Hermitian Algebras

Introduction

A Hermitian algebra is an associative algebra carrying an involution and a Hermitian form that the involution turns into the adjoint of multiplication. The data are a product, an involution and a form, and one axiom ties them together:

$$ \langle xy, z \rangle = \langle y, x^{\dagger} z \rangle . $$

The axiom says that the involution computes the adjoint of every multiplication operator. It is the algebraic content of the whole layer, and it needs nothing else: no positivity of the form, no completeness of the algebra, no norm and no convergence.

This article is the base of the four operator groups of Sesqualgebras with a degree-2 form: the Hermitian pair $\mathrm{Sd}_x(y) = xyx^{\dagger}$, the one-sided operators, and their signed variants all take the dagger fixed here as their involution. It is the Part I companion of Hilbert Algebras, which adds the two hypotheses this article removes — the positive definiteness of the form and the completion of the algebra. Positivity, the cone, Cauchy–Schwarz, the completion and the modular theory are Hilbert Algebras and the later articles of Part II; the adjoint, the radical, the isometry group and the unitary slice are here.

The base is a commutative ring $F$ with an involution $\sigma$, and the collapse at $\sigma = \mathrm{id}$ returns the bilinear theory of Algebras with a degree-2 form.

The Axioms

Definition

Let $A$ be a unital associative algebra over a commutative ring $F$ carrying an involution $\sigma$, and let $x \mapsto x^{\dagger}$ be an involution of $A$, $\sigma$-semilinear over $F$:

$$ (xy)^{\dagger} = y^{\dagger}x^{\dagger}, \qquad (x^{\dagger})^{\dagger} = x, \qquad 1^{\dagger} = 1, \qquad (ax)^{\dagger} = \sigma(a)x^{\dagger} . $$

A Hermitian form on $A$ is a map $\langle \cdot, \cdot \rangle : A \times A \to F$ that is

  • additive in each argument, with $\langle ax, y \rangle = a\langle x, y\rangle$ and $\langle x, ay \rangle = \sigma(a)\langle x, y\rangle$ for $a \in F$;
  • Hermitian: $\langle x, y \rangle = \sigma(\langle y, x \rangle)$, so that $\mathfrak{Q}(x) = \langle x, x\rangle$, the value at the two equal arguments, lies in the fixed ring $F^\sigma$;
  • non-degenerate: $\langle x, y \rangle = 0$ for every $y$ forces $x = 0$.

A Hermitian algebra is the triple $(A, \bar{\cdot}, \langle \cdot, \cdot \rangle)$ with the form tied to the involution by the adjoint axiom

$$ \langle xy, z \rangle = \langle y, x^{\dagger} z \rangle \qquad \text{for all } x, y, z \in A . $$

Two of the three data are constitutive of the algebra — the product and the involution — and the third is constitutive of the Hermitian structure. The axiom is what forbids an arbitrary pairing: it is the statement that the involution is the adjoint operation of the algebra for the form. A Hilbert algebra is the case in which the form is, in addition, positive definite; that hypothesis is not made here.

The Adjoint Axiom

Proposition. The adjoint axiom is equivalent to

$$ \boxed{\;\langle x^{\dagger}y, z \rangle = \langle y, xz \rangle\;} . $$

Proof. Apply the axiom to the pair $(y^{\dagger}, x^{\dagger})$ and use $y^{\dagger}x^{\dagger} = (xy)^{\dagger}$, obtaining $\langle (xy)^{\dagger}, z\rangle = \langle x^{\dagger}, yz\rangle$; reading the axiom on the pair $(x^{\dagger},y)$ gives the displayed relation. The non-degeneracy of the form is what makes the adjoint unique.

Remark (a third identity is not equivalent). The identity $\langle xy,z\rangle = \langle x, zy^{\dagger}\rangle$ is sometimes listed as a third equivalent form of the axiom. It is not a consequence of it, even for a non-degenerate form: the compatible form $\tau(xGy^{\dagger})$ with $G = \operatorname{diag}(1,-1)$ on $\mathsf{M}_2$ fails it on $307\,008$ of the $531\,441$ triples of matrices with entries in $\{-1,0,1\}$, while the trace form $\tau(xy^{\dagger})$ satisfies it because $(zy^{\dagger})^{\dagger} = yz^{\dagger}$ and the trace is cyclic. The identity is an extra hypothesis, and it is discussed in Sesqualgebras with a Form.

Proposition. Let $L_x : A \to A$ be the left multiplication $L_x(y) = xy$ and let $R_x(y) = yx$ be the right multiplication. Then

$$ \boxed{\;L_{x^{\dagger}} = L_x^{*}\;}, $$

the adjoint being taken with respect to $\langle \cdot, \cdot \rangle$.

