The K Norm and the Indefinite Structure

Introduction

This entry reads the eleventh section of the mathematical category: the fourth and last norm of the algebra, the only one that can be negative.

$$ \mathrm{Kr}(\tilde Q)=|Q_t|^{2}-\lVert\vec Q\rVert^{2}, \qquad \vec Q=Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z . $$

The first claim is that this norm measures a different kind of thing from the other three. The Hermitian norm measures size. The quaternionic norm measures invertibility. The plain norm is the complex bilinear partner of the quaternionic one. The K norm measures a sign: it says whether the scalar part of the element dominates its vector part, or is dominated by it, or exactly balances it. It is real-valued, homogeneous of degree two, and it is multiplicative in neither direction.

The second claim is that the sign cuts the algebra into two sectors and a wall. On one side the value is positive and the complex line that the element spans is positive definite. On the other side the value is negative and the line is negative definite. On the wall the value is zero and the line is totally isotropic: the element spans a null line, and the form vanishes on the whole line. The wall is a real cone of real dimension seven with a single point of singularity.

The third claim is that the wall is not the massless set. The zero divisors of the plain product form a cone of real dimension six. Lightlike and massless are therefore different conditions in the full algebra: an element can be lightlike and massive, and it can be massless and not lightlike. The two conditions coincide on the material sector, and only there. That coincidence is the reason the two words are used as synonyms in physics, and the reason the synonym is a theorem about one sector rather than about the algebra.

The fourth claim is that no spectral refinement of this partition exists. The obvious temptation is a plane with the K value on one axis and the square of the vector part on the other, in imitation of the plane that organises the spectral strata of the tenth section. That plane is declined, and the reason is that its two axes are invariants of different groups. The isometry group of the form carries a null element of one spectral type to a null element of the other type, and a similarity of the algebra moves the value of the norm while leaving the spectrum alone. The refinement that does exist pairs the K value with the algebraic norm.

The fifth claim is the reading, and it is stated here so that it can be judged against the mathematics. The K norm is the weight that a state space carries when it keeps a time direction. An inner product that tells a time direction from the spatial ones is necessarily indefinite, and the K norm is the algebra's own indefinite form. The price is stated in the same place: the positive region of an indefinite form is not convex, not pointed and not traced, so it cannot be a probability cone. A Lorentzian state space must carry two forms, one for its inertia and one for its probability, and the algebra supplies both.

The strongest objection is that a norm which can be negative is not a norm at all. The word is used here in the sense of a quadratic form: a real homogeneous function of degree two whose polar form is non-degenerate. The K norm qualifies, and what it lacks is positivity. The objection is right about the word and blind to the content, because the lack of positivity is the whole of what the norm says. A positive norm cannot distinguish a time direction from a spatial one; an indefinite norm can do nothing else.

A second objection is that the entry renames a known object and calls the renaming a result. The Krein form is standard in operator theory, and its indefinite geometry is standard with it. The answer is that what is not standard is the pair of theorems that the algebra supplies and the general theory does not. The first is the coincidence of the two null sets on the material sector, which turns a physical synonym into a sector statement. The second is the transitivity of the isometry group on the null set, which forbids the spectral split that general theory would seem to allow.

The interpretation labels continue the category's sequence. The H Norm: Amplitude and Frame carries I.8.a to I.8.k, The N Norm: Mass, Angle and the Cone carries I.9.a to I.9.l, and the five entries from The Partition: Two Invariants and Four Regimes to The Vacuum and the Nilpotent Ladder carry I.10.a to I.10.bd. This entry carries I.11.a to I.11.l.

The readings are readings. Every physical name below is attached to an algebraic object that is defined without it, and §The Limits states what the readings do not carry. In particular no dynamics, no field equation and no statistics are supplied here, and the words state, weight and light cone name roles inside the algebra.

The Source Table

Mathematical Content Source Mathematical Article
the norm, its homogeneity, its non-multiplicativity, the two sectors, the null set, the two partitions, the six restrictions, and the four reasons against a spectral split Kr-Regularity and Degeneracy in the Biquaternion Algebra
the positive and the negative regions, their parametrisations, their topologies, their products and their inversion The Kr-Regular Sector
the null set, the doubly null Peirce lines, the isotropic units, the Witt index and the products of null elements The Kr-Degenerate Sector
the four forms, their Gram matrices, their inertia and the four algebraic norms The Four Forms and the Four Algebraic Norms over the Biquaternion Algebra
the fundamental symmetry $J={}^{\natural}$ and the bridge to the definite form The Fundamental Symmetry of the Form
the level sets, the complex hyperbolic ball, the real hyperbolic slice and the boundary The Krein Level Sets and the Hyperbolic Structure
the isometry group $\mathrm{U}(1,3)$, its boosts, its Cartan decomposition and its action The Krein Isometry Group and Its Cartan Decomposition
the multiplicative norm, the group of invertibles and the zero-divisor cone N-Regularity and N-Degeneracy in the Biquaternion Algebra
the rank of the one-sided multiplications, the two-sided operator and the polar decomposition Relations between the N Norm and Spectral Regularity and Degeneracy
the four products and their relations, the annihilators and the two-sided operator The Four General Products and Their Relations over the Biquaternion Algebra
the minimal Hermitian idempotents, the Peirce lines and their normal form Hermitian Idempotents and the Peirce Decomposition
the modulus, its branch and its failure at the algebraic null cone The Complete Polar Decomposition of an N-Regular Biquaternion

The first row owns the partition. The other rows supply what the entry uses and does not build: the topology of the two regions, the strata of the null set, the inertia of the form, the bridge to the definite companion, the group that preserves the value, the six restrictions, and the operator theory that follows the algebraic norm instead of the value.

The Results in Brief

The Norm

  • The K norm is $\mathrm{Kr}(\tilde Q)=|Q_t|^{2}-\lVert\vec Q\rVert^{2}$; it is real-valued, homogeneous of degree two under complex scalars, and non-degenerate as a form.
  • It is the real companion of the plain norm $P=Q_t^{2}-B^{2}$, in the sense that the two have the same inertia and the K norm is the plain one read on moduli.
  • It is not multiplicative in either direction: two null elements can have the regular product $2e_t$, and the regular element $e_t+\tfrac12\vec e_x$ has a square of value $-7/16$.
  • Both conjugations preserve it, quaternionic and Hermitian alike; no algebra multiplication preserves it except a scalar of modulus one.
  • Its realified inertia is $(2,6)$, and its inertia as a form on $\mathbb{C}^{4}$ is $(1,3)$: positive on the centre, negative on the vector subspace.

The Two Sectors

  • The positive region is $\{|Q_t|>\lVert\vec Q\rVert\}$, the negative region $\{|Q_t|<\lVert\vec Q\rVert\}$, the null set the equality.
  • The line $\mathbb{C}\tilde Q$ is positive definite on the positive region, negative definite on the negative region, and totally isotropic on the null set; the value of the norm, its sign and the definiteness of the line are one datum.
  • The positive region is $\mathbb{C}^{\times}\times$ the open unit ball of $\mathbb{C}^{3}$, of homotopy type $S^{1}$; the negative region is $\mathbb{R}_{>0}\times D^{2}\times S^{5}$, of homotopy type $S^{5}$; the two are not homeomorphic.
  • The two regions are open and dense, and the null set is their common boundary; both regions are cones with the apex deleted, and neither is convex.
  • The Krein cone $P_K=\{\mathrm{Kr}\ge0\}$ is closed and star-shaped but neither convex nor pointed.

The Null Set

  • $\mathcal K=\{\mathrm{Kr}=0\}=\mathbb{C}^{\times}\cdot(e_t+S^{5})$, of real dimension seven, smooth off the origin, homotopy equivalent to $S^{1}\times S^{5}$.
  • Every non-zero null element has $Q_t\neq0$; the null set meets the group of invertibles and the zero-divisor set, and its own origin is neither.
  • The isometry group $\mathrm{U}(1,3)$ acts transitively on $\mathcal K\setminus\{0\}$, so no continuous invariant survives inside the wall except the value, which is constant there.
  • The null set splits into the doubly null set $\mathcal K_{\mathrm d}=\mathcal K\cap\mathcal N=\mathbb{C}^{\times}\cdot(e_t+iS^{2})$, of real dimension four, and the isotropic units $\mathcal K_{\mathrm u}$, its open dense complement of real dimension seven.
  • The doubly null set off the origin is exactly the union of the Peirce lines $\mathbb{C}(e_t+i\vec u)$, and every point of it is spectrally regular, with $B^{2}=-Q_t^{2}\neq0$.

