The Algebra Generated by an Element and the Chinese Remainder Decomposition
Introduction
This article studies the smallest subalgebra of an associative algebra that contains a given element. Everything that can be said about one element without looking at the rest of the algebra is already contained in that subalgebra: its dimension, its idempotents, the way it splits into components, and the minimal polynomial that governs the element. The central result is that the subalgebra is a quotient of the polynomial algebra by the minimal polynomial, and that the Chinese remainder theorem splits it whenever the minimal polynomial factors into coprime pieces.
The article assumes the definitions of Algebras, Associative Algebras and Unital Algebras, the polynomial algebra of Polynomial Algebras, and the first isomorphism theorem of Ideals and Quotients of Algebras. The base is a field $k$ throughout, except in §The Case of a Commutative Ring; the algebra $A$ is associative, unital and finite-dimensional over $k$. The eigenvalue decomposition is the special case in which every factor of the minimal polynomial is linear, and it is the subject of The Eigenvalue Decomposition of an Element. The Peirce decomposition of $A$ by the idempotents produced here is The Centraliser and the Diagonal Peirce Part and Off-Diagonal Peirce Components and Noncentral Idempotents; the nilpotent part that survives when the factors are not coprime is Nilpotent Elements and the Jordan–Chevalley Decomposition.
The Subalgebra Generated by an Element
Definition
Let $A$ be a unital associative $k$-algebra and let $a \in A$. The algebra generated by $a$ is the intersection of all unital subalgebras of $A$ that contain $a$; it is written $k[a]$. Equivalently, it is the set of all polynomial expressions in $a$,
$$ k[a] = \{ f(a) : f \in k[x] \}, $$
with the convention that the powers are the powers in $A$ and that the constant polynomial $1$ gives the identity $1_A$. The equivalence is immediate: the set of polynomial expressions is a subalgebra containing $a$, and every subalgebra containing $a$ contains all its powers and hence all its polynomial expressions.
The algebra $k[a]$ is commutative, even when $A$ is not, because it is generated by the single element $a$ and its powers commute with one another. It contains $1_A$, it is finite-dimensional over $k$ because $A$ is, and it is the image of the evaluation map of the next subsection. If $A$ is not unital, one applies the construction to the unitisation $A^+$ of Unital Algebras, §The Unitisation, or works with the non-unital algebra of polynomials with zero constant term; the unital case is the one used in this article.
The evaluation map
The evaluation map at $a$ is
$$ \varphi_a : k[x] \longrightarrow A, \qquad \varphi_a(f) = f(a). $$
It is a homomorphism of unital $k$-algebras: the sum and the product of two polynomials are sent to the sum and the product of their values, and the constant polynomial $1$ is sent to $1_A$. Its image is $k[a]$ by definition, so the first isomorphism theorem gives an isomorphism of $k$-algebras
$$ k[a] \;\cong\; k[x]/\ker \varphi_a . $$
The kernel is an ideal of $k[x]$. Since $k$ is a field, $k[x]$ is a principal ideal domain, so the kernel is generated by one polynomial. If $A \neq 0$, the kernel is nonzero: the powers $1, a, a^2, \dots$ cannot all be linearly independent in the finite-dimensional space $A$, so some nonzero polynomial vanishes at $a$. The kernel is then generated by a unique monic polynomial, the minimal polynomial of $a$, written $m_a$. Thus
$$ \boxed{\;k[a] \;\cong\; k[x]/(m_a)\;}. $$
The isomorphism is the one induced by $\varphi_a$, and it sends the class of $x$ to $a$.
The minimal polynomial
The minimal polynomial $m_a$ is the monic polynomial of least degree with $m_a(a) = 0$. It is unique, it generates the kernel of $\varphi_a$, and it divides every polynomial $f$ with $f(a) = 0$. Its degree is the dimension of $k[a]$:
$$ \dim_k k[a] = \deg m_a . $$
Indeed, the classes of $1, x, \dots, x^{d-1}$ with $d = \deg m_a$ form a basis of $k[x]/(m_a)$: they span because every polynomial can be divided by $m_a$, and they are independent because a relation of degree less than $d$ would give a nonzero polynomial of degree less than $d$ vanishing at $a$, against the minimality of $m_a$. Consequently
$$ 1_A,\ a,\ a^2,\ \dots,\ a^{d-1} $$
is a basis of $k[a]$, and every higher power is reduced to this basis by the relation $m_a(a) = 0$. If
$$ m_a(x) = x^d + c_{d-1}x^{d-1} + \cdots + c_1 x + c_0, $$
then
$$ a^d = -c_{d-1}a^{d-1} - \cdots - c_1 a - c_0 1_A . $$
The degree is bounded by $\dim_k A$, since $k[a] \subseteq A$. For $A = M_n(k)$, the Cayley–Hamilton theorem gives the stronger bound $\deg m_a \le n$; for a general algebra there is no characteristic polynomial and no such bound beyond $\dim_k A$. The minimal polynomial of an element of a general algebra is therefore a genuinely algebraic object and not a matrix invariant in disguise.
