Isometries and the Unitary Group of a Form

Introduction

An operator preserves the form of the layer when it preserves every pairing, $h(Tx,Ty) = h(x,y)$, and the collection of such operators is the isometry group of the form: the unitary group $\mathrm{U}(A,h)$ of the sesquilinear layer. This article defines it, proves its two equivalent descriptions, and identifies the algebraic conditions under which the two descriptions separate.

Two facts organise the article. The condition $h(Tx,Ty) = h(x,y)$ is equivalent, for a nonsingular form and an operator admitting an adjoint, to the single algebraic equation $T^{*}T = 1$ in the adjoint of The Form-Adjoint of an Operator: on a nonsingular module the form is recovered from the adjoint, so the geometric hypothesis becomes an equation in the algebra of operators. And the adjoint operation restricted to the group is inversion, $T^{-1} = T^{*}$: the involution of the operator algebra is the group-theoretic inverse, which is why the form's group is an algebraic group with a prescribed involution and not a metric object.

The group is the stabiliser of the form under congruence (Congruence and the Stabiliser of a Form), its elements are the operators whose adjoint is their inverse, and its Lie algebra is the space of the anti-self-adjoint operators (The Unitary Group of a Form and Its Lie Algebra). The elements of the algebra whose dagger is their inverse act by isometries by left multiplication, and the unitary slice of Hermitian Algebras is the bridge between the group of the form and the group of the algebra. The metric aspects — boundedness, completeness, compactness of the group for a definite form — are Part II and Unitary and Isometric Operators of the Form. Throughout, $(A,*,h)$ is a sesqualgebra with a form over a base $(R,\varsigma)$, $h$ is nonsingular, the operators are the $R$-linear endomorphisms of $A$ admitting an adjoint, and $T^{*}$ denotes that adjoint.

The Isometry Group

Definition. An operator $T$ is an isometry of $h$ when $h(Tx,Ty) = h(x,y)$ for all $x, y \in A$. The isometries form the isometry set $\operatorname{Iso}(A,h)$, and the unitary group $\mathrm{U}(A,h)$ is the group of the invertible isometries.

Proposition. The isometry set is a monoid under composition, and the unitary group is its group of invertibles.

Proof. The identity is an isometry, the composite of two isometries is an isometry, and $\mathrm{U}(A,h)$ is by definition the set of the isometries invertible in the monoid. The group law is the composition, the inverse of an isometry is an isometry because $h(T^{-1}x, T^{-1}y) = h(TT^{-1}x, TT^{-1}y) = h(x,y)$ applying the isometry to the pair $(T^{-1}x, T^{-1}y)$.

Proposition (the equation of an isometry). Let $T$ admit the adjoint $T^{*}$. Then

$$ \boxed{\;T \in \operatorname{Iso}(A,h) \iff T^{*}T = 1\;} . $$

Proof. $h(Tx, Ty) = h(x, T^{*}Ty)$ for every pair by the definition of the adjoint, and $h(x,y) = h(x, T^{*}Ty)$ for every pair iff $h(x, (T^{*}T - 1)y) = 0$ for every $x, y$; the nonsingularity of $h$ makes this equivalent to $T^{*}T = 1$.

Corollary (the finite-dimensional case). Over a field and in finite dimension with $h$ non-degenerate, every isometry is invertible: if $Tx = 0$ then $h(x,y) = h(Tx,Ty) = 0$ for every $y$, and non-degeneracy gives $x = 0$, so $T$ is injective and hence bijective. Then

$$ \boxed{\;\mathrm{U}(A,h) = \{T : T^{*}T = 1\} = \{T : TT^{*} = 1\}\;} . $$

Over a general ring the implication fails — an injective endomorphism of a free module of finite rank need not be surjective — and the group is the set of the invertible isometries, which is the reading of the topological article Unitary and Isometric Operators of the Form in the complete case, where the unilateral shift is isometric without being unitary.

Inversion Is the Adjoint

Proposition. On $\mathrm{U}(A,h)$ the adjoint is the inverse:

$$ T^{-1} = T^{*} \qquad \text{for every } T \in \mathrm{U}(A,h) . $$

Proof. $T^{*}T = 1$ and $T$ is invertible, so $T^{*} = T^{-1}$ multiplying on the right by $T^{-1}$; taking adjoints in $TT^{*} = 1$ gives the same relation in the other order.

Corollary. The adjoint operation is an anti-automorphism of order two of the algebra of operators, $(ST)^{*} = T^{*}S^{*}$, and its restriction to $\mathrm{U}(A,h)$ is the group inverse; on the group it therefore satisfies $(ST)^{*} = T^{*}S^{*}$ and $(T^{*})^{*} = T$, which is the group-theoretic statement that inversion reverses the product.

Proposition (the determinant). Over a field, for $T \in \mathrm{U}(A,h)$ the determinant satisfies

$$ \boxed{\;\det(T)\,\varsigma(\det(T)) = 1\;} . $$

Proof. Taking determinants in $T^{*}T = 1$ and using that the adjoint is $\varsigma$-semilinear, $\det(T^{*}) = \varsigma(\det(T))$; the determinant of the identity is $1$.