The Two Partitions

  • Positive implies unit: no positive element is a zero divisor. Every zero divisor has non-positive value, and the equality case is the doubly null set.
  • The converses fail: $e_t+2\vec e_x$ is a negative unit, $e_t+\vec e_x$ is a null unit, $\vec e_x+i\vec e_y$ is a negative zero divisor, $e_t+i\vec e_x$ is a null zero divisor.
  • Inversion preserves the sign and rescales the value by the square of the modulus: $\mathrm{Kr}(\tilde Q^{-1})=\mathrm{Kr}(\tilde Q)/|N(\tilde Q)|^{2}$.
  • The algebraic norm is onto $\mathbb{C}$ on the null set, so it is not a function of the K norm; $e_t$ and $\vec e_x$ have the same algebraic norm and values of both signs.

The Absence of a Spectral Split

  • The K value is not a coefficient of the characteristic polynomial, and it is not invariant under similarity.
  • Every K stratum meets both spectral types, and the doubly null set off the origin is entirely spectrally regular.
  • The refinement that exists pairs the K value with the algebraic norm $N$, not with the discriminant; on the two sectors that pair degenerates, because $\mathrm{Kr}=-N$ on matter and $\mathrm{Kr}=+N$ on information.

The Operators

  • The rank of left multiplication follows the algebraic norm: four on the units, two on the zero divisors off the origin, zero at the origin.
  • The two-sided operator $\mathrm{Sd}_{\tilde Q}(\tilde X)=\tilde Q\tilde X\tilde Q^{*}$ is invertible on the invertibles and of rank one on the null cone off the origin.
  • The polar decomposition exists exactly on the invertibles, and the K norm imposes no further restriction; the K adjoint exists everywhere, because the form is non-degenerate.
  • The level sets of the value are single orbits of $\mathrm{U}(1,3)$: of type $\mathrm{U}(1,3)/\mathrm{U}(3)$ for a positive value and $\mathrm{U}(1,3)/\mathrm{U}(1,2)$ for a negative one, both of real dimension seven.

The Fourth Norm

The K norm is the diagonal of the general quaternionic sesquilinear form.

$$ \langle\tilde P,\tilde Q\rangle_{\natural*}=\sum_\mu\varepsilon_\mu P_\mu\overline{Q_\mu}, \qquad \varepsilon=(1,-1,-1,-1), \qquad \mathrm{Kr}(\tilde Q)=\langle\tilde Q,\tilde Q\rangle_{\natural*}=|Q_t|^{2}-\lVert\vec Q\rVert^{2}. $$

Four properties follow from the formula and from nothing else. The value is real, because the diagonal of a Hermitian form is real. It is homogeneous of degree two under complex scalars, $\mathrm{Kr}(\lambda\tilde Q)=|\lambda|^{2}\mathrm{Kr}(\tilde Q)$, so its zero set is a cone and its non-zero level sets are homogeneous. Its polar form is non-degenerate, with Gram matrix $E=\mathrm{diag}(1,-1,-1,-1)$ in the coefficient basis. Its inertia is $(1,3)$ as a form on $\mathbb{C}^{4}$ and $(2,6)$ on the eight real parameters.

The norm is the plain norm read on moduli. The plain norm is $P(\tilde Q)=Q_t^{2}-B^{2}$ with $B^{2}=Q_x^{2}+Q_y^{2}+Q_z^{2}$, and the quaternionic norm is $N(\tilde Q)=Q_t^{2}+B^{2}$. Replacing each square by the square of the modulus in the first gives the K norm, and doing the same in the second gives the Hermitian norm. So the two real norms are the two complex norms read on moduli, and each real norm keeps the inertia of its complex partner while losing its complex value.

What each of the four decides is the reason there are four.

norm formula value in multiplicative decides
plain, $P$ $Q_t^{2}-B^{2}$ $\mathbb{C}$ no the complex bilinear cone
quaternionic, $N$ $Q_t^{2}+B^{2}$ $\mathbb{C}$ yes invertibility
Hermitian, $H$ $\lvert Q_t\rvert^{2}+\lVert\vec Q\rVert^{2}$ $\mathbb{R}$, positive no size, amplitude, topology
K, $\mathrm{Kr}$ $\lvert Q_t\rvert^{2}-\lVert\vec Q\rVert^{2}$ $\mathbb{R}$, signed no a position on the metric cone

The K norm is not multiplicative, and both directions of failure matter. Two null elements multiply to a regular one, and two regular elements multiply to a null one.

$$ (e_t+\vec e_x)(e_t-\vec e_x)=2e_t, \qquad \mathrm{Kr}=4, $$

$$ \Bigl(e_t+\tfrac12\vec e_x\Bigr)^{2}=\tfrac34e_t+\vec e_x, \qquad \mathrm{Kr}=-\tfrac{7}{16}, \qquad \mathrm{Kr}\Bigl(e_t+\tfrac12\vec e_x\Bigr)=\tfrac34, $$

$$ \vec e_x\bigl(\vec e_x+i\vec e_y\bigr)=-e_t+i\vec e_z, \qquad \mathrm{Kr}=0, \qquad \mathrm{Kr}(\vec e_x)=-1,\quad \mathrm{Kr}(\vec e_x+i\vec e_y)=-2 . $$

The first two lines are the two halves of the failure. A null product of regular factors says that the value is not submultiplicative in the strong sense: it can drop to zero. A regular product of null factors says that the value is not bounded below either: it can rise from zero. Nothing survives except homogeneity.

What does survive is the invariance under the two conjugations. Quaternionic and Hermitian conjugation preserve the moduli of the four coefficients, so

$$ \mathrm{Kr}(\tilde Q^{\natural})=\mathrm{Kr}(\tilde Q)=\mathrm{Kr}(\tilde Q^{*}). $$

No algebra multiplication preserves the value, and this is the structural difference from the other three norms. The Hermitian norm is invariant under both-sided $*$-unitary multiplication. The quaternionic norm is multiplicative, so it is preserved by a multiplication exactly when the factor has norm one. The K norm is preserved by no multiplication except that of a scalar of modulus one: a left factor of norm one such as $\vec e_x$ sends $e_t+2\vec e_x$, of value $-3$, to $\vec e_x-2e_t$, of value $3$. The value is a property of the element in the frame, not a property the element carries through a product.

The word regular in this section is metric and not ring-theoretic. A ring-regular element is a non-zero-divisor. A K-regular element is one of non-zero value. The two readings differ: $e_t+\vec e_x$ is K-degenerate and an invertible element, while $\vec e_x+i\vec e_y$ is ring-degenerate and K-regular. The only implication between them runs one way, and it is the theorem of §The Two Partitions.

The Wall

The value of the norm sorts the algebra into three sets.

$$ \mathbb{B}^{\natural*}_{+}=\{\mathrm{Kr}>0\}, \qquad \mathcal K=\{\mathrm{Kr}=0\}, \qquad \mathbb{B}^{\natural*}_{-}=\{\mathrm{Kr}<0\}, \qquad \mathbb{B}=\mathbb{B}^{\natural*}_{+}\uplus\mathcal K\uplus\mathbb{B}^{\natural*}_{-}. $$

The three sets have five equivalent descriptions each, and the equivalence is the theorem. An element is positive exactly when $|Q_t|>\lVert\vec Q\rVert$, exactly when its complex line is non-degenerate and positive definite, exactly when its real plane $\operatorname{span}_{\mathbb{R}}\{\tilde Q,i\tilde Q\}$ is non-degenerate and positive definite. The same holds with the inequalities reversed, and the null set is the case in which the line is totally isotropic and lies in its own orthogonal complement.

The value, the sign and the definiteness of the line are one datum. On the line, homogeneity reduces the form to one number:

$$ \mathrm{Kr}(\lambda\tilde Q)=|\lambda|^{2}\mathrm{Kr}(\tilde Q), \qquad \lambda\in\mathbb{C}. $$

So the restriction to $\mathbb{C}\tilde Q$ is positive definite, negative definite or identically zero according to the sign of that number. There is no element whose line is definite with a sign other than the sign of its value.