Proof. $\langle L_xy, z\rangle = \langle xy, z\rangle = \langle y, x^{\dagger}z\rangle = \langle y, L_{x^{\dagger}}z\rangle$ for all $y, z$, which is the defining relation of the adjoint, and non-degeneracy makes the adjoint unique. The right multiplication is subtler: the identity $R_{x^{\dagger}} = R_x^{*}$ is the third identity $\langle yx,z\rangle = \langle y,zx^{\dagger}\rangle$ refused above, it holds for the trace form $\langle x,y\rangle = \tau(x^{\dagger}y)$ because $\tau$ is cyclic, and in general the adjoint of $R_x$ is the right multiplication by the transpose $x^{t}$, defined by $\langle yx,z\rangle = \langle y,zx^{t}\rangle$ and equal to $x^{\dagger}$ only for the trace form; the computation is in The Form-Adjoint of the Multiplications.

The two identities are the adjoint axiom in operator form, and they are the origin of the operator theory of the layer. They hold for a form of any signature and over any commutative base with an involution.

The Correspondence

Proposition. The form and the involution determine each other through the functional $\ell = \langle \cdot, 1\rangle$:

$$ \boxed{\;\langle x, y \rangle = \langle x^{\dagger}y, 1 \rangle\;} . $$

Proof. Apply the adjoint axiom with $y$ replaced by $x^{\dagger}y$ and $z = 1$, and use $(x^{\dagger}y)^{\dagger} = y^{\dagger}x$. Conversely, a $\sigma$-linear functional $\ell$ on $A$ produces the Hermitian form $h_\ell(x, y) = \ell(y^{\dagger}x)$, and the two constructions are inverse. This is the correspondence of The Sesquilinear Form and the Conjugation, stated here for the involution of the algebra; it uses neither positivity nor completeness.

The Dagger of a Clifford Algebra

The algebras of the category carry their involution intrinsically. With the $\mathbb{Z}/2$-grading $A = A^0 \oplus A^1$ and the grade involution $\alpha$ acting by $\alpha(x) = (-1)^k x$ on the degree-$k$ part, two anti-involutions are defined on every coefficient ring:

  • reversion $x^{r}$, the anti-automorphism fixing each vector;
  • the conjugate $x^{\natural} = \alpha(x^{r})$, which negates each vector.

They agree on the even part and differ on the odd part, their signs on a $k$-blade being $(-1)^{k(k-1)/2}$ and $(-1)^{k(k+1)/2}$. The involution $\sigma$ of the base acts on the coefficients and commutes with both. The dagger is the composite

$$ x^{\dagger} = \sigma(\alpha(x^{r})) = \sigma(x^{\natural}), $$

an anti-automorphism of order two, and the conjugate reversion $x^{r,\sigma} = \sigma(x^{r})$ is the other anti-involution of order two, the two being related by $x^{\dagger} = \alpha(x^{r,\sigma})$. The construction, the table of signs, the Klein four-group on the biquaternions and the trace form $\tau(x^{\dagger}y)$ are Hilbert Algebras; the point taken here is that they are available over every commutative base and use no positivity.

Example. For the biquaternions $\mathbb{B} = \mathbb{C}\otimes_{\mathbb{R}}\mathbb{H} \cong M_2(\mathbb{C})$ the conjugations ${}^{\natural}$, $\sigma$ and $\bar{\cdot}$ commute and $\bar{\cdot} = \sigma\circ{}^{\natural}$ is the dagger; the form $\operatorname{Trd}(\tilde Q_1\bar{\tilde Q_2})$ is the Hermitian one, while $\operatorname{Trd}(\tilde Q_1\tilde Q_2)$ is the symmetric form of the ordinary trace. The difference between the two is exactly the dagger, and it is stated in Hilbert Algebras.

The Self-Adjoint and Anti-Hermitian Elements

The involution cuts the algebra in two. An element is self-adjoint when $x^{\dagger} = x$ and anti-self-adjoint when $x^{\dagger} = -x$, and the involution is $F$-semilinear, so the self-adjoint elements form a module over the fixed ring $F^\sigma$ and the anti-self-adjoint elements another. When $2$ is invertible in $F$ every element decomposes uniquely,

$$ x = \tfrac{1}{2}\bigl(x + x^{\dagger}\bigr) + \tfrac{1}{2}\bigl(x - x^{\dagger}\bigr), $$

into a self-adjoint part and an anti-self-adjoint part. The decomposition is the algebraic substitute for the splitting of a complex number into a real part and an imaginary part, and it carries no order.

