<!-- Generated from https://www.teclamairme.com/#group-T10 by readviews.py; never edited by hand. -->

# Glossary cards — Chapter ᚈ(T): The caboose

*Teclam Áirme, the Mathematical Compendium — a copy for AI readers of https://www.teclamairme.com/#group-T10, as of 2026-09-30. Links are relative to https://www.teclamairme.com/fortheaireaders/.*

[Guide for AI readers](index.md) · [The chapter](ch10.md)

*Linear algebra: determinant, trace, norm, eigenvalues, and the Kronecker delta. Each card: its symbol, type, one-line headline, the longer note, the reference entries it cites and the cards it links to.*

### the letter Tinne — `ᚈ` {#g-og-tinne}

*Ogham letter*

Ogham letter name meaning ‘a bar of metal’ or ‘an ingot’; the medieval glossators read it as the holly.

Tinne is not a tree name. On the evidence of the kennings it means ‘bar of metal, ingot’, and it is probably cognate with Old Irish tend, ‘strong’, or tind, ‘brilliant’. The three Old Irish word-oghams, the Bríatharogaim ascribed to Morann mac Moín, to Mac ind Óc, and to Cú Chulainn, give trian roith, “one of three parts of a wheel”; smiur gúaile, “marrow of charcoal”; and trian n-airm, “one of three parts of a weapon.” The arboreal tradition glosses cuileand, holly, and reaches it through the wheel: “a third of a wheel is tinne, that is, because holly is one of the three timbers of the chariot-wheel.” Modern tree lore and ogham divination descend from the medieval plant glosses by way of Robert Graves’s The White Goddess of 1948, which read the letter order as an ancient seasonal calendar of tree magic. Celtic scholarship rejects that reconstruction; the associations in circulation today are modern, not ancient.

Cites: [2.15](ch02-refs.md#r-2.15), [2.16](ch02-refs.md#r-2.16), [2.17](ch02-refs.md#r-2.17), [2.19](ch02-refs.md#r-2.19)

### determinant of A = [[1, 2], [2, 4]] — `det A` {#g-detA}

*Integer or rational*

Zero because the second row is twice the first; the matrix squashes the plane onto a line.

The determinant measures how a linear map scales area or volume. Seki Takakazu reached it in 1683 and Leibniz independently in a letter of 1693; Gauss coined its name in 1801 for something else, and Cauchy fixed the name on this meaning in 1812. For A with rows (1, 2) and (2, 4) it is 1·4 − 2·2 = 0: the second row is twice the first, so the plane is flattened onto the line through (1, 2), and A has no inverse. Integer entries give an integer determinant.

Cites: [10.20](ch10-refs.md#r-10.20), [10.3](ch10-refs.md#r-10.3), [10.4](ch10-refs.md#r-10.4)

See also: [identity, inverse, and transpose](ch10-cards.md#g-identity)

### trace of A — `tr A` {#g-trA}

*Integer or rational*

1 + 4 = 5.

The trace, the sum of the diagonal entries, is 1 + 4 = 5, and it equals the sum of the eigenvalues, 5 + 0. The name is a translation of the German Spur, which is why some older books write sp A. It is an integer because the entries are.

Cites: [10.4](ch10-refs.md#r-10.4), [10.6](ch10-refs.md#r-10.6)

### Frobenius norm of A — `‖A‖_{F}` {#g-frob}

*Integer or rational*

√(1 + 4 + 4 + 16) = 5.

The Frobenius norm, named for Ferdinand Georg Frobenius, is the square root of the sum of the squares of all entries: √(1 + 4 + 4 + 16) = √25 = 5. It also equals the square root of the sum of the squared singular values, and this matrix has one nonzero singular value, 5. The entries were chosen so that the norm is an integer.

Cites: [10.4](ch10-refs.md#r-10.4), [10.7](ch10-refs.md#r-10.7)

See also: [the largest singular value](ch10-cards.md#g-svd)

### largest eigenvalue of A — `λ_{max}(A)` {#g-lmax}

*Integer or rational*

5; the other eigenvalue is 0.

The eigenvalues of A are the numbers λ for which A stretches some direction by exactly λ; they are the roots of λ² − 5λ = 0, so 5 and 0. The direction (1, 2) is stretched by 5 and the perpendicular direction (2, −1) is sent to zero. The word eigen is Hilbert’s, from about 1904; the idea is older, going back to Euler and Lagrange on rotating bodies. Here the largest eigenvalue is the integer 5.