The two non-zero regions are parameterised by a scale and a distortion.

$$ \tilde Q=Q_t\bigl(e_t+W\bigr),\quad Q_t\neq0,\ \lVert W\rVert<1 \qquad\text{(positive)}, $$

$$ \tilde Q=r\bigl(we_t+W\bigr),\quad r>0,\ w\in D^{2},\ \lVert W\rVert=1 \qquad\text{(negative)} . $$

The two parameterisations are the reason the two regions are topologically different. The positive region is a punctured complex plane times an open ball, of homotopy type $S^{1}$. The negative region is a positive half-line times an open disc times a five-sphere, of homotopy type $S^{5}$. The two are both of real dimension eight and they are not homeomorphic, so the wall separates two components that cannot be deformed into one another.

The wall is the common boundary of both. The norm is a real polynomial, so its non-zero set is open and its zero set is closed. Off the origin the gradient in the eight real coordinates does not vanish, so the null set is a smooth hypersurface of real dimension seven and of empty interior. Every point of it is a limit of positive and of negative elements, because the norm has no local extremum there: the origin is its only critical point, and near the origin $e_t$ and $\vec e_x$ give values of both signs.

The closure of the positive region is the Krein cone, and the cone is not what a probability cone is.

$$ P_K=\{\mathrm{Kr}\ge0\}=\{|Q_t|\ge\lVert\vec Q\rVert\}. $$

Three failures of conehood are provable from the formula. The cone is not pointed, because it is invariant under $\tilde Q\mapsto-\tilde Q$, so it contains opposite rays and $P_K\cap(-P_K)=P_K$. The cone is not convex: the null elements $e_t+\vec e_y$ and $-e_t+\vec e_y$ both lie in it, and their midpoint $\vec e_y$ has value $-1$. The cone is star-shaped with respect to the origin, closed, and invariant under positive real scalars, and those three properties are all it has.

The wall is a single orbit, and that is the strongest statement in the section. The isometry group of the form acts transitively on $\mathcal K\setminus\{0\}$, so two non-zero null elements are related by an isometry whatever their coefficients are. Nothing continuous distinguishes one null element from another except the scale, which the action also moves. The wall has one shape and no internal structure.

The Null Set

The null set is a real cone of real dimension seven, smooth off its apex, and it is parameterised by a scale and a direction of the unit five-sphere.

$$ \mathcal K=\{\mathrm{Kr}=0\}=\{|Q_t|=\lVert\vec Q\rVert\} =\bigl\{Q_t(e_t+W)\ :\ Q_t\in\mathbb{C}^{\times},\ W\in S^{5}\bigr\}. $$

Every non-zero null element has a non-zero scalar part. The scalar part is complex and its modulus equals the norm of the vector part, so a vanishing scalar part forces a vanishing vector part, which is the origin. The apex is therefore the only null element with $Q_t=0$, and it is the only singular point of the cone.

The null set is not the zero-divisor set, and the dimensions separate them.

set condition real dimension contains units contains zero divisors
the K null set $\mathrm{Kr}=0$ $7$ yes yes
the algebraic null cone $N=0$ $6$ no yes, off the origin
their intersection $\mathrm{Kr}=0$ and $N=0$ $4$ no yes, off the origin

The two witnesses are the shortest proof that neither set contains the other. The element $e_t+\vec e_x$ is null and an invertible element, with $N=2$. The element $\vec e_x+i\vec e_y$ is a zero divisor and not null, with value $-2$. The intersection is not empty: it is the doubly null set.

The null set splits into two strata, and the split is the trace of the algebraic partition.

$$ \mathcal K_{\mathrm d}=\mathcal K\cap\mathcal N=\mathbb{C}^{\times}\cdot\bigl(e_t+iS^{2}\bigr), \qquad \mathcal K_{\mathrm u}=\mathcal K\setminus\mathcal K_{\mathrm d}. $$

The doubly null stratum is the union of the Peirce lines, and each line carries an idempotent.

$$ \mathcal K_{\mathrm d}\setminus\{0\}=\bigcup_{\vec u\in S^{2}}\mathbb{C}(e_t+i\vec u), \qquad \tilde\Pi_{\vec u}=\tfrac12\bigl(e_t+i\vec u\bigr), \qquad \tilde\Pi_{\vec u}^{2}=\tilde\Pi_{\vec u}. $$

The two strata are separated by the algebraic norm and by nothing else. On $\mathcal K_{\mathrm u}$ the algebraic norm takes every non-zero complex value, so the stratum is a cone of invertibles. On $\mathcal K_{\mathrm d}$ it vanishes identically, so the punctured stratum is a set of zero divisors. The K value is zero on both, which is the precise sense in which the metric partition cannot see the difference.

The doubly null stratum is entirely spectrally regular off the origin, and this is the sharpest single fact about it. On $\mathcal K_{\mathrm d}$ the two equations give $Q_t^{2}+B^{2}=0$ with $Q_t\neq0$, hence

$$ B^{2}=-Q_t^{2}\neq0 . $$

So a spectral subdivision of the doubly null stratum would have nothing to subdivide. Every point of the stratum is spectrally regular, and the only spectrally degenerate point of the stratum is the origin, which lies in it as its apex and in no line.

No non-zero null element is nilpotent, and the two cones meet only at the apex of one of them. A nilpotent has $Q_t=0$, and the null set off the origin has $Q_t\neq0$. The nilpotent cone is a different four-dimensional cone, and it lies entirely in the open negative region of the K norm: for a pure isotropic vector $\vec\Upsilon\neq0$,

$$ \mathrm{Kr}(\vec\Upsilon)=-\lVert\vec\Upsilon\rVert^{2}<0 . $$

The products of null elements do not stay null, and the three possible outcomes all occur.

$$ (e_t+\vec e_x)(e_t-\vec e_x)=2e_t\ \ (\mathrm{Kr}=4), \qquad (e_t+\vec e_x)^{2}=2\vec e_x\ \ (\mathrm{Kr}=-4), $$

$$ (e_t+\vec e_x)(e_t+i\vec e_x)=(1-i)e_t+(1+i)\vec e_x\ \ (\mathrm{Kr}=0). $$

The maximal totally isotropic subspace is a line over $\mathbb{C}$ and a plane over $\mathbb{R}$. The Witt index is one over $\mathbb{C}$, because a totally isotropic subspace meets the negative-definite vector space only at the origin and therefore injects into a one-dimensional quotient. Over $\mathbb{R}$ the index is two, the extra dimension being the realification of the complex line. A real isotropic plane is never a complex line, so it is never a Peirce line.

The Two Partitions

The algebra carries two partitions, and the entry's job is to say how they meet. The metric partition is by the sign of the K norm. The algebraic partition is by invertibility: the group of invertibles, the origin, and the zero-divisor cone. The pair of numbers $(\mathrm{Kr},N)$ records both, and the table gives one witness per cell.

$\mathrm{Kr}$ unit zero divisor
positive $e_t$, pair $(1,1)$ none
negative $e_t+2\vec e_x$, pair $(-3,5)$ $\vec e_x+i\vec e_y$, pair $(-2,0)$
null $e_t+\vec e_x$, pair $(0,2)$ $e_t+i\vec e_x$, pair $(0,0)$

The origin sits outside the table. It has the pair $(0,0)$ and is neither an invertible element nor a zero divisor, so it is the one point that the algebraic partition separates from everything else and the one point at which the metric partition is degenerate without being informative.

The comparison is one-sided, and the one implication is a theorem.

Theorem (positive implies invertible). If $\mathrm{Kr}(\tilde Q)>0$ then $\tilde Q$ has an inverse.

Proof. Positivity gives $|Q_t|>\lVert\vec Q\rVert$, hence $Q_t\neq0$. If $\tilde Q$ were a zero divisor then $Q_t^{2}=-B^{2}$, so

$$ |Q_t|^{2}=\Bigl|\sum_kQ_k^{2}\Bigr|\le\sum_k|Q_k|^{2}=\lVert\vec Q\rVert^{2}, $$

against positivity. A non-zero element that is not a zero divisor is invertible.

The same inequality run in the other direction says where the zero divisors live.

Corollary. Every zero divisor has $\mathrm{Kr}\le0$, and the equality case is the doubly null set.

