \(1={{\,\mathrm{nrd}\,}}(1)={{\,\mathrm{nrd}\,}}(f(1))=f(1)^2\) so \(f(1)=1\). We have \(f(j)^2={{\,\mathrm{nrd}\,}}(f(j))={{\,\mathrm{nrd}\,}}(j)=b\) and similarly \(f(ij)^2=ab\) since \(j,ij\in B^0\). Thus

$$\begin{aligned} 1&={{\,\mathrm{trd}\,}}(i)=b^{-1}{{\,\mathrm{trd}\,}}((ij)j)=b^{-1}T(ij,j)= \\&=b^{-1}T(f(ij),f(j))={{\,\mathrm{trd}\,}}(b^{-1}f(ij)f(j))={{\,\mathrm{trd}\,}}(f(i)) \end{aligned}$$

and similarly \({{\,\mathrm{nrd}\,}}(f(i))={{\,\mathrm{nrd}\,}}(i)=a\), thus \(f(i)^2+f(i)+a=0\). Finally,

$$\begin{aligned} f(i)f(j)=b^{-1}f(ij)f(j)^2=f(ij) \end{aligned}$$

and

$$\begin{aligned} f(j)f(i)&= b^{-1}f(j)f(ij)f(j) = b^{-1} f(j)(f(j)f(ij)+T(f(j),f(ij))) \\&=f(ij)+f(j)=(f(i)+1)f(j)=\overline{f(i)}f(j) \end{aligned}$$

so f is an isomorphism of F-algebras. Therefore the functor is full and faithful, yielding an equivalence of categories. \(\square \)

Corollary 6.4.8

The maps \(B \mapsto Q={{\,\mathrm{nrd}\,}}|_{B^0} \mapsto C=V(Q)\) yield bijections

that are functorial with respect to F.

Proof. The remaining parts of the bijection follow as in the proof of Corollary 5.2.6. \(\square \)

We now turn to identifying the matrix ring in characteristic 2.

Definition 6.4.9

A quadratic form \(H :V \rightarrow F\) is a hyperbolic plane if \(H \simeq [1,0]\).

Recall that \([1,0]:F^2 \rightarrow F\) is given by the quadratic form \(x^2+xy=x(x+y)\), visibly isometric to the quadratic form xy. Definition 6.4.9 agrees with Definition 5.4.1 after a change of basis.

Lemma 6.4.10

If Q is nondegenerate and isotropic then \(Q \simeq H \boxplus Q'\) with H a hyperbolic plane.

Proof. We repeat the proof of Lemma 5.4.2. \(\square \)

We may again characterize division quaternion algebras by examination of the reduced norm as a quadratic form as in Main Theorem 5.4.4 and Theorem 5.5.3.

Theorem 6.4.11

Let \(B=\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) (with \({{\,\mathrm{char}\,}}F=2\)). Then the following are equivalent:

  1. (i)

    \(B \simeq \displaystyle {\biggl [\frac{1,1}{F}\biggr )} \simeq {{\,\mathrm{M}\,}}_2(F)\);

  2. (ii)

    B is not a division ring;

  3. (iii)

    The quadratic form \({{\,\mathrm{nrd}\,}}\) is isotropic;

  4. (iv)

    The quadratic form \({{\,\mathrm{nrd}\,}}|_{B^0}\) is isotropic;

  5. (v)

    The binary form [1, a] represents b;

  6. (vi)

    \(b \in {{\,\mathrm{Nm}\,}}_{K|F}(K^\times )\) where \(K=F[i]\); and

  7. (vii)

    The conic \(C :=V({{\,\mathrm{nrd}\,}}|_{B^0}) \subset \mathbb P ^2\) has an F-rational point.

Proof. Only condition (v) requires significant modification in the case \({{\,\mathrm{char}\,}}F = 2\); see Exercise 6.13. \(\square \)

Lemma 6.4.12

Let \(K \supset F\) be a quadratic extension of fields. Then K is a splitting field for B if and only if there is an injective F-algebra homomorphism \(K \hookrightarrow B\).