Cites: [10.4](ch10-refs.md#r-10.4), [10.8](ch10-refs.md#r-10.8)

### the Kronecker delta off the diagonal — `δ_{12}` {#g-kronecker}

*Integer or rational*

δᵢⱼ is 1 when i = j and 0 otherwise, so δ₁₂ = 0; this δ is not Feigenbaum’s.

Leopold Kronecker’s delta, from his work on bilinear forms of the 1860s, is δᵢⱼ = 1 when i = j and 0 otherwise. Off the diagonal it is 0, so δ₁₂ = 0, and the identity matrix has entries δᵢⱼ. It is an integer-valued symbol and is not Feigenbaum’s δ of Chapter ᚃ(F).

Cites: [10.5](ch10-refs.md#r-10.5)

See also: [Feigenbaum’s δ](ch05-cards.md#g-feig), [identity, inverse, and transpose](ch10-cards.md#g-identity)

### two perpendicular vectors — `(1, 2) · (2, −1)` {#g-dot0}

*Integer or rational*

The dot product 1·2 + 2·(−1) = 0: the two eigenvectors of A are at right angles, as a real symmetric matrix’s eigenvectors for distinct eigenvalues always are.

The dot product 𝐮 · 𝐯 = Σ uᵢvᵢ, given its dot by Gibbs in lecture notes privately printed in 1881, and reached independently by Heaviside, is zero for two nonzero vectors exactly when they are perpendicular; the name dot product arrived with Gibbs and Wilson’s Vector Analysis of 1901. Here (1, 2) and (2, −1) are the eigenvectors of A for the eigenvalues 5 and 0, and their dot product is 2 − 2 = 0, an instance of the spectral theorem: a real symmetric matrix has an orthonormal basis of eigenvectors, and eigenvectors belonging to distinct eigenvalues are always perpendicular. It is the same zero as the orthogonal states of Chapter ᚍ(K) in two real dimensions. An integer, hence the tag.

Cites: [10.16](ch10-refs.md#r-10.16), [10.8](ch10-refs.md#r-10.8)

See also: [the spectral theorem](ch10-cards.md#g-spectralthm)

### a nilpotent matrix squared — `N², N = [[0, 1], [0, 0]]` {#g-nilp}

*Integer or rational*

A nonzero matrix whose square is the zero matrix: N shifts the plane, and shifting twice sends everything to 0.

A matrix N is nilpotent when some power of it is the zero matrix; the smallest nonzero example is 2 × 2, N = [[0, 1], [0, 0]], with N² = 0. Its only eigenvalue is 0, its determinant and trace are 0, yet N itself is not 0: the algebra of matrices has zero divisors, unlike the reals, which is the same lesson the sedenions teach in Chapter ᚗ(V). Jordan’s normal form goes back to 1870, and with it every complex matrix is a diagonalizable part plus a nilpotent part that commutes with it, the Jordan–Chevalley decomposition. The zero matrix hands home the integer 0.

Cites: [10.9](ch10-refs.md#r-10.9)

See also: [the real numbers](ch01-cards.md#g-reals), [the sedenions](ch24-cards.md#g-sedenions)

### the rank–nullity theorem — `rank A + null A` {#g-ranknull}

*Integer or rational*

1 + 1 = 2: the dimension of what A keeps plus the dimension of what it kills is the dimension of the plane.

The rank of a matrix is the dimension of its image, the nullity the dimension of its kernel, and the rank–nullity theorem says they add to the number of columns. A squashes the plane onto a line, so its rank is 1, and it kills the line through (2, −1), so its nullity is 1; the sum is 2. Sylvester named the nullity in 1884 and the word rank is Frobenius’s, from 1878 or 1879. A dimension count, hence the integer tag.

Cites: [10.10](ch10-refs.md#r-10.10), [10.4](ch10-refs.md#r-10.4)

### the Cayley–Hamilton theorem for A — `A² − 5A` {#g-cayham}

*Integer or rational*

Every square matrix satisfies its own characteristic equation; for A that equation is λ² − 5λ = 0, so A² − 5A is the zero matrix.

The characteristic polynomial of A = [[1, 2], [2, 4]] is det(λI − A) = λ² − 5λ, and the Cayley–Hamilton theorem says a matrix is a root of its own characteristic polynomial: A² − 5A = 0, which one can check by hand since A² = [[5, 10], [10, 20]] = 5A. Hamilton proved a case for quaternions in 1853, Cayley stated it in 1858 and proved it only for 2 × 2 matrices, which is exactly this size, and Frobenius proved the general case in 1878. The value is the zero matrix, which hands home the integer 0.