Proof. The zero divisors are the non-zero elements with $N=0$, which is the inequality just proved. Equality needs $|Q_t|^{2}=\lVert\vec Q\rVert^{2}$ as well, hence both conditions, hence $\mathcal K_{\mathrm d}$.

The three converses fail, and each failure has a witness of one line. A negative unit is $e_t+2\vec e_x$. A negative zero divisor is $\vec e_x+i\vec e_y$. A null unit is $e_t+\vec e_x$. So the group of invertibles meets all three strata, and the zero-divisor set meets two of them.

Inversion preserves the sign and rescales the value.

$$ \mathrm{Kr}\bigl(\tilde Q^{-1}\bigr)=\frac{\mathrm{Kr}(\tilde Q)}{\lvert N(\tilde Q)\rvert^{2}} . $$

Proof. $\tilde Q^{-1}=(Q_t-\vec Q)/N(\tilde Q)$, so the scalar part of the inverse is $Q_t/N$ and its vector part $- \vec Q/N$, and the value is $(|Q_t|^{2}-\lVert\vec Q\rVert^{2})/|N|^{2}$.

The formula says that the sign is a property of the element and the value is a property of the element and of its scale. Inverting changes the value by a positive factor and never changes the sign, so the two non-zero strata are stable under inversion wherever inversion exists, and the null set is stable on its units. It is not stable under multiplication: the four witnesses of §The Fourth Norm move between all three strata.

The algebraic norm is not a function of the K norm, and the reason is that it is onto on the null set.

$$ \mathrm{Kr}\bigl(c(e_t+\vec e_x)\bigr)=0, \qquad N\bigl(c(e_t+\vec e_x)\bigr)=2c^{2}, \qquad c\in\mathbb{C}^{\times}. $$

So the wall carries every non-zero algebraic norm, and the value carries none. In the other direction $e_t$ and $\vec e_x$ have the same algebraic norm, $1$, and values $1$ and $-1$. The two invariants are independent, and neither decides the other.

No Spectral Refinement

The temptation is explicit, and it is the temptation the tenth section creates. There the plane of the norm and the discriminant organises four strata, because both coordinates are coefficients of one characteristic polynomial, $p_{\tilde Q}(X)=X^{2}-2Q_tX+N$ with discriminant $4B^{2}$. The same construction with the K value on one axis and $B^{2}$ on the other would give a regular and a degenerate article with two spectral halves each. That plane is declined here, for four reasons that are independent of one another.

The first reason is that the isometry group is transitive on the wall, so it moves the spectral type.

$$ B^{2}(e_t+\vec e_x)=1, \qquad B^{2}\Bigl(e_t+\tfrac{1}{\sqrt2}(\vec e_x+i\vec e_y)\Bigr)=\tfrac12-\tfrac12=0, \qquad \mathrm{Kr}=0\ \text{for both}. $$

An isometry of the form therefore carries a spectrally regular null element to a spectrally degenerate one. A stratum of the proposed plane would not be stable under the group that defines the plane's first axis, and a subdivision that a group destroys is not a subdivision of the object the group acts on.

The second reason is that the K value is not a coefficient of the characteristic polynomial. Take $Q_t=1$ with the two vector parts $(i,0,0)$ and $(1,i,i)$. Both give $B^{2}=-1$ and $N=0$, so both have the same polynomial $X^{2}-2X$ and the same spectrum $\{0,2\}$, and their values are

$$ 1-1=0, \qquad 1-(1+1)=-2 . $$

One polynomial, two values, of two different signs. The K norm is not a function of $p_{\tilde Q}$, and $p_{\tilde Q}$ is not a function of the K norm.

The third reason is that the K value is not invariant under similarity. Conjugation by the unit $\tilde U=e_t+2i\vec e_x$ sends $\vec e_y$ to

$$ \tilde U\vec e_y\tilde U^{-1}=-\tfrac53\vec e_y-\tfrac43 i\vec e_z, $$

so the square of the vector part stays $1$, as it must, and the value moves from $-1$ to $-41/9$. A similarity is the natural change of element in the spectral problem, and it is not a symmetry of the weight. An isometry is a symmetry of the weight, and it is not a symmetry of the spectral type. The two problems are carried by different groups and neither group contains the other.

The fourth reason is that every stratum meets both spectral types, and one stratum meets only one.

K stratum real dimension spectrally degenerate witness spectrally regular witness
positive region $8$ $e_t$, with $B^{2}=0$ $e_t+\tfrac12\vec e_x$, with $B^{2}=\tfrac14$
negative region $8$ $\vec e_x+i\vec e_y$, with $B^{2}=0$ $\vec e_x$, with $B^{2}=1$
doubly null set off the origin $4$ none $e_t+i\vec e_x$, with $B^{2}=-1$
isotropic units $7$ $e_t+\tfrac{1}{\sqrt2}(\vec e_x+i\vec e_y)$ $e_t+\vec e_x$, with $B^{2}=1$

The third row is decisive. A spectral split would separate the origin from the Peirce lines, although the algebraic partition groups them, both being zero divisors and both being doubly null. It would also cut the null set into two pieces of dimensions five and seven, neither of which is invariant under the isometries. The plane is definable and it groups the wrong things.

The refinement that does exist pairs the K value with the algebraic norm.

$\mathrm{Kr}$ $N\neq0$ $N=0$
positive non-empty: $e_t$ empty
negative non-empty: $e_t+2\vec e_x$ non-empty: $\vec e_x+i\vec e_y$
null non-empty: $e_t+\vec e_x$ non-empty: $e_t+i\vec e_x$

Five of the six cells are non-empty and the empty one is the theorem of §The Two Partitions. Each cell is invariant under one of the two groups, the value under the isometries and the algebraic norm under the algebra automorphisms, so each cell is carved by a genuine invariant. This is the partition that the operator theory follows, and it is why the two sector articles are cut by the algebraic norm and not by the discriminant.

On the two sectors the pair degenerates to a single axis, and the degeneration explains why the physical readings are exact there. On the material sector $\mathrm{Kr}=-N$ and on the informational sector $\mathrm{Kr}=+N$, so on either sector the K value is a function of the algebraic norm and carries no independent information. The independence of the two is a property of the complexification, and it is stated as such.

The Six Restrictions

The restriction of the K norm to each of the six remarkable subspaces has a definite inertia, and the inertias decide the algebraic status of the null elements.

subspace real dimension inertia of $\\mathrm{Kr}$ non-zero null elements algebraic status
the centre $\mathbb{B}_{C}$ $2$ $(2,0)$ none —
the vector subspace $\mathbb{B}_{V}$ $6$ $(0,6)$ none —
the quaternion subspace $\mathbb{B}_{H}$ $4$ $(1,3)$ a cone of dimension $3$ every one is invertible
the anti-quaternion subspace $\mathbb{B}_{iH}$ $4$ $(1,3)$ a cone of dimension $3$ every one is invertible
the informational sector $\mathbb{B}_{\Psi}$ $4$ $(1,3)$ a cone of dimension $3$ every one is a zero divisor
the material sector $\mathbb{B}_{\Phi}$ $4$ $(1,3)$ a cone of dimension $3$ every one is a zero divisor

Two of the six are definite and four are Lorentzian. On the centre the form is $|Q_t|^{2}$, positive definite, and on the vector subspace it is $-\lVert\vec Q\rVert^{2}$, negative definite. Neither carries a non-zero null element, so on those two subspaces the K norm decides nothing beyond its sign.

The four Lorentzian restrictions are where the algebraic status matters, and it splits two and two.

On the quaternion subspace the coefficients are real, so the two norms are

$$ N=q_t^{2}+\lVert\vec q\rVert^{2}, \qquad \mathrm{Kr}=q_t^{2}-\lVert\vec q\rVert^{2}. $$

A null element has $q_t^{2}=\lVert\vec q\rVert^{2}$, hence $N=2q_t^{2}\neq0$ off the origin, and it is invertible. The same computation on the anti-quaternion subspace, where all four coefficients are purely imaginary, gives $\mathrm{Kr}=q_t^{2}-\lVert\vec q\rVert^{2}$ and $N=-(q_t^{2}+\lVert\vec q\rVert^{2})$, which is also non-zero off the origin.

On the informational sector the scalar part is real and the vector part purely imaginary. A null element has $a_0^{2}=\lVert\vec a\rVert^{2}$, which is exactly the vanishing of $N=a_0^{2}-\lVert\vec a\rVert^{2}$, so every non-zero null element is a zero divisor. The material sector is the same statement multiplied by $i$, with the signs of both norms reversed.