Proof. If \(\iota :K \hookrightarrow B\) and \(K=F(\alpha )\), then \(1 \otimes \alpha - \iota (\alpha ) \otimes 1\) is a zerodivisor in \(B \otimes _F K\), since

$$\begin{aligned} \left( 1 \otimes \alpha - \iota (\alpha ) \otimes 1\right) \left( 1 \otimes \alpha - \iota (\overline{\alpha }) \otimes 1\right) =0, \end{aligned}$$
(6.4.13)

and so \(B \otimes _F K \simeq {{\,\mathrm{M}\,}}_2(K)\) and K is a splitting field.

Conversely, let \(K=F(\alpha )\) and suppose \(B \otimes _F K \simeq {{\,\mathrm{M}\,}}_2(K)\). If \(B \simeq {{\,\mathrm{M}\,}}_2(F)\), we can take the embedding mapping \(\alpha \) to a matrix with the same rational canonical form. So we suppose that \(B=\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) is a division ring. By Theorem 6.4.11(iv) (over K) and 6.4.3, there exist \(x,y,z,u,v,w \in F\) not all zero such that

$$\begin{aligned} (x+u\alpha )^2 + b(y+v\alpha )^2 + b(y+v\alpha )(z+w\alpha ) + ab(z+w\alpha )^2 = 0; \end{aligned}$$
(6.4.14)

expanding and rewriting into the powers of \(\alpha \) gives

$$\begin{aligned} (u^2+bv^2+bvw+abw^2)\alpha ^2 + (vz+wy)b\alpha + (x^2+by^2+byz+abz^2) = 0. \end{aligned}$$
(6.4.15)

Let \(\beta :=x+ yj+zij\) and \(\gamma :=u+ vj+wij\). Then \(\gamma \in B^\times \), since \(\gamma =0\) implies \({{\,\mathrm{nrd}\,}}(\beta )=0\) and yet B is a division ring. Then the equation (6.4.15) can be written

$$\begin{aligned} {{\,\mathrm{nrd}\,}}(\gamma )\alpha ^2 + {{\,\mathrm{trd}\,}}(\beta \gamma )\alpha + {{\,\mathrm{nrd}\,}}(\beta ) = 0; \end{aligned}$$

then a direct calculation shows that the element

$$\begin{aligned} \mu =\beta \gamma ^{-1} = {{\,\mathrm{nrd}\,}}(\gamma )^{-1}\beta \gamma \end{aligned}$$

satisfies the same equation as (6.4.15) in the variable \(\alpha \), so there is an embedding \(K \hookrightarrow B\) defined by \(\alpha \mapsto \mu \). \(\square \)

Exercises

Throughout these exercises, we let F be a field (of any characteristic, unless specified).

  1. 1.

    Recall the primitive element theorem from Galois theory: if \(K \supseteq F\) is a separable field extension of finite degree, then there exists \(\alpha \in K\) such that \(K=F(\alpha )\)—and hence \(K \simeq F[x]/(f(x))\) where \(f(x) \in F[x]\) is the minimal polynomial of \(\alpha \).

    Extend this theorem to algebras as follows. Let B be a separable, commutative, finite-dimensional F-algebra. Show that \(B \simeq F[x]/(f(x))\) for some \(f(x) \in F[x]\).

\(\triangleright \) 2.:

Let B be a quaternion algebra over F and let \(K \subset B\) be a separable quadratic F-algebra. Show that there exists \(b \in F^\times \) such that \(B \simeq \displaystyle {\biggl (\frac{K,b}{F}\biggr )}\) (as in 6.1.5).

  1. 3.

    Let \(F{}^\text{sep }\) be a separable closure of F and let B be a quaternion algebra over F. Show that \(B \otimes _F F{}^\text{sep }\simeq {{\,\mathrm{M}\,}}_2(F{}^\text{sep })\). [More generally, see Exercise 7.24.]

\(\triangleright \) 4.:

Let K be a separable quadratic F-algebra and let \(u,b \in F^\times \). Show that \(\displaystyle {\biggl (\frac{K,b}{F}\biggr )} \simeq \displaystyle {\biggl (\frac{K,ub}{F}\biggr )}\) if and only if \(u \in {{\,\mathrm{nrd}\,}}(K^\times )={{\,\mathrm{Nm}\,}}_{K|F}(K^\times )\).