At the trivial base involution this reads $\det(T)^{2} = 1$, the classical statement for the orthogonal group; the determinant-one part is the kernel of the determinant map, and it is the group of the layer in the classical theory, The Orthogonal Lie Algebra.

The Operators of the Elements

Definition. An element $u \in A$ is left-unitary when $u^{*}u = 1$ and unitary when $u^{*}u = uu^{*} = 1$; the unitaries are the unitary slice $\mathrm{U}(A)$ of Hermitian Algebras.

Proposition. Let $L_u$ be the left multiplication by $u$. Then

  • $L_u$ is an isometry of $h$ exactly when $u$ is left-unitary;
  • $L_u$ is a unitary operator exactly when $u$ is unitary.

Proof. $h(ux, uy) = h(x, u^{*}uy)$ by the compatibility of Sesqualgebras with a Form, so the left side equals $h(x,y)$ for all $x,y$ iff $h(x, (u^{*}u - 1)y) = 0$ for all $x,y$, which is $u^{*}u = 1$ by nonsingularity. Invertibility of $L_u$ is invertibility of $u$ in $A$, and for a unitary element $u^{-1} = u^{*}$ gives the second clause.

The assignment $u \mapsto L_u$ is an injective $R$-algebra homomorphism on a unital algebra, so the unitary slice embeds in the unitary group of the form, and the elements of the slice act on $A$ by form-preserving transformations. This is the bridge between the algebra and the group: the group of the form contains the group of the algebra, and the two agree for the forms whose isometries are all of the shape $L_u$, which is a property of the algebra and not a theorem of the layer.

Examples

The Trace Form

On $M_n(\mathbb{C})$ with $h(X,Y) = \tau(XY^{*})$ the adjoint is given by the dagger, $T^{*}(X) = T^{\dagger}X$ read in the basis of the matrices, so the isometry equation $T^{*}T = 1$ is $T^{\dagger}T = 1$, the unitary group $\mathrm{U}(n)$: the group of the form is the classical unitary group.

The Orthogonal Case

At the trivial base involution and $* = \mathrm{id}$ the formula reads $T^{\mathrm{t}}T = 1$ and the group is the orthogonal group of the symmetric form, $\operatorname{O}(A,\mathfrak{Q})$, with determinant-one part $\operatorname{SO}(A,\mathfrak{Q})$ of The Orthogonal Lie Algebra.

The Biquaternion Slice

On $\mathbb{B}$ with the dagger of Hermitian Algebras, the unitary slice $U = \{u : u^{\dagger}u = uu^{\dagger} = 1\}$ is the compact real form of the algebra of the biquaternions, and the isometries of the Lorentzian form $\operatorname{Sc}(\bar{\tilde Q}\tilde Q')$ form the indefinite group of the layer; the two are The Unitary Slice and the Compact Real Form with Hermitian Adjoint and The Form on the Biquaternion Algebra as a General Plain Sesquilinear Form.

Summary

  • An isometry of $h$ is an operator with $h(Tx,Ty) = h(x,y)$; the isometries form a monoid and the unitary group $\mathrm{U}(A,h)$ is its group of invertibles.
  • For a nonsingular form and an operator with an adjoint, $T$ is an isometry exactly when $T^{*}T = 1$.
  • Over a field and in finite dimension with $h$ non-degenerate every isometry is invertible, so the equation alone defines the group; over a general ring the injectivity does not give the surjectivity and the two notions separate.
  • On the group the adjoint is the inverse, $T^{-1} = T^{*}$, so the involution of the operators is the group inverse; the determinant obeys $\det(T)\varsigma(\det(T)) = 1$.
  • The left multiplication $L_u$ is an isometry exactly when $u^{*}u = 1$ and is a unitary operator exactly when $u$ is unitary, which embeds the unitary slice in the unitary group of the form.
  • The definitions are polynomial, so the group is algebraic; the boundedness and the compactness of the definite case are Part II.

Summary of Notation

symbol meaning
$\operatorname{Iso}(A,h)$ the isometry monoid of the form
$\mathrm{U}(A,h)$ the unitary group of the form
$T^{*}$ the adjoint of $T$ for $h$
$L_u$ the left multiplication by $u$
$\mathrm{U}(A)$ the unitary slice $\{u : u^{*}u = uu^{*} = 1\}$
$\operatorname{O}(A,\mathfrak{Q})$, $\operatorname{SO}(A,\mathfrak{Q})$ the orthogonal group and its determinant-one part

Further Reading

  • Max-Albert Knus, Quadratic and Hermitian Forms over Rings, Grundlehren der mathematischen Wissenschaften 294 (Springer, 1991), for the unitary group of a form over a ring with an involution.
  • Jean Dieudonné, La géométrie des groupes classiques, Ergebnisse der Mathematik 5 (Springer, 1955), for the classical groups attached to a form and their generation.
  • Larry C. Grove, Classical Groups and Geometric Algebra, Graduate Studies in Mathematics 39 (American Mathematical Society, 2002), for the orthogonal and unitary groups as isometry groups.