The sharp consequence is that lightlike and massless coincide on the two sectors and nowhere else. On $\mathbb{B}_{\Phi}$ the two conditions are one equation, because $\mathrm{Kr}=-N$. On $\mathbb{B}_{\Psi}$ they are also one equation, because $\mathrm{Kr}=+N$. On the quaternion subspaces they are two, and the null elements there are massive.

The table records six restrictions and not a cover. A generic null element has all four coefficients non-zero and lies in none of the six subspaces. The six rows are the charts, and the generic point is the reason the readings of the next section are attached to the charts and not to the algebra as a whole.

The Operators

The operator theory follows the algebraic norm, and the K value follows nothing. This is the practical consequence of the two partitions, and it is worth stating as a table.

operator follows behaviour on the K strata
left multiplication $L_{\tilde Q}$ $N$ rank $4$ on every invertible element, rank $2$ on every zero divisor off the origin, rank $0$ at the origin
the two-sided operator $\mathrm{Sd}_{\tilde Q}$ $N$ invertible on the invertibles, of rank $1$ on the null cone off the origin, zero at the origin
the polar decomposition $N$ exists on the units, fails on the cone
the K adjoint the form exists everywhere, because the form is non-degenerate

The rank of left multiplication is the sharpest of the four. It is four on the whole group of invertibles, which includes the positive region, the negative units and the null units; it is two on the whole zero-divisor cone, which includes the negative zero divisors and the doubly null stratum. So the rank does not follow the sign of the value: $e_t+\vec e_x$, of value zero, has rank four, while $\vec e_x+i\vec e_y$, of value $-2$, has rank two.

The two-sided operator is the one that sees the algebraic cone, and it sees it as a rank drop.

$$ \mathrm{Sd}_{\tilde Q}(\mathbb{B})=\mathbb{C}\tilde Q\tilde Q^{*}, \qquad \operatorname{rank}\mathrm{Sd}_{\tilde Q}=1, \qquad N(\tilde Q)=0,\ \tilde Q\neq0 . $$

The reason is that $\mathrm{Sd}_{\tilde Q}(\tilde Q^{\natural})=\tilde Q\tilde Q^{\natural}\tilde Q^{*}=N\tilde Q^{*}=0$, and the matrix model writes $\tilde Q=uv^{*}$ with $v^{*}u=0$. Off the cone the operator is invertible, with $\mathrm{Sd}_{\tilde Q}^{-1}=\mathrm{Sd}_{\tilde Q^{-1}}$.

The singular set of the map $X\mapsto\tilde QX\tilde Q^{*}$ is therefore the algebraic null cone, on which it loses three-quarters of its rank, and that set is not the K null set. A negative zero divisor is a singular case for the operator and a regular case for the value; a null unit is the reverse. The two failures are different, and the entry keeps them apart.

The polar decomposition exists exactly on the group of invertibles, so the K norm adds no condition. The modulus requires $\varrho^{2}e^{2i\theta_{\mathrm{pha}}}=N(\tilde Q)$ with $\varrho\,e^{i\theta_{\mathrm{pha}}}\neq0$, which is the algebraic condition and nothing more. On the positive region the decomposition therefore exists throughout, and on the negative region it exists exactly off the zero divisors.

The K adjoint is defined everywhere, and it is where the form enters the operator theory.

$$ T^{\dagger}=JT^{*}J, \qquad \langle T\tilde X,\tilde Y\rangle_{\natural*}=\langle\tilde X,T^{\dagger}\tilde Y\rangle_{\natural*} . $$

The $J$-self-adjoint and the $J$-unitary operators are the subject of the cited article. The point for this entry is that the adjoint needs only non-degeneracy, and non-degeneracy is a property of the form that survives at every point of the algebra, including the wall.

The level sets of the value are single orbits of the isometry group. For $c>0$ the set $\{\mathrm{Kr}=c\}$ is $\mathrm{U}(1,3)/\mathrm{U}(3)$, for $c<0$ it is $\mathrm{U}(1,3)/\mathrm{U}(1,2)$, and both are of real dimension seven. The positive family is the complex hyperbolic ball, the negative family is its exterior, and the wall is the boundary at infinity of the first. That geometry is the subject of The Krein Level Sets and the Hyperbolic Structure and is quoted here, not rebuilt.

The Interpretations

I.11.a — The Fourth Norm Is an Inertia, Not a Magnitude

Read: the four norms of §The Fourth Norm, the reality of the K value, and the sign that no other norm of the algebra can carry.

The Hermitian norm is a size and it is never negative. The quaternionic norm is an invertibility and it is complex. The K norm is the only one of the four whose value is real and signed, and that is the whole of its content. It does not rank elements by size, it does not decide invertibility, and it does not carry a phase. It sorts.

The reading is that the algebra carries one number per element that is purely a metric position. Physics has such a number on every four-vector: the squared interval, positive inside the light cone, negative outside, zero on it. The K norm is the algebra's own version of that number, taken on the coefficients rather than on a chart.

The comparison with the plain norm fixes what is specific to the K norm. The plain norm $P=Q_t^{2}-B^{2}$ has the same inertia and a complex value; it is the bilinear object. The K norm replaces every square by the square of a modulus, and the replacement does two things at once: it makes the value real and it makes the value invariant under both conjugations. A metric position must be real and must not care which conjugation is applied, and those two requirements produce the K norm and nothing else.

The boundary. An inertia is a sign pattern and not a physical quantity. That the algebra's inertia is $(1,3)$ is a statement about the coefficient basis; that the physical metric has one time and three space directions is a reading of that statement, and the two are not the same proposition.

I.11.b — The Weight of a State Space That Keeps a Time Direction

Read: the split into two sectors and a wall, the positivity of the form on the centre and its negativity on the vector subspace, and the bridge $\langle\tilde P,\tilde Q\rangle_{\natural*}=\langle\tilde P,J\tilde Q\rangle_{*}$ to the definite companion.

A state space that keeps a time direction must carry an indefinite inner product. If the inner product is positive definite, no direction is distinguished and the space has no clock. If one direction counts with a plus and the others with a minus, the space has a clock and nothing else has changed. The K norm is exactly such a form: plus on the centre, minus on the vector subspace, and the fundamental symmetry $J={}^{\natural}$ is the operator that turns the indefinite form into the definite one.

The reading is that the K norm is the weight of such a space and the Hermitian norm is its probability. The two are related by the fundamental symmetry and by nothing else: the indefinite form is the definite form with the parity inserted. So the pair $(H,\mathrm{Kr})$ is the pair that a Lorentzian state space needs, one form for the norm of a state and one for the inertia of the space it lives in.

The corpus already has the definite half, and it has it in one place. The state slice of the informational sector is the Bloch ball, with the pure states on the boundary and the mixed states inside. That state space is non-relativistic, and the entry of that sector says so: a boost carries a state off the slice. The K norm is the form that a relativistic state space would carry, and the algebra supplies it while it does not supply the state space.

The boundary. No Hilbert space is constructed here, no inner product is completed, and no dynamics is supplied. The statement is that the algebra contains one indefinite form with the right inertia and one definite form with the right positivity, and that the two are one object read through the parity. Whether a physical state space is built on them is a separate question.

I.11.c — Positive Implies Invertible: a Timelike Element Is Never Massless

Read: the theorem that a positive element is invertible, its corollary that a zero divisor has non-positive value, and the three witnesses that the converses fail.

The one implication of the partition has a physical form: an element with a positive metric weight is invertible. Nothing with a positive value is a zero divisor, so nothing with a positive value collapses. The collapse cone lies entirely in the closed negative half-space, and it touches the wall only on the Peirce lines.

The reading is that the interior of the light cone is safe. A four-momentum inside the cone has a positive squared interval and is invertible in the algebra; a four-momentum on the cone is a zero divisor on the material chart; a four-momentum outside the cone is spacelike and has no mass shell. The passage from unit to zero divisor happens at the wall, and the wall is where the mass vanishes.

The implication is one way, and the failure of the converse is the interesting part. an invertible element can be negative, and it can be null. So invertibility is strictly weaker than positivity, and the algebra contains units of every sign pattern. The reading is that reversibility is not the same property as being inside the cone: a spacelike element is perfectly invertible and a lightlike element can be invertible too.