  1. 5.

    Let B be a quaternion algebra over F, and let \(K_0 \supseteq F\) be a quadratic field. Prove that there exists a separable extension \(K \supseteq F\) linearly disjoint from \(K_0\) over F (i.e., \(K \otimes _F K_0\) is a domain) such that K splits F.

  2. 6.

    Suppose \({{\,\mathrm{char}\,}}F = 2\) and let \(a \in F\) and \(b \in F^\times \).

    1. (a)

      Show that \(\displaystyle {\biggl [\frac{a,b}{F}\biggr )} \simeq \displaystyle {\biggl [\frac{a,ab}{F}\biggr )}\) if \(a \ne 0\).

    2. (b)

      Show that if \(t \in F\) and \(u \in F^\times \), then \(\displaystyle {\biggl [\frac{a,b}{F}\biggr )} \simeq \displaystyle {\biggl [\frac{a+(t+t^2),bu^2}{F}\biggr )}\).

  3. 7.

    Let \({{\,\mathrm{char}\,}}F = 2\) and let \(B=\displaystyle {\biggl [\frac{a,b}{F}\biggr )}\) be a quaternion algebra over F. Compute the left regular representation \(\lambda :B \rightarrow {{\,\mathrm{End}\,}}_K(B) \simeq {{\,\mathrm{M}\,}}_2(K)\) where \(K=F[i]\) as in 2.3.8.

\(\triangleright \) 8.:

Suppose \({{\,\mathrm{char}\,}}F = 2\). Let \(M \in {{\,\mathrm{M}\,}}_n(F)\) be a symmetric matrix with n odd, and suppose that all diagonal entries of M are zero. Show that \(\det M = 0\).

\(\triangleright \) 9.:

Let \({{\,\mathrm{char}\,}}F=2\) and let B be a division F-algebra with a standard involution. Prove that either the standard involution is the identity (and so B is classified by Exercise 3.9), or that the conclusion of Theorem 3.5.1 holds for B: namely, that either \(B=K\) is a separable quadratic field extension of F or that B is a quaternion algebra over F. [Hint: Replace conjugation by i by the map \(\phi (x)=ix+xi\), and show that \(\phi ^2=\phi \). Then diagonalize and proceed as in the case \({{\,\mathrm{char}\,}}F \ne 2\).]

\(\triangleright \) 10.:

Let \({{\,\mathrm{char}\,}}F=2\). Show that the even Clifford algebra \({{\,\mathrm{Clf}\,}}^0 Q\) of a nondegenerate ternary quadratic form \(Q:V \rightarrow F\) is a quaternion algebra over F.

\(\triangleright \) 11.:

Prove Lemma 6.3.7, that every quadratic form over F with \({{\,\mathrm{char}\,}}F=2\) has a normalized basis.

\(\triangleright \) 12.:

Let \({{\,\mathrm{char}\,}}F=2\) and let \(Q:V \rightarrow F\) be a quadratic form over F with discriminant \(d \in F^\times /F^{\times 2}\) and \(\dim _F V=n\) odd. Show that \(Z({{\,\mathrm{Clf}\,}}Q) \simeq F[x]/(x^2-d)\) and that there is a unique \(\zeta \in Z({{\,\mathrm{Clf}\,}}Q) \cap {{\,\mathrm{Clf}\,}}^1 Q\) such that \(\zeta ^2=1\).

\(\triangleright \) 13.:

Prove Theorem 6.4.11.

\(\triangleright \) 14.:

Let \(Q :=Q' \boxplus Q''\) be an orthogonal sum of two anisotropic quadratic forms over F (with F of arbitrary characteristic). Show that Q is isotropic if and only if there exists \(c \in F^\times \) that is represented by both \(Q'\) and \(-Q''\).

\(\triangleright \) 15.:

Let B be a quaternion algebra over F (with F of arbitrary characteristic). Show that \(F=(B^0)^\perp \).

\(\triangleright \) 16.:

Prove Wedderburn’s little theorem in the following special case: a quaternion algebra over a finite field with even cardinality is not a division ring. [Hint: See Exercise 3.16.]