Cites: [10.11](ch10-refs.md#r-10.11), [10.4](ch10-refs.md#r-10.4)

See also: [the characteristic polynomial](ch10-cards.md#g-charpoly), [the quaternions](ch22-cards.md#g-quaternions)

### identity, inverse, and transpose — `I, A⁻¹, Aᵀ` {#g-identity}

*Function or operator*

I has δᵢⱼ as its entries and changes nothing; A⁻¹ undoes A and exists only when det A ≠ 0, so this A has none; Aᵀ swaps rows and columns.

The identity matrix I has ones on the diagonal and zeros elsewhere, the Kronecker delta as its entries, and I·A = A for every A. The inverse A⁻¹ satisfies A·A⁻¹ = I and exists exactly when det A ≠ 0; the matrix of this chapter has determinant 0 and so no inverse, which is what squashing the plane onto a line means. The transpose Aᵀ swaps rows and columns; A here is symmetric, Aᵀ = A, which is why its eigenvectors are perpendicular. Cayley published the first matrix inverse in 1853 and the first abstract definition of a matrix, with its algebra, in 1858; the word matrix itself is Sylvester’s, from 1850. Tools, hence the tag.

Cites: [10.12](ch10-refs.md#r-10.12), [10.2](ch10-refs.md#r-10.2), [10.21](ch10-refs.md#r-10.21), [10.3](ch10-refs.md#r-10.3)

See also: [determinant of A = [[1, 2], [2, 4]]](ch10-cards.md#g-detA)

### the largest singular value — `σ₁(A)` {#g-svd}

*Integer or rational*

5, the same as the Frobenius norm here, because A has only one nonzero singular value; singular values measure how far any matrix stretches, eigenvalues how far a square one stretches the directions it keeps on their own line.

The singular values of a matrix are the square roots of the eigenvalues of AᵀA, and the singular value decomposition A = UΣVᵀ, developed by Beltrami and Jordan in 1873 and 1874 and made numerical by Golub and Kahan in 1965, factors any matrix into a rotation, a stretch, and a rotation. A has rank 1, so one nonzero singular value, σ₁ = 5, and the Frobenius norm √(Σσᵢ²) is the same 5. The SVD is used in least squares, principal components, and image compression. An integer here, hence the tag.

Cites: [10.13](ch10-refs.md#r-10.13), [10.8](ch10-refs.md#r-10.8)

See also: [Frobenius norm of A](ch10-cards.md#g-frob), [least squares](ch10-cards.md#g-leastsq)

### Cramer’s rule — `x = det Aᵢ / det A` {#g-cramer}

*Function or operator*

Solve a square system by determinants; it fails for this A because det A = 0, which is exactly when the system has no unique solution.

Colin Maclaurin was teaching the rule to his students by ~1729, and Gabriel Cramer published it in 1750: the i-th unknown of a system A𝐱 = 𝐛 is det Aᵢ divided by det A, where Aᵢ has its i-th column replaced by 𝐛. The rule divides by det A, so it breaks down exactly when the determinant is zero, which is also when the system has no unique solution: the system of this chapter’s A either has no solution or a whole line of them. That division is the reason for the rule of this book that a denominator must be provably nonzero. A rule, hence the tool tag.

Cites: [10.14](ch10-refs.md#r-10.14), [10.23](ch10-refs.md#r-10.23), [10.3](ch10-refs.md#r-10.3)

See also: [determinant of A = [[1, 2], [2, 4]]](ch10-cards.md#g-detA)

### vector spaces, bases, and dimension — `dim V, basis` {#g-vspace}

*Function or operator*

A vector space is anything you can add and scale; a basis is a smallest spanning set, and every basis has the same size, the dimension.

Peano gave the axioms of a vector space in 1888, following Grassmann’s 1844 Ausdehnungslehre (Chapter ᚌ(G)), and the theorem that every basis of a space has the same number of elements is what makes dimension well defined; ℝ² has dimension 2, which is the 2 of the rank–nullity bracket, and the space of all polynomials has countably infinite dimension. Linear independence, span, subspaces, and linear maps are the vocabulary; matrices are what linear maps look like once a basis is chosen. Concepts, hence the tool tag.

Cites: [10.15](ch10-refs.md#r-10.15)

### tensor and cross product — `𝐮 ⊗ 𝐯, 𝐮 × 𝐯` {#g-tensor}

*Function or operator*

The tensor product builds a matrix from two vectors, and A itself is (1, 2) ⊗ (1, 2); the cross product of two vectors in space is a vector perpendicular to both whose length is the area of their parallelogram, so it is 0 for parallel vectors.