The boundary. The theorem is about the plain product and the K norm, and its physical reading uses the material chart. On that chart the implication is nearly trivial, because the value is minus the algebraic norm and positivity is the timelike condition. In the full algebra the implication is not trivial, because the two conditions are independent there, and that is where the theorem earns its place.

I.11.d — The Positive Region Is a Clock Times a Subluminal Distortion

Read: the parametrisation $\tilde Q=Q_t(e_t+W)$ with $\lVert W\rVert<1$, and its negative counterpart $\tilde Q=r(we_t+W)$ with $\lvert w\rvert<1$ and $\lVert W\rVert=1$.

Every positive element is a complex scale times a distortion of modulus less than one, and every negative element is a positive scale times a unit direction and a time component of modulus less than one. The two parametrisations are not a convenient choice of coordinates: they are the positivity and the negativity written out.

The reading uses the rapidity, and the rapidity is the corpus's own word. Writing $\lVert W\rVert=\tanh\Xi$ turns the inequality $\lVert W\rVert<1$ into the statement that $\Xi$ is real and finite, and turns the positive region into the set of elements of finite rapidity. A distortion of modulus one is an infinite rapidity, and it is the wall.

On the material chart the inequality is the velocity bound. The vector part of a four-momentum over its scalar part has modulus $\lvert\vec p\rvert/p_0$, and $\lvert\vec p\rvert

The boundary. The parametrisation is exact and the notation $W$ is a coordinate. The name distortion is a reading of the pair (scale, remainder), and the identification of the remainder with a boost is a reading of the informational and material charts. Nothing in the parametrisation forces either name.

I.11.e — Inversion Keeps the Character and Rescales the Weight

Read: the formula $\mathrm{Kr}(\tilde Q^{-1})=\mathrm{Kr}(\tilde Q)/\lvert N(\tilde Q)\rvert^{2}$ and the corollary that each non-zero stratum is inversion-stable.

Running an element backwards never changes its side of the wall, and it always changes its weight. The factor is positive, so the sign survives inversion and the value does not. A positive element has a positive inverse, a negative element a negative inverse, and a null element a null inverse; the magnitudes differ by the square of the modulus.

The reading is that causal character is invariant under reversal and the weight is not. A process and its reverse are both inside the cone or both outside it. Their weights are different numbers, and the difference is the square of the algebraic norm, which is the multiplicative measure of the element.

The asymmetry is the point. A quantity that survives reversal unchanged is a charge or a label; a quantity that changes by a positive factor is a measure. The K value is the second. The algebraic norm is the first: it is multiplicative, so inversion carries it to its reciprocal and the weight's rescaling is written in terms of it.

The boundary. The formula is one line of algebra and the reading is a statement about processes in general. No time-reversal operator is constructed, and the entry does not claim that inversion is the physical reversal of a process. It claims that the two strata are stable under it and that the value is not.

I.11.f — The Sign Is Not a Composition Law

Read: the four witnesses of non-multiplicativity, and the contrast with the quaternionic norm, which is multiplicative in the strict sense.

The sign of the weight is a property of one element, not a rule for products. Two positive elements can have a negative product, two negative elements a positive one, two regular elements a null one, and two null elements a regular one. Every combination occurs. There is no addition law, no multiplication law, and no inequality except the one that says a zero divisor cannot be positive.

The reading is that the causal character is not a conserved charge. A conserved label would need a composition law, and the only composition law the algebra has is the multiplicativity of $N$. The K value has none, so it cannot label a conserved quantity; it labels a position.

This is the sharpest contrast in the section, and it is worth stating as a pair. The interval is a law: the modulus of a product is the product of the moduli, up to the phase carried by the product. The weight is not a law: it can vanish, change sign, or grow across a product with no constraint other than homogeneity. One measure composes, the other locates.

The boundary. The witnesses are elements and not processes, and the entry does not claim that physical composition is the algebra's product. It claims that any reading in which the K value is a conserved quantity is excluded by the algebra, and that exclusion is exact.

I.11.g — Inside Is a Circle Times a Ball, Outside Is a Sphere Times a Disc

Read: the two parametrisations, the homotopy types $S^{1}$ and $S^{5}$, and the theorem that the two regions are not homeomorphic.

The two sides of the wall have different topology, so they are not two halves of one space. The positive region is a punctured plane times an open ball and it retracts onto a circle: inside the cone, the only topology is the phase. The negative region is a half-line times a disc times a five-sphere and it retracts onto a five-sphere: outside the cone, a full five-sphere of directions survives.

The reading is that a time direction has no freedom beyond its phase and a spatial direction has a sphere of freedom. Inside the cone there is one circle, the clock. Outside there is a five-sphere, and its five dimensions are the six real parameters of a complex three-vector constrained to unit modulus, minus the overall phase. The asymmetry is not put in by hand; it is the asymmetry of the defining inequality.

The consequence is that no deformation carries one region to the other without crossing the wall. A path that changes the sign of the weight must pass through zero, and at zero the element spans a null line. The wall is therefore not an artefact of the parametrisation: it separates two components of the complement of the null set, and the components are separated by a topological invariant and not only by an inequality.

The boundary. Homotopy type is a statement about the regions as they sit in the algebra, and no physical significance is claimed for the number five. The reading is that the two sides are qualitatively different, and the entry states the difference in the strongest form the mathematics offers.

I.11.h — Lightlike Is Not Massless

Read: the two cones, of real dimensions seven and six, their intersection of real dimension four, and the two witnesses $e_t+\vec e_x$ and $\vec e_x+i\vec e_y$.

In the full algebra the K null set and the zero-divisor set are different objects. The first has codimension one and contains units; the second has codimension two and contains only zero divisors. The element $e_t+\vec e_x$ is lightlike and has algebraic norm $2$; the element $\vec e_x+i\vec e_y$ is a zero divisor and has value $-2$. Neither set contains the other, and their intersection is the doubly null set.

The reading is that on the light cone and massless are two different statements, and physics uses them as one because it works on one sector. On the material chart the K norm is minus the algebraic norm, so the two conditions are one equation. On the informational chart the K norm equals the algebraic norm, and the two conditions are again one equation. On the quaternion subspaces they are two equations and the null elements there are massive.

The doubly null stratum is where the two conditions meet, and it is the union of the Peirce lines. Those lines are parameterised by a real unit vector and a complex scale, they are of real dimension four, and each of them carries an idempotent $\tfrac12(e_t+i\vec u)$. The intersection of the two cones is therefore not a generic cone but a cone of idempotent lines.

The boundary. The reading identifies the light cone of physics with the material restriction of the K null set, and that identification is a reading of the chart. The algebra's statement is the pair of dimensions and the two witnesses, and those are exact.

I.11.i — The Pure State Is Doubly Null

Read: the coincidence $\mathrm{Kr}=N$ on the informational sector, the determinant form of the value on the unit-weight slice, and the identification of the doubly null set with the Peirce lines.

On the informational sector the K value is the algebraic norm, and on the unit-weight slice the algebraic norm is the determinant. For a state $\tfrac12(e_t+i\vec s_{\Psi})$ with $\vec s_{\Psi}$ real and $\lvert\vec s_{\Psi}\rvert\le1$,

$$ \mathrm{Kr}=\tfrac14\bigl(1-\lvert\vec s_{\Psi}\rvert^{2}\bigr), $$

so the three K strata are the three classes of the state space. Positive is the interior of the Bloch ball, the mixed states. Null is the sphere $\lvert\vec s_{\Psi}\rvert=1$, the pure states. Negative is the exterior, the set of unit-weight Hermitian elements that are not positive and therefore are not states.

The reading is that the weight measures mixedness, and that a pure state has none of it. The value is largest at the maximally mixed state $\tfrac12e_t$, where it is $\tfrac14$, and it vanishes exactly on the pure states. A pure state is a zero divisor and a doubly null element at once, and the third statement is that it is spectrally regular: on the doubly null set $B^{2}=-Q_t^{2}\neq0$, so the two eigenvalues are distinct.

The three statements are one. The purity, the vanishing of the weight and the vanishing of the algebraic norm are the vanishing of the determinant. The spectral regularity is the statement that a rank-one two-by-two matrix has eigenvalues one and zero, which never coincide. So the entry's coincidences are the coincidences of the determinant.