For every $x$ the products $x^{\dagger}x$ and $xx^{\dagger}$ are self-adjoint, $(x^{\dagger}x)^{\dagger} = x^{\dagger}x$, and the form at the two equal arguments reads $\mathfrak{Q}(x) = \langle x^{\dagger}x, 1\rangle$ by the correspondence. These elements are the Hermitian squares of the algebra, and the set of their finite sums is the algebraic positive cone of Hermitian Squares and the Algebraic Positive Cone. Whether that cone meets its negative only at $0$ is the question of the positivity of the involution, and it is the question that decides whether the algebra is a Hilbert Algebra; it is Part II's.

The Radical and Non-Degeneracy

Definition. The left radical and the right radical of the form are

$$ \operatorname{rad}_l = \{x : \langle x, y\rangle = 0 \ \text{for all } y\}, \qquad \operatorname{rad}_r = \{x : \langle y, x\rangle = 0 \ \text{for all } y\}, $$

and the form is non-degenerate when the common radical vanishes.

Proposition. For a Hermitian form the two radicals coincide, $\operatorname{rad}_l = \operatorname{rad}_r = \operatorname{rad}$, and the common radical is a left ideal and a right ideal. If the form is associative, in the sense that $\langle xy, z\rangle = \langle x, yz\rangle$ for all $x, y, z$, then the radical is a two-sided ideal.

Proof. Since the form is Hermitian, $\langle x, y\rangle = \sigma(\langle y, x\rangle)$ with $\sigma$ an involution, so $\langle x, y\rangle = 0$ is equivalent to $\langle y, x\rangle = 0$ and the two radicals coincide. If $x \in \operatorname{rad}$ and $a \in A$ then $\langle ax, y\rangle = \langle x, a^{\dagger}y\rangle = 0$ for every $y$, so $ax \in \operatorname{rad}$ and the radical is a left ideal; if the form is associative then also $\langle xa, y\rangle = \langle x, ay\rangle = 0$, so it is a right ideal as well.

When the radical is a two-sided ideal the form descends to the quotient $A/\operatorname{rad}$, and the descended form is non-degenerate; this is the algebraic reduction that the definite theory never needs, because a positive definite form has a zero radical. The reduction is the companion, on a single algebra, of the passage from an indefinite space to its definite quotient, and the indefinite case is The Indefinite Case and the Signature.

The Isometry Group and the Unitary Slice

Definition. An $F$-linear map $T : A \to A$ is an isometry of the form when

$$ \langle Tx, Ty \rangle = \langle x, y \rangle \qquad \text{for all } x, y \in A , $$

and the isometries form the isometry group $\operatorname{Iso}(A, \mathfrak{Q})$.

The isometry of a form is the solution set of a system of polynomial equations over $F$, so the group is algebraic: it is defined by an equation and not by a distance. Its determinant-one part is written $\operatorname{SIso}(A, \mathfrak{Q})$, and it is the group of the layer.

Definition. The unitary slice is the set of elements whose dagger is their inverse,

$$ U = \{x \in A : x^{\dagger}x = xx^{\dagger} = 1\}. $$

It is a subgroup of the group of invertibles: it is closed under the product because $(xy)^{\dagger}(xy) = y^{\dagger}x^{\dagger}xy = 1$, it contains $1$, and it is closed under inversion because the two defining relations are exchanged by the dagger. On the slice the dagger is the inverse, $x^{\dagger} = x^{-1}$, so the Hermitian pair and the inner conjugation agree there and only there; this is the identity that the four operator groups below exploit. The slice, its relation to the compact real form and the definite case are The Unitary Slice and the Compact Real Form with Hermitian Adjoint.

Proposition. An element of the unitary slice acts by an isometry of the algebra under both the inner conjugation and the Hermitian pair.

Proof. For $u \in U$ and the inner conjugation, $\langle uxu^{-1}, uyu^{-1}\rangle = \langle x, y\rangle$ is the statement that the form is invariant under the automorphism; for the Hermitian pair $\mathrm{Sd}_u(z) = uzu^{\dagger}$ the same computation with $u^{\dagger} = u^{-1}$ gives the identity. The two actions therefore agree on the slice, which is the reason the slice is the meeting point of the involutive structure and the orthogonal structure.