The outer or tensor product 𝐮 ⊗ 𝐯 is the matrix with entries uᵢvⱼ; every rank-one matrix is one, and A = (1, 2) ⊗ (1, 2), which is why A squashes the plane. The cross product 𝐮 × 𝐯, Gibbs’s and Heaviside’s from the 1880s and Grassmann’s wedge in disguise (Chapter ᚌ(G)), is defined in three dimensions and vanishes for parallel vectors; (1, 2, 0) × (2, 4, 0) = 0 is the same collapse as det A = 0. Tensors of higher rank, Ricci and Levi-Civita’s 1900 calculus, are the language of general relativity. Operations, hence the tool tag.

Cites: [10.16](ch10-refs.md#r-10.16)

See also: [the seven-dimensional cross product](ch23-cards.md#g-cross7), [determinant of A = [[1, 2], [2, 4]]](ch10-cards.md#g-detA)

### the characteristic polynomial — `p_A(λ) = det(λI − A)` {#g-charpoly}

*Function or operator*

The polynomial whose roots are the eigenvalues; for A it is λ² − 5λ = λ(λ − 5), so the eigenvalues are 0 and 5 and the trace 5 is the sum of the roots.

The characteristic polynomial p_A(λ) = det(λI − A) has the eigenvalues as roots; its coefficients are, up to sign, the trace and the determinant for a 2 × 2 matrix, so p_A(λ) = λ² − (tr A)λ + det A = λ² − 5λ + 0 here, which is why the trace 5 and the determinant 0 of this chapter reappear as the sum and product of the eigenvalues. Cauchy used the polynomial in 1829 and named the characteristic equation in 1840. A polynomial, keyed as a tool.

Cites: [10.17](ch10-refs.md#r-10.17), [10.4](ch10-refs.md#r-10.4), [10.6](ch10-refs.md#r-10.6)

See also: [determinant of A = [[1, 2], [2, 4]]](ch10-cards.md#g-detA)

### maps, arrows, and composition — `f: V → W, x ↦ Ax, g ∘ f` {#g-maps}

*Function or operator*

→ names a map between spaces, ↦ says where one element goes, and ∘ composes two maps; A ∘ A is A², and matrix multiplication is composition.

A linear map is written f: V → W, and x ↦ Ax says what it does to a vector. Composition g ∘ f means do f, then g, and the reason matrix multiplication is defined as it is (rows times columns) is that the product matrix is the matrix of the composed map. Here A ∘ A = A² = 5A by the Cayley–Hamilton entry. Notation, hence the tool tag.

Cites: [10.18](ch10-refs.md#r-10.18)

See also: [the Cayley–Hamilton theorem for A](ch10-cards.md#g-cayham)

### the Nine Chapters on the Mathematical Art — `Rectangular Arrays` {#g-ninechapters}

*Function or operator*

The Chinese classic whose eighth chapter solves systems of simultaneous linear equations by reducing an array of coefficients with column operations: Gaussian elimination, two thousand years before Gauss.

Chapter 8, Rectangular Arrays, poses eighteen problems that reduce to systems of simultaneous linear equations, and solves them by writing the coefficients as an augmented array and reducing it to triangular form with elementary column operations—exactly the method now called Gaussian elimination. Most historians date the text to about 200 BC, with 206 BC to 50 AD covering its different parts, so the method predates the name by two millennia. Gauss met a six-by-six system in astronomy: from observations of the asteroid Pallas taken between 1803 and 1809 he obtained six linear equations in six unknowns, and in 1810 he devised the notation for symmetric elimination that nineteenth-century hand computers used on least-squares normal equations. It is the oldest ancestor of everything else in this chapter: the grid of coefficients that Sylvester would later name a matrix.

Cites: [10.1](ch10-refs.md#r-10.1), [10.22](ch10-refs.md#r-10.22)

See also: [Gaussian elimination](ch10-cards.md#g-gausselim)

### orthogonal projection — `P² = P = Pᵀ` {#g-orthproj}

*Function or operator*

Drop each vector straight down onto a subspace. Doing it twice is the same as doing it once, and what is discarded is perpendicular to what is kept.

A projection is a linear map with P² = P—once you have landed on the subspace, landing again changes nothing—and it is orthogonal when the part thrown away is perpendicular to the part kept, equivalently when the matrix is also symmetric. That perpendicularity is what makes the projection the nearest point of the subspace, which is why least squares is a projection and not merely an analogy. This chapter’s matrix is five times such a map: A/5 sends the plane onto the line y = 2x and leaves nothing of the perpendicular direction, which is the whole reason its determinant is 0, its rank is 1, and its second eigenvalue is 0. A tool of the framework, uncolored.

