The Two Forms and the Two Algebraic Norms of the Complex Space
Introduction
A form on an algebra is the scalar part of a product. The complex space carries two products, so it carries two forms, and each form read at its two equal arguments is an algebraic norm. The forms are $P(Q_1,Q_2)=Q_1Q_2$, the plain form, and $H(Q_1,Q_2)=Q_1\bar Q_2$, the Hermitian form. The norms are $P(Q)=Q^{2}$ and $H(Q)=Q\bar Q$.
The two are as different as two forms on one space can be. The plain form is symmetric, complex-valued, of realified inertia $(1,1)$, isotropic, with no positive cone and no order; its norm is complex, anisotropic and not real. The Hermitian form is conjugate-symmetric, positive definite, of realified inertia $(2,0)$, multiplicative; its norm is the squared modulus. One function, $H$, is at once a form, a norm and a square of the algebra.
The two products are The Two Products of the Complex Space, and the general convention of a form is Product and Form Conventions; the norm of the definite form, its multiplicativity, the invertibility criterion and the group of invertibles are Complex Norm and Invertibility. The positivity and the cone are The Positivity and the Ray of Squares of the Complex Space, and the Hermitian modules over the algebra are The Hermitian Modules and the Unitary Witt Group over the Complex Algebra.
Interpretation. The elements are $Q_1=q_1+iq'_1$ and $Q_2=q_2+iq'_2$, with real coordinates. A form is written with two arguments and a norm with one, and the brackets $\langle\cdot,\cdot\rangle$ and $\langle\cdot,\cdot\rangle_{*}$ name the two forms, the plain and the Hermitian. A bracket is reserved for a form and is never an element. The letters of the forms are $P$ and $H$.
The Two Forms
Definition. The plain form is $P(Q_1,Q_2)=Q_1Q_2$, the scalar part of the plain product, and the Hermitian form is $H(Q_1,Q_2)=Q_1\bar Q_2$, the scalar part of the sesquilinear product.
Proposition (there is no third form). A form of the space is the scalar part of a product of the space, so the space carries two forms, and any other reading of the two slots is a conjugate of one of them.
Proof. The space carries the two products $Q_1Q_2$ and $Q_1\bar Q_2$ (The Two Products of the Complex Space), so it carries at most their two scalar parts. A reading with the bar in the first slot gives $\bar Q_1Q_2=\overline{Q_1\bar Q_2}$, whose value is the conjugate of $H(Q_1,Q_2)$, and a form and its conjugate form are one object with two signs, not two forms.
| form | product | the form | the algebraic norm |
|---|---|---|---|
| plain | $Q_1Q_2$ | $P(Q_1,Q_2)=Q_1Q_2$ | $P(Q)=Q^{2}$ |
| Hermitian | $Q_1\bar Q_2$ | $H(Q_1,Q_2)=Q_1\bar Q_2$ | $H(Q)=Q\bar Q$ |
The Plain Form
Definition. The plain form is $P(Q_1,Q_2)=Q_1Q_2$, the multiplication of the field.
Proposition. The plain form is $\mathbb{C}$-bilinear, symmetric and non-degenerate.
Proof. Bilinearity is the distributivity of the field in each variable and the associativity of the scalars; symmetry is the commutativity of the multiplication. Non-degeneracy: if $P(Q_1,Q_2)=0$ for all $Q_2$, then $Q_1=0$ by the choice $Q_2=1$, and the field has no zero divisor.
The Gram Matrix
Over $\mathbb{C}$ the space is one-dimensional with the basis $(1)$, and the Gram matrix is the one-by-one matrix with the single entry $P(1,1)=1$.
The Realified Inertia
Proposition. The realified symmetric bilinear form $\operatorname{Re}P$ has the Gram matrix $\mathrm{diag}(1,-1)$ on the real basis $(1,i)$ and the inertia $(1,1)$.
Proof. $\operatorname{Re}P(1,1)=1$, $\operatorname{Re}P(1,i)=0$, $\operatorname{Re}P(i,i)=-1$.
The Null Lines
Proposition. The realified form is isotropic: $\operatorname{Re}P(Q,Q)=0$ for the two non-zero elements $Q=1+i$ and $Q=1-i$.
Proof. $(1\pm i)^{2}=\pm2i$, whose real part is zero.
The null cone of $\operatorname{Re}P$ is therefore the pair of real lines spanned by $1+i$ and $1-i$, that is, the two diagonals of the real plane.
Remark (the two readings of isotropy). The complex-valued form $Q_1Q_2$ has no non-zero isotropic vector, because $Q^{2}=0$ forces $Q=0$ in a field. The realified form does have them. The two statements do not conflict: isotropy is a property of a real form, and the complex form is not a real form. The word isotropic is therefore read on the realification throughout this article.