Examples
- The scalar algebra. For $A = k$ and $a = \lambda$, the minimal polynomial is $x - \lambda$ and $k[a] = k$.
- The polynomial quotient. For $A = k[x]/(p)$ with $p$ monic and $a$ the class of $x$, the minimal polynomial is $p$ and $k[a] = A$; the dimension is $\deg p$.
- The diagonal matrix. For $A = M_2(k)$ and $a = \operatorname{diag}(1,2)$, the minimal polynomial is $(x-1)(x-2)$ and $k[a]$ is the algebra of diagonal matrices, of dimension two. The identity and $a$ already form a basis.
- A field inside a matrix algebra. For $A = M_2(\mathbb{R})$ and $a$ the rotation matrix of angle $\pi/2$,
$$ a = \begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix}, $$
one has $a^2 = -I$, so $m_a = x^2 + 1$ and $k[a] \cong \mathbb{C}$. The algebra generated by a real matrix can be a field extension of $\mathbb{R}$, and then it has no idempotent other than $0$ and $1$. - The companion matrix. For $A = M_3(\mathbb{Q})$ and
$$ C = \begin{pmatrix} 0 & 0 & 0 \\ 1 & 0 & 1 \\ 0 & 1 & 0 \end{pmatrix}, $$
one has $C^2 = \begin{pmatrix} 0 & 0 & 0 \\ 0 & 1 & 0 \\ 1 & 0 & 1 \end{pmatrix}$ and $C^3 = C$, so $m_C = x^3 - x$ and $\mathbb{Q}[C]$ has dimension three. The matrix is the companion matrix of its own minimal polynomial. - The dual numbers. For $A = \mathbb{R}[\varepsilon]/(\varepsilon^2)$ and $a = \varepsilon$, the minimal polynomial is $x^2$ and $k[a] = A$, of dimension two. The element is nonzero and nilpotent, and the algebra is local. - The Jordan block. For $A = M_2(k)$ and $a = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix}$, the minimal polynomial is $(x-\lambda)^2$, and $k[a] = \{\alpha I + \beta (a - \lambda I)\}$ is the two-dimensional algebra of upper triangular Toeplitz matrices with equal diagonal entries.
What the subalgebra sees
The algebra $k[a]$ sees the powers of $a$ and nothing else. Two elements with the same minimal polynomial have isomorphic subalgebras, whatever the ambient algebras; the isomorphism is the one sending $a$ to the other element, and it is well defined precisely because the minimal polynomials agree. The embedding $k[a] \subseteq A$ need not be an ideal, need not be a direct summand, and need not be central. The dimension $\deg m_a$ can be much smaller than $\dim_k A$: a scalar matrix in $M_n(k)$ generates the one-dimensional algebra $k I$, and a cyclic matrix generates an $n$-dimensional commutative subalgebra of the $n^2$-dimensional matrix algebra.
The converse question, how much of $A$ is determined by $a$, is the subject of the centraliser: for a cyclic matrix the answer is that $k[a]$ is the whole centraliser, while for a scalar matrix the centraliser is all of $A$. This is The Centraliser and the Diagonal Peirce Part.
The Chinese Remainder Decomposition
Coprime factors
Two polynomials $m_1, m_2 \in k[x]$ are coprime if $\gcd(m_1,m_2) = 1$, equivalently if there exist $u_1, u_2 \in k[x]$ with $u_1 m_1 + u_2 m_2 = 1$. A family $m_1, \dots, m_r$ is pairwise coprime if every pair is. For a monic polynomial $m$, a factorisation
$$ m = m_1 m_2 \cdots m_r $$
into pairwise coprime monic factors is unique up to the order of the factors once the factors are required to be powers of distinct irreducibles; the general statement used below does not need uniqueness.