Cites: [10.24](ch10-refs.md#r-10.24)

See also: [least squares](ch10-cards.md#g-leastsq)

### the spectral theorem — `A = QΛQᵀ` {#g-spectralthm}

*Function or operator*

A real symmetric matrix has real eigenvalues and an orthonormal basis of eigenvectors, so it is nothing but a set of stretches along perpendicular axes.

The spectral theorem says that a real symmetric matrix can be written QΛQᵀ with Λ diagonal and Q orthogonal: its eigenvalues are real, its eigenvectors for distinct eigenvalues are perpendicular, and it acts by stretching along a set of perpendicular axes and nothing else. Cauchy proved it for symmetric matrices, and von Neumann carried it to operators on a Hilbert space, where it became perhaps the most important result of operator theory, and where quantum mechanics uses it to attach probabilities to the values of a measurement. The chapter’s A is symmetric, which is exactly why its eigenvectors (1, 2) and (2, −1) meet at right angles and why the sentence about perpendicular eigenvectors is a theorem rather than a coincidence. A theorem, uncolored.

Cites: [10.25](ch10-refs.md#r-10.25)

See also: [Hilbert space](ch15-cards.md#g-hilbert)

### Gaussian elimination — `Ax = b → Ux = c` {#g-gausselim}

*Function or operator*

Reduce a system of linear equations to triangular form by row or column operations, then solve it by back-substitution.

The method appears in the eighth chapter of the Chinese Nine Chapters on the Mathematical Art, compiled over generations up to the first century AD, which solves systems of simultaneous equations by reducing an array of coefficients with column operations. Carl Friedrich Gauss used it in 1810 to solve six simultaneous equations arising from observations of the asteroid Pallas, devising a symmetric-elimination notation that nineteenth-century hand computers used for least-squares problems; his name attached to the method even though it predates him by nearly two thousand years. It underlies the row reduction used to compute rank, solve linear systems, and invert matrices.

Cites: [10.27](ch10-refs.md#r-10.27)

See also: [the Nine Chapters on the Mathematical Art](ch10-cards.md#g-ninechapters)

### least squares — `min‖Ax−b‖²` {#g-leastsq}

*Function or operator*

Least squares chooses the solution that minimizes the sum of the squared errors, the standard way to fit a line, curve, or overdetermined system when no exact solution exists.

For an inconsistent system Ax = b, the least-squares solution minimizes ‖Ax − b‖², and it is characterized by the normal equations AᵀAx = Aᵀb, exactly the condition that the residual be orthogonal to the column space of A, the same orthogonality that orthogonal projection formalizes. Adrien-Marie Legendre published the method in 1805 and Carl Friedrich Gauss claimed to have used it since 1795; historians have largely accepted Gauss’s claim, though Legendre keeps priority of publication. The singular value decomposition computes it more stably than the normal equations, as orthogonal methods do, because it avoids forming AᵀA.

Cites: [10.30](ch10-refs.md#r-10.30)

See also: [orthogonal projection](ch10-cards.md#g-orthproj)

### zero divisor — `ab=0, a,b≠0` {#g-zerodiv}

*Non-real object*

A zero divisor is a ring element that multiplies with some nonzero element to give zero; a nonzero one breaks the rule, true in all the ordinary number systems, that a product is zero only when one of its factors is.

In the integers, the rationals, the reals, the complex numbers, the quaternions, and the octonions, ab = 0 forces a = 0 or b = 0; rings and algebras with zero divisors, such as the integers modulo 6, where 2 × 3 = 0, or the sedenions, lose that guarantee along with the ability to divide safely. Their presence is what keeps a ring from being an integral domain, and in the Cayley–Dickson ladder they first appear at the sedenions, the sixteen-dimensional algebra after the octonions, and persist in every algebra above them, as in the trigintaduonions. A nonzero matrix whose square is zero, such as the nilpotent N of Chapter ᚈ(T), is itself a zero divisor: multiplied by itself, it gives nothing.

Cites: [10.33](ch10-refs.md#r-10.33)

See also: [the integers](ch01-cards.md#g-ints), [a nilpotent matrix squared](ch10-cards.md#g-nilp), [the octonions](ch23-cards.md#g-octonions), [the quaternions](ch22-cards.md#g-quaternions), [the real numbers](ch01-cards.md#g-reals), [the sedenions](ch24-cards.md#g-sedenions), [the trigintaduonions](ch24-cards.md#g-trig)