What Belongs Elsewhere

  • The general theory of $\sigma$-sesquilinear and Hermitian forms over an involution ring — the Gram matrix, congruence, the unitary Witt group and the Wall group — belongs to Hermitian Forms over an Involution Ring and the Unitary Witt Group with Hermitian Adjoint and to Hermitian Forms on a Hermitian Algebra with Hermitian Adjoint.
  • The symmetric and antisymmetric decomposition of a form, its dictionary of Hermitian, anti-Hermitian and alternating kinds, and the collapse at the trivial involution belong to The Sesquilinear Form and the Conjugation.
  • The positivity of the form, the positive cone $\{\sum_i x_i^{\dagger}x_i\}$, Cauchy–Schwarz, the completion and the modular theory are Hilbert Algebras, Self-Adjoint Elements and the Positive Cone and The Modular Structure of a Hermitian Algebra in Part II; nothing of that layer is used here.
  • The operators built from the dagger — the blade form, the Hermitian pair $\mathrm{Sd}_x(y) = xyx^{\dagger}$, the one-sided operators and their signed variants — are the four groups of this category that follow.
  • The Clifford algebra, its grading, its intrinsic anti-involutions and the trace form are Clifford Algebras, The Grading of the Clifford Algebra with Signed Inner Conjugation and Hilbert Algebras.

Summary

  • A Hermitian algebra is an associative algebra with an involution $\bar{\cdot}$ and a non-degenerate Hermitian form tied by the adjoint axiom $\langle xy, z\rangle = \langle y, x^{\dagger}z\rangle$.
  • The axiom is $\langle xy, z\rangle = \langle y, x^{\dagger}z\rangle$, equivalently $\langle x^{\dagger}y, z\rangle = \langle y, xz\rangle$; it says that $L_{x^{\dagger}} = L_x^{*}$. The third identity $\langle xy, z\rangle = \langle x, zy^{\dagger}\rangle$, equivalently $R_{x^{\dagger}} = R_x^{*}$, is not a consequence: it is the cyclicity of the functional of the form, and it holds for the trace form.
  • The form and the involution determine each other through $\ell = \langle \cdot, 1\rangle$, with $\langle x, y\rangle = \langle x^{\dagger}y, 1\rangle$.
  • The dagger of a Clifford algebra is the composite $x^{\dagger} = \sigma(\alpha(x^{r}))$ of reversion and the involution of the base, available with no positivity.
  • Every element splits as $x = \tfrac12(x + x^{\dagger}) + \tfrac12(x - x^{\dagger})$ into a self-adjoint and an anti-self-adjoint part, and the Hermitian squares $x^{\dagger}x$ generate the algebraic positive cone.
  • The left and the right radicals are a left and a right ideal, and they coincide and are two-sided when the form is associative; the form descends to a non-degenerate form on the quotient.
  • The isometry group is algebraic, and the unitary slice $U = \{x : x^{\dagger}x = xx^{\dagger} = 1\}$ is the subgroup on which the dagger is the inverse and the Hermitian pair is the inner conjugation.
  • Positivity, the cone, the completion and the modular theory are Hilbert Algebras and Part II; the operator groups of this category are the four that follow.

Summary of Notation

symbol meaning
$A$ a unital associative algebra
$F$, $\sigma$ the commutative base ring and its involution
$x^{\dagger}$ the involution of $A$, $\sigma$-semilinear over $F$
$x^{r}$, $x^{\natural}$ reversion and the conjugate $\alpha(x^{r})$
$x^{r,\sigma}$ the conjugate reversion $\sigma(x^{r})$
$\alpha$ the grade involution of the $\mathbb{Z}/2$-grading
$\langle \cdot, \cdot \rangle$ the Hermitian form of the algebra
$\mathfrak{Q}(x)$ the value at the two equal arguments $\langle x, x\rangle$, valued in $F^\sigma$
$\ell$ the functional $\langle \cdot, 1\rangle$ of the correspondence
$L_x$, $R_x$ the left and the right multiplication by $x$
$T^{*}$ the adjoint of $T$ for the form
$\operatorname{rad}_l$, $\operatorname{rad}_r$ the left and the right radical
$\operatorname{Iso}(A, \mathfrak{Q})$ the isometry group of the form
$U$ the unitary slice $\{x : x^{\dagger}x = xx^{\dagger} = 1\}$
$\mathrm{Sd}_x$ the Hermitian pair $y \mapsto xyx^{\dagger}$
$F^\sigma$ the fixed ring of the base involution

Further Reading

  • Max-Albert Knus, Quadratic and Hermitian Forms over Rings, Grundlehren der mathematischen Wissenschaften 294 (Springer, 1991), for Hermitian forms over a ring with an involution, the adjoint involution and the unitary group.
  • Winfried Scharlau, Quadratic and Hermitian Forms (Springer, 1985), for the non-degenerate case, the radical and the isometry group.
  • Ian R. Porteous, Clifford Algebras and the Classical Groups (Cambridge University Press, 1995), for reversion, Clifford conjugation and the dagger on a Clifford algebra.