The theorem
Theorem (Chinese remainder theorem for polynomials). Let $m = m_1 \cdots m_r$ with the $m_i$ pairwise coprime and monic. Then the map
$$ \boxed{\;k[x]/(m) \longrightarrow \prod_{i=1}^{r} k[x]/(m_i), \qquad f \bmod m \longmapsto (f \bmod m_1, \dots, f \bmod m_r)\;} $$
is an isomorphism of $k$-algebras. In particular,
$$ \deg m = \sum_{i=1}^{r} \deg m_i . $$
Proof. The map is well defined because $m_i$ divides $m$, and it is a $k$-algebra homomorphism because reduction modulo each $m_i$ is one. It is injective: if $f \equiv 0 \pmod{m_i}$ for every $i$, then every $m_i$ divides $f$, and the pairwise coprimality gives $m \mid f$. It is surjective: given residues $f_i \bmod m_i$, one constructs $f$ by Bezout. For each $i$, the coprimality of $m_i$ and $m/m_i$ gives $u_i m_i + v_i (m/m_i) = 1$; the polynomial $g_i = v_i (m/m_i)$ satisfies $g_i \equiv 1 \pmod{m_i}$ and $g_i \equiv 0 \pmod{m_j}$ for $j \neq i$. Then $f = \sum_i f_i g_i$ has the required residues. The dimension statement follows from the isomorphism and from $\dim k[x]/(m) = \deg m$.
The idempotents
The polynomials $g_i$ of the proof are the canonical idempotents of the factorisation. Reduced modulo $m$, they satisfy
$$ g_i^2 \equiv g_i, \qquad g_i g_j \equiv 0 \ (i \neq j), \qquad \sum_{i=1}^{r} g_i \equiv 1 \pmod m . $$
Their images in $k[a]$,
$$ \varepsilon_i = g_i(a), $$
are therefore idempotents of $k[a]$ that are pairwise orthogonal and sum to $1_A$. They give the internal form of the Chinese remainder decomposition,
$$ \boxed{\;k[a] = k[a]\varepsilon_1 \oplus \cdots \oplus k[a]\varepsilon_r\;}, $$
a direct sum of $k$-algebras, where the $i$-th summand has identity $\varepsilon_i$ and is isomorphic to $k[x]/(m_i)$ by $f(a)\varepsilon_i \mapsto f \bmod m_i$. The external form of the theorem is the product of the same quotients. The two forms carry the same information: the projections of the product correspond to multiplication by the $\varepsilon_i$.
The element $a$ decomposes accordingly. Put $a_i = a \varepsilon_i$. Then
$$ a = a_1 + \cdots + a_r, \qquad a_i a_j = 0 \ (i \neq j), $$
and $a_i$ is the image of the class of $x$ in the $i$-th factor, so its minimal polynomial is $m_i$. The study of $a$ is thereby reduced to the study of the $r$ elements $a_i$, each governed by one factor. This reduction is the algebraic content of the decomposition: it separates the eigenvalues, the field extensions and the nilpotent blocks according to the irreducible factors of $m_a$.
The two descriptions
There are two ways to read the same decomposition, and they are worth keeping apart.
- As a product of quotients. $k[a] \cong \prod_i k[x]/(m_i)$. This description is useful when the factors are fields or local algebras and one wants to compute componentwise.
- As a sum of ideals. $k[a] = \bigoplus_i k[a]\varepsilon_i$, with $k[a]\varepsilon_i \cong k[x]/(m_i)$. This description is useful inside $A$, because the $\varepsilon_i$ are actual elements of $A$ and can be multiplied with arbitrary elements of $A$.
The second description is internal to $k[a]$, not to $A$. The idempotents $\varepsilon_i$ need not be central in $A$, so $A$ itself does not split as a product. The Peirce decomposition of $A$ by the $\varepsilon_i$ is the correct substitute, and it is the subject of The Centraliser and the Diagonal Peirce Part and Off-Diagonal Peirce Components and Noncentral Idempotents. If $a$ is central in $A$, then every $\varepsilon_i$ is central, and $A$ splits as $\prod_i A\varepsilon_i$.
The primary decomposition
The most important factorisation is into powers of distinct irreducible polynomials,
$$ m_a = f_1^{e_1} \cdots f_r^{e_r}, \qquad f_i \text{ irreducible and distinct}. $$
The factors $f_i^{e_i}$ are pairwise coprime, so the theorem applies. The factors $k[x]/(f_i^{e_i})$ are local algebras with residue field $k[x]/(f_i)$; the factor is a field exactly when $e_i = 1$. This is the primary decomposition of $k[a]$. The idempotents $\varepsilon_i$ are the generalized eigenprojectors when the $f_i$ are linear, and the block projections in general. The exponents $e_i$ measure the nilpotent thickness of the components, and $\max_i e_i$ is the nilpotency index of the nilpotent part of $a$ in the Jordan–Chevalley decomposition of Nilpotent Elements and the Jordan–Chevalley Decomposition.