Remark (no positive cone, no order). The plain form takes complex values and its real part is indefinite, so there is no set of positive elements for it and no order attached to it. The inertia $(1,1)$ says exactly why: the set where $\operatorname{Re}P$ is non-negative is a double cone of two opposite components, not a positive cone, and a double cone with two components defines no order. The Hermitian form is the one that carries positivity, and the only order of the complex field is the order inherited from its real part, which is not compatible with the multiplication, as The Specificities of the Complex Algebra shows.
The Hermitian Form
Definition. The Hermitian form is $H(Q_1,Q_2)=Q_1\bar Q_2$.
Proposition. The Hermitian form is conjugate-symmetric, $\mathbb{C}$-linear in the first slot and conjugate-linear in the second, and positive definite: $H(Q,Q)\ge0$ with equality only at $Q=0$.
Proof. Conjugate symmetry: $\overline{H(Q_2,Q_1)}=\overline{Q_2\bar Q_1}=Q_1\bar Q_2$. The slot rules are the slot rules of the sesquilinear product. Definiteness: $Q\bar Q=q^{2}+(q')^{2}$, a sum of two real squares.
Proposition. The realified form $\operatorname{Re}H$ has the Gram matrix $I_2$ on the real basis $(1,i)$ and the inertia $(2,0)$.
Proof. $\operatorname{Re}H(1,1)=1$, $\operatorname{Re}H(1,i)=0$, $\operatorname{Re}H(i,i)=1$.
Proposition. The Hermitian form is multiplicative: $H(Q_1Q_2)=H(Q_1)H(Q_2)$.
Proof. $H(Q_1Q_2)=Q_1Q_2\overline{Q_1Q_2}=Q_1\bar Q_1Q_2\bar Q_2$, the modulus being multiplicative.
Remark (both Gram matrices over $\mathbb{C}$). Over $\mathbb{C}$ both forms have the Gram matrix $[1]$: the plain form because $P(1,1)=1$ and the Hermitian form because $H(1,1)=1$. The two forms are separated by their symmetry and not by their matrix, and that is exactly the point: over $\mathbb{C}$ a symmetric form and a conjugate-symmetric form can have the same Gram matrix and behave in opposite ways on the realification.
The Two Algebraic Norms
The Plain Norm
Definition. The plain norm is $P(Q)=P(Q,Q)=Q^{2}$.
Proposition. The plain norm is complex-valued, multiplicative and anisotropic: $P(Q_1Q_2)=P(Q_1)P(Q_2)$, and $P(Q)=0$ only at $Q=0$.
Proof. The multiplicativity is $(Q_1Q_2)^{2}=Q_1^{2}Q_2^{2}$; the anisotropy is the absence of a zero divisor.
The plain norm takes every non-zero complex value exactly twice, at $Q$ and at $-Q$, and it is neither real, nor non-negative, nor definite.
The Hermitian Norm
Definition. The Hermitian norm is $H(Q)=H(Q,Q)=Q\bar Q=q^{2}+(q')^{2}$.
Proposition. The Hermitian norm is real, non-negative, definite and multiplicative, and it is the square of the modulus.
Proof. The reality and the value are the coordinates; the definiteness is the proposition of §The Hermitian Form; the multiplicativity is the multiplicativity of $H$ read at the two equal arguments.
The Two Norms Compared
| the property | $P(Q)=Q^{2}$ | $H(Q)=Q\bar Q$ |
|---|---|---|
| the value | complex | real, non-negative |
| definite | no | yes |
| multiplicative | yes | yes |
| the zero set | $\{0\}$ | $\{0\}$ |
| the range | $\mathbb{C}$ | $[0,\infty)$ |
| the level set of $1$ | the two points $\pm1$ | the unit circle |
Remark (the precise sense of the only one). Both norms are multiplicative, and both vanish only at the origin. Only the Hermitian norm is real and definite, and that is the sense in which it is the norm of the algebra: it is the one that measures, the other one multiplying values without measuring anything.
The Two Forms on the Two Lines
The Real Line
Proposition. On the real line the two forms coincide: $P(q_1,q_2)=H(q_1,q_2)=q_1q_2$ for real $q_1,q_2$.
Proof. The conjugate of a real number is itself.
On the real line both forms are positive definite, and both norms are the square of the real coordinate.
The Imaginary Line
Proposition. On the imaginary line the two forms are opposite: $P(iq'_1,iq'_2)=-q'_1q'_2=-H(iq'_1,iq'_2)$.
Proof. $(iq'_1)(iq'_2)=-q'_1q'_2$, and $(iq'_1)\overline{(iq'_2)}=q'_1q'_2$.
Theorem. The restriction of $\operatorname{Re}P$ to the real line is positive definite and its restriction to the imaginary line is negative definite; the restriction of $\operatorname{Re}H$ is positive definite on both. Hence the inertia $(1,1)$ of the plain form and the inertia $(2,0)$ of the Hermitian form.