The boundary. The state slice and the Bloch ball are the informational sector's reading, and the entry uses them as given. That a pure state has no weight is a statement about the determinant, and the name mixedness is a reading of the determinant on the unit-weight slice.

I.11.j — The Weight Is Not a Spectral Quantity

Read: the two elements with the same characteristic polynomial and different values, the similarity that moves the value and not the spectrum, and the general absence of a spectral refinement.

The K value is not a function of the characteristic polynomial and it is not preserved by similarity. Two elements with the polynomial $X^{2}-2X$ have values $0$ and $-2$. A similarity that fixes the square of the vector part moves the value from $-1$ to $-41/9$. In the other direction two elements of the same value have different vector squares, $\tfrac12$ and $0$.

The reading is that the weight is a property of the frame and the spectrum is a property of the modes. A change of frame is an isometry, and it preserves the weight; a change of basis is a similarity, and it preserves the spectrum. The two transformations are different groups, neither contains the other, and there is no coordinate system in which both are diagonal at once.

The physics of the distinction is the distinction between an observable and a label. The spectrum of an element is what a mode decomposition would read: two eigenvalues, their separation, their coincidence. The weight is what a frame would read: which side of the cone, and how far. A theory needs both, and the entry's negative result is that it cannot get the second by refining the first.

The boundary. On each of the two sectors the weight is a function of the algebraic norm, so the distinction collapses there and the entry says so. The independence is a property of the complexification, and the readings that use it are readings of the full algebra and not of a chart.

I.11.k — The Isometry Group Is the Gauge Group of the Weight

Read: the transitivity of $\mathrm{U}(1,3)$ on the punctured null set, its preservation of the value, and its failure to preserve the spectral type.

The group that preserves the weight is $\mathrm{U}(1,3)$, of real dimension sixteen, and it acts transitively on the wall. Every non-zero null element is carried to every other by an isometry, so inside the wall there is no invariant except the value, which is constant. The group is larger than the Lorentz group, and the extra directions are the ones that change the phase of the scalar part and rotate the three complex directions among themselves.

The reading is that this group is the gauge group of the frame and not the Lorentz group. Lorentz is the subgroup that preserves the material real slice. The whole of $\mathrm{U}(1,3)$ preserves the form, and a form-preserving transformation that is not Lorentz moves an element off the sector. The gauge reading is a name for the invariance of the weight, and it is not a derivation of a gauge field.

The transitivity is what forbids the spectral split, and the two statements are one theorem. Because the group is transitive on the wall, a function invariant under it is constant there; because the square of the vector part is not constant there, it is not an invariant of the weight. A classification by the weight can therefore have no second axis inside the wall.

The boundary. No gauge field, no connection and no curvature is constructed. The group is the isometry group of one form on one algebra, and the identification of it with a physical gauge group is not made.

I.11.l — The Indefinite Cone Is Not a State Space

Read: the three failures of conehood, the Krein cone $P_K$, and the contrast with the positive cone of the Hermitian form.

The positive region of an indefinite form is not a probability cone. It is not pointed, because it contains opposite rays. It is not convex, because two null elements on opposite sides have their midpoint outside it. It is not traced, because the form defines no trace on the algebra. A probability cone needs all three, so the K cone cannot be one.

The reading is that this is the price of a Lorentzian state space, and it is paid in one coin. A definite form gives a convex cone, a norm and a normalisable state; it also gives no time direction. An indefinite form gives the time direction and loses the convexity. A theory that wants both must carry two forms and use each for what it can do.

The algebra is explicit about which form does which. The Hermitian norm supplies the amplitude, the topological size and the unit-weight slice; the polar decomposition, the Born weight and the Bloch ball all follow it. The K norm supplies the inertia, the side of the cone and the isometry group; the indefinite geometry and the $J$-adjoint follow it. Neither can do the other's work, and the entry's two partitions are the statement of that division.

The boundary. Nothing here says that a physical state space is indefinite. It says that if one is, its positivity cone is not convex, and that the algebra contains exactly one form for each of the two roles. The choice of which form is the probability is a reading, and the entry makes it in favour of the Hermitian one.

The Limits

The object is a quadratic form and the readings are readings of a quadratic form. The whole mathematical content is one real polynomial of degree two on eight real coordinates, together with its polar form, its inertia and its isometry group. Everything the entry names — weight, light cone, mass, mixedness, state space — is a name for what that polynomial means when a chart supplies a physical vocabulary, and the names carry no content beyond the polynomial and the chart.

No dynamics is supplied. The norm says which side of the wall an element is on and says nothing about how an element changes. A propagation law that keeps an element on one level set, one that moves it across the wall, and one that does neither are all compatible with the norm. Conservation of the weight is not implied and is not excluded; it is simply not in the object.

No statistics is supplied. The indefinite inner product of operator theory comes with a Krein space structure, a $J$-orthogonal decomposition and an indefinite spectral theory, and none of that is built here. The entry has no Hilbert space, no completion, no self-adjoint operator on a Krein space and no probability rule beyond the identification of the Hermitian norm with the one the informational sector already uses.

No measurement theory is supplied. The state slice, the Born weight and the projection postulate belong to the informational sector's entry, and this entry refers to them and does not restate them. The K norm is the form of a relativistic state space and it is not a rule for outcomes; the rule for outcomes is the definite form's.

No gauge field is supplied. The group $\mathrm{U}(1,3)$ is the isometry group of the form on the coefficient space. It is not promoted to a local symmetry, no connection is chosen, and no field strength is computed. The word gauge in I.11.k names an invariance and not a theory.

No scale is fixed. The norm is homogeneous of degree two, so every non-zero element has a one-parameter family of rescalings that stay on its level set. Nothing in the form selects an invertible element, and the choice of a level set on which to normalise an element is a separate act. The physical readings use the mass shell, and the mass shell is a chart-dependent choice of level.

No topology is supplied beyond the homotopy type. The two regions retract onto a circle and a five-sphere, the wall onto a circle times a five-sphere, and that is the whole of the topology claimed. There is no index, no characteristic class and no invariant that distinguishes two elements of one level set.

The two sectors are charts and not subalgebras with a physics of their own. The identities $\mathrm{Kr}=-N$ on matter and $\mathrm{Kr}=+N$ on information are exact, and every sector reading in this entry is a reading through one of them. A statement that holds on one chart and fails on the other is a statement about the chart, and the entry says which chart it is using at each place.

The form is standard and the arrangement is not. The Krein form, its inertia, its isometry group, its hyperbolic level sets and its indefinite operator theory are classical. What this entry contributes is the placement: the norm inside a four-norm family, the comparison of its partition with the algebraic one, the identification of the doubly null stratum with the Peirce lines, and the refusal of a spectral split with the four reasons for it.