Proof. The two lines are orthogonal for both realified forms, and the values on a line are those of the propositions above. A form whose restrictions to two orthogonal lines are of signs $(+,-)$ has inertia $(1,1)$, and one whose restrictions are $(+,+)$ has inertia $(2,0)$.
Remark (the count of signs). The whole difference between the two forms is the sign on the imaginary line. The plain form multiplies two imaginary numbers into a negative real, and the Hermitian form conjugates one of them first and so multiplies two into a positive. Every statement of this article is a consequence of that one sign.
The One Function That Is a Form, a Norm and a Square
Remark (the triple identity of $H$). The function $H$ is three things at once. With two arguments it is the Hermitian form; with two equal arguments it is the algebraic norm; and its value at two equal arguments is the square of the algebra, $Q\bar Q$, which is also the Hermitian square of The Classes of Element of the Complex Algebra. No other function of the complex space has the three readings. The plain form is a form and it has a diagonal, but the diagonal is complex and indefinite, so it measures nothing; the modulus $|Q|$ measures, but it is a square root rather than a square.
Proposition (the polarisation of the plain norm). The plain norm polarises in two terms:
$$ P(Q_1,Q_2)=\tfrac12\bigl[P(Q_1+Q_2)-P(Q_1)-P(Q_2)\bigr]. $$
Proof. Expand $(Q_1+Q_2)^{2}=Q_1^{2}+2Q_1Q_2+Q_2^{2}$.
Proposition (the polarisation of the Hermitian norm). The Hermitian norm polarises in four terms:
$$ H(Q_1,Q_2)=\tfrac12\bigl[H(Q_1+Q_2)-H(Q_1)-H(Q_2)\bigr]+\tfrac{i}{2}\bigl[H(Q_1+iQ_2)-H(Q_1)-H(Q_2)\bigr]. $$
Proof. $H(Q_1+Q_2)=H(Q_1)+H(Q_2)+2\operatorname{Re}H(Q_1,Q_2)$ and $H(Q_1+iQ_2)=H(Q_1)+H(Q_2)+2\operatorname{Im}H(Q_1,Q_2)$, so the two brackets are $2\operatorname{Re}H(Q_1,Q_2)$ and $2\operatorname{Im}H(Q_1,Q_2)$.
Remark (why two terms and four). A square polarises in two terms exactly when the form is bilinear, and in four when it is sesquilinear. The plain norm is the square of a bilinear form and the Hermitian norm is the square of a sesquilinear one, and the arity of the polarisation is the arity of the form. This is the same distinction as the one between the two products, read on the diagonal.
Summary
The complex space carries two forms, $P(Q_1,Q_2)=Q_1Q_2$ and $H(Q_1,Q_2)=Q_1\bar Q_2$, one for each of its two products, and no third. The plain form is symmetric, non-degenerate, of realified inertia $(1,1)$, isotropic on the two null lines $q=\pm q'$, with no positive cone and no order; its norm $Q^{2}$ is complex, anisotropic and multiplicative. The Hermitian form is conjugate-symmetric, positive definite, of realified inertia $(2,0)$, multiplicative; its norm $Q\bar Q$ is real, non-negative and definite, and it polarises in four terms where the plain norm polarises in two. On the real line the two forms coincide, on the imaginary line they are opposite, and the whole difference between the two inertias is the sign of the square of an imaginary number. One function, $H$, is at once a form, a norm and a square.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $P(Q_1,Q_2)=Q_1Q_2$ | the plain form; the plain norm is $P(Q)=Q^{2}$ |
| $H(Q_1,Q_2)=Q_1\bar Q_2$ | the Hermitian form; the norm is $H(Q)=Q\bar Q$ |
| $\langle\cdot,\cdot\rangle$ | the bracket of the plain form |
| $\langle\cdot,\cdot\rangle_{*}$ | the bracket of the Hermitian form |
| $\operatorname{Re}P$, $\operatorname{Re}H$ | the realified forms, of inertias $(1,1)$ and $(2,0)$ |
| $1\pm i$ | the two isotropic elements of the realified plain form |
Further Reading
- Max-Albert Knus, Alexander Merkurjev, Markus Rost and Jean-Pierre Tignol, The Book of Involutions, Colloquium Publications 44 (American Mathematical Society, 1998), for symmetric and Hermitian forms over an algebra with involution, their inertias and their classification.
- Tsit-Yuen Lam, Introduction to Quadratic Forms over Fields, Graduate Studies in Mathematics 67 (American Mathematical Society, 2005), for symmetric bilinear forms, isotropy, the Witt decomposition and the inertia of a real form.
- Paul R. Halmos, Finite-Dimensional Vector Spaces (Springer, second edition, 1974), for sesquilinear forms, their Gram matrices and the polarisation identities.
- Richard S. Pierce, Associative Algebras, Graduate Texts in Mathematics 88 (Springer, 1982), for the trace form and the norm form of an algebra and their behaviour under the realification.