Summary

result content
the norm $\mathrm{Kr}(\tilde Q)=\lvert Q_t\rvert^{2}-\lVert\vec Q\rVert^{2}$; real, homogeneous of degree two, non-degenerate
the inertia $(1,3)$ on $\mathbb{C}^{4}$, $(2,6)$ on the eight real parameters; positive on the centre, negative on the vector subspace
the companions $\mathrm{Kr}$ is the plain norm $P$ read on moduli, $H$ is the quaternionic norm $N$ read on moduli
the bridge $\langle\tilde P,\tilde Q\rangle_{\natural*}=\langle\tilde P,J\tilde Q\rangle_{*}$ with $J={}^{\natural}$, the fundamental symmetry
multiplicativity none: $(e_t+\vec e_x)(e_t-\vec e_x)=2e_t$ of value $4$; $(e_t+\tfrac12\vec e_x)^{2}=\tfrac34\vec e_t+\vec e_x$ of value $-\tfrac{7}{16}$
the invariance preserved by both conjugations; preserved by no algebra multiplication except a scalar of modulus one
the three sets $\mathbb{B}^{\natural*}_{+}$ of dimension $8$, $\mathcal K$ of dimension $7$, $\mathbb{B}^{\natural*}_{-}$ of dimension $8$
the line criterion $\mathbb{C}\tilde Q$ is positive definite, negative definite or totally isotropic according to the sign of the value
the topologies $\mathbb{B}^{\natural*}_{+}\simeq S^{1}\times B^{6}$, $\mathcal K\simeq S^{1}\times S^{5}$, $\mathbb{B}^{\natural*}_{-}\simeq S^{5}\times D^{2}$; no two are homeomorphic
the parametrisations $Q_t(e_t+W)$ with $\lVert W\rVert<1$; $r(we_t+W)$ with $\lvert w\rvert<1$ and $\lVert W\rVert=1$
the cone $P_K=\{\lvert Q_t\rvert\ge\lVert\vec Q\rVert\}$; closed and star-shaped, neither convex nor pointed
the transitivity $\mathrm{U}(1,3)$ acts transitively on $\mathcal K\setminus\{0\}$
the doubly null set $\mathcal K_{\mathrm d}=\mathbb{C}^{\times}\cdot(e_t+iS^{2})$, dimension $4$, the union of the Peirce lines $\mathbb{C}(e_t+i\vec u)$
the doubly null element $\tilde\Pi_{\vec u}=\tfrac12(e_t+i\vec u)$, an idempotent; spectrally regular, $B^{2}=-Q_t^{2}\neq0$
the isotropic units $\mathcal K_{\mathrm u}$, dimension $7$, open and dense in $\mathcal K$; $N$ takes every non-zero value there
positive implies invertible no positive element is a zero divisor; every zero divisor has $\mathrm{Kr}\le0$, with equality on $\mathcal K_{\mathrm d}$
the converses fail: $e_t+2\vec e_x$ negative unit, $e_t+\vec e_x$ null unit, $\vec e_x+i\vec e_y$ negative zero divisor
inversion $\mathrm{Kr}(\tilde Q^{-1})=\mathrm{Kr}(\tilde Q)/\lvert N(\tilde Q)\rvert^{2}$; the sign is preserved, the value rescaled
the independence $e_t$ and $\vec e_x$ have $N=1$ and values $1,-1$; $c(e_t+\vec e_x)$ has value $0$ and any $2c^{2}$
no spectral split the value is not a coefficient of $p_{\tilde Q}$, is not similarity-invariant, and is moved by the isometries
the same polynomial $Q_t=1$ with vector parts $(i,0,0)$ and $(1,i,i)$: both $X^{2}-2X$, values $0$ and $-2$
the similarity witness conjugation by $e_t+2i\vec e_x$ keeps $B^{2}=1$ and moves the value from $-1$ to $-41/9$
the refinement the pair $(\mathrm{Kr},N)$ has five non-empty cells of six, the empty one being positive and degenerate
the six restrictions $(2,0)$ on the centre, $(0,6)$ on the vector subspace, $(1,3)$ on each of the four remaining subspaces
the sector identities $\mathrm{Kr}=-N$ on $\mathbb{B}_{\Phi}$, $\mathrm{Kr}=+N$ on $\mathbb{B}_{\Psi}$; null is massless on both and only there
the state slice $\mathrm{Kr}=\tfrac14(1-\lvert\vec s_{\Psi}\rvert^{2})=N$; positive mixed, null pure, negative not a state
the operators $L_{\tilde Q}$ of rank $4$ on the invertibles and $2$ on the zero divisors; $\mathrm{Sd}_{\tilde Q}$ invertible on the invertibles and of rank $1$ on the null cone
the adjoint the K adjoint $T^{\dagger}=JT^{*}J$ exists everywhere; the polar decomposition exists exactly on the invertibles
the level sets $\{\mathrm{Kr}=c\}$ is $\mathrm{U}(1,3)/\mathrm{U}(3)$ for $c>0$ and $\mathrm{U}(1,3)/\mathrm{U}(1,2)$ for $c<0$, both of dimension $7$
the geometry the positive family is the complex hyperbolic ball, its boundary the wall, the level sets the Bergman spheres
the readings weight, inertia, light cone, mass, mixedness, gauge frame, and the price of an indefinite state space

Summary of Notation

object notation content
the element $\tilde Q=Q_te_t+Q_x\vec e_x+Q_y\vec e_y+Q_z\vec e_z$ four complex coefficients
the K norm $\mathrm{Kr}(\tilde Q)$ $\lvert Q_t\rvert^{2}-\lVert\vec Q\rVert^{2}$, real
the plain norm $P(\tilde Q)$ $Q_t^{2}-B^{2}$, complex
the quaternionic norm $N(\tilde Q)$ $Q_t^{2}+B^{2}$, complex, multiplicative
the Hermitian norm $H(\tilde Q)$ $\lvert Q_t\rvert^{2}+\lVert\vec Q\rVert^{2}$, positive
the vector square $B^{2}$ $Q_x^{2}+Q_y^{2}+Q_z^{2}$
the vector norm $\lVert\vec Q\rVert^{2}$ $\lvert Q_x\rvert^{2}+\lvert Q_y\rvert^{2}+\lvert Q_z\rvert^{2}$
the polar form $\langle\tilde P,\tilde Q\rangle_{\natural*}$ the sesquilinear form of which $\mathrm{Kr}$ is the diagonal
the fundamental symmetry $J={}^{\natural}$ the quaternionic conjugation; $\langle\tilde P,\tilde Q\rangle_{\natural*}=\langle\tilde P,J\tilde Q\rangle_{*}$
the positive region $\mathbb{B}^{\natural*}_{+}$ $\{\mathrm{Kr}>0\}$, real dimension $8$
the negative region $\mathbb{B}^{\natural*}_{-}$ $\{\mathrm{Kr}<0\}$, real dimension $8$
the null set $\mathcal K$ $\{\mathrm{Kr}=0\}$, real dimension $7$
the doubly null set $\mathcal K_{\mathrm d}$ $\mathcal K\cap\mathcal N$, real dimension $4$
the isotropic units $\mathcal K_{\mathrm u}$ $\mathcal K\setminus\mathcal K_{\mathrm d}$, real dimension $7$
the algebraic null cone $\mathcal N$ $\{N=0\}$, real dimension $6$
the Krein cone $P_K$ $\{\mathrm{Kr}\ge0\}$
the distortion $W$ the vector part over the scalar part, of norm $<1$ on the positive region
the rapidity $\Xi$ $\lVert W\rVert=\tanh\Xi$
the Peirce idempotent $\tilde\Pi_{\vec u}$ $\tfrac12(e_t+i\vec u)$ for a real unit $\vec u$
the isometry group $\mathrm{U}(1,3)$ the group preserving the form, real dimension $16$
the K adjoint $T^{\dagger}$ $JT^{*}J$
the two-sided operator $\mathrm{Sd}_{\tilde Q}$ $\tilde X\mapsto\tilde Q\tilde X\tilde Q^{*}$
the charts $\mathbb{B}_{\Phi}$, $\mathbb{B}_{\Psi}$ the anti-Hermitian and the Hermitian sectors, of real dimension $4$

Further Reading

The mathematics of this entry. Kr-Regularity and Degeneracy in the Biquaternion Algebra carries the partition and the four reasons against a spectral split. The Kr-Regular Sector carries the two non-zero regions with their parametrisations, products and inversion. The Kr-Degenerate Sector carries the wall, its two strata and the Peirce lines.

The form itself. The Four Forms and the Four Algebraic Norms over the Biquaternion Algebra places the K norm among the four. The Fundamental Symmetry of the Form carries the operator $J$ and the bridge to the definite form. Remarkable Subspaces under the General Quaternionic Sesqualgebra carries the six restrictions.

The geometry of the level sets. The Krein Level Sets and the Hyperbolic Structure carries the complex hyperbolic ball, its boundary and the real hyperbolic slice. The Krein Isometry Group and Its Cartan Decomposition carries the group and its action.

The algebraic partition. N-Regularity and N-Degeneracy in the Biquaternion Algebra carries the multiplicative norm and the zero-divisor cone. Relations between the N Norm and Spectral Regularity and Degeneracy carries the operator ranks. The Four General Products and Their Relations over the Biquaternion Algebra carries the two-sided operator and the annihilators.

The earlier entries of the category. The H Norm: Amplitude and Frame reads amplitude and frame. The N Norm: Mass, Angle and the Cone reads mass, angle and the cone. The partition of the tenth section is read in five entries, from The Partition: Two Invariants and Four Regimes to The Vacuum and the Nilpotent Ladder. This entry is the fourth and last of the four norms and it closes the norm family.

What this entry changes. It keeps the mathematics of the earlier reading and changes the placement: the form is read as one of four norms, its partition is compared with the algebraic one, and the spectral split that the earlier reading leaves open is declined with four reasons.