The Two Extreme Cases
An irreducible minimal polynomial
If $m_a$ is irreducible, then $k[a] \cong k[x]/(m_a)$ is a field, and its only idempotents are $0$ and $1_A$. There is no nontrivial Chinese remainder decomposition, because there is only one factor. The element $a$ behaves like a primitive element of a field extension: it generates a field inside $A$, and the rest of $A$ is a module over that field. The rotation matrix of §Examples is the model case, with $k = \mathbb{R}$ and $k[a] \cong \mathbb{C}$. The same phenomenon occurs for the companion matrix of an irreducible polynomial over $\mathbb{Q}$, which generates a number field inside a matrix algebra.
An irreducible minimal polynomial does not mean that $a$ has no eigenvalues: it means that $a$ has no eigenvalues in $k$. Over the algebraic closure the polynomial splits, and the eigenvalues appear there, but the idempotents that separate them lie in $A \otimes_k \bar k$ and not in $A$ in general. The passage to the algebraic closure is the subject of Splitting Fields and Algebraic Closure; the distinction between separable and inseparable irreducible polynomials is the subject of Separable Elements and Split Minimal Polynomials.
A power of a linear polynomial
If $m_a = (x-\lambda)^d$, then $k[a] \cong k[x]/((x-\lambda)^d)$ is a local algebra with maximal ideal $(a-\lambda)$. Its only idempotents are $0$ and $1$, so again there is no nontrivial decomposition. The element is the sum of the scalar $\lambda$ and the nilpotent $a - \lambda$, and the algebra is the smallest example of a non-reduced algebra:
$$ k[a] = k \oplus k(a-\lambda) \oplus \cdots \oplus k(a-\lambda)^{d-1}, \qquad (a-\lambda)^d = 0 . $$
The dual numbers are the case $d = 2$, and the Jordan block is the matrix case. The decomposition of $a$ into a scalar and a nilpotent is the simplest Jordan–Chevalley decomposition; the general case is Nilpotent Elements and the Jordan–Chevalley Decomposition.
A squarefree polynomial
If $m_a$ is squarefree, then $k[a]$ is reduced, and a finite-dimensional reduced commutative $k$-algebra is a product of fields. The factors are the quotients $k[x]/(f_i)$ for the distinct irreducible factors $f_i$ of $m_a$. Two subcases are worth naming.
- Split squarefree. If every $f_i$ is linear, then $m_a = \prod_{i=1}^{r}(x-\lambda_i)$ with the $\lambda_i$ distinct in $k$, and $k[a] \cong k^r$. The idempotents $\varepsilon_i$ are the Lagrange idempotents, and $a = \sum_i \lambda_i \varepsilon_i$. This is the eigenvalue decomposition of The Eigenvalue Decomposition of an Element.
- Not split squarefree. If some $f_i$ has degree greater than one, the corresponding factor is a field extension of $k$, and $k[a]$ is a product of fields, one of which is larger than $k$. The element has eigenvalues only in the extension, not in $k$, and the decomposition over $k$ separates the irreducible factors without separating the roots inside a factor.
The squarefree condition is exactly the condition that $k[a]$ be reduced, and the split condition is exactly the presence of all eigenvalues in the base field. The two conditions together are the hypothesis of the eigenvalue decomposition.
Worked Examples
The companion matrix of $x^3 - x$
Let $A = M_3(\mathbb{Q})$ and let
$$ C = \begin{pmatrix} 0 & 0 & 0 \\ 1 & 0 & 1 \\ 0 & 1 & 0 \end{pmatrix}. $$
The minimal polynomial is $m_C = x^3 - x = x(x-1)(x+1)$. The three factors are linear and pairwise coprime, so the Chinese remainder theorem gives $\mathbb{Q}[C] \cong \mathbb{Q} \times \mathbb{Q} \times \mathbb{Q}$, of dimension three. The canonical idempotents are obtained from the Lagrange interpolation formula of The Lagrange–Sylvester Eigenprojector Formula:
$$ \varepsilon_0 = I - C^2 = \begin{pmatrix} 1 & 0 & 0 \\ 0 & 0 & 0 \\ -1 & 0 & 0 \end{pmatrix}, $$
$$ \varepsilon_1 = \frac{C^2 + C}{2} = \begin{pmatrix} 0 & 0 & 0 \\ \tfrac12 & \tfrac12 & \tfrac12 \\ \tfrac12 & \tfrac12 & \tfrac12 \end{pmatrix}, $$
$$ \varepsilon_{-1} = \frac{C^2 - C}{2} = \begin{pmatrix} 0 & 0 & 0 \\ -\tfrac12 & \tfrac12 & -\tfrac12 \\ \tfrac12 & -\tfrac12 & \tfrac12 \end{pmatrix}. $$
A direct computation, or the theorem, gives
$$ \varepsilon_0^2 = \varepsilon_0, \quad \varepsilon_1^2 = \varepsilon_1, \quad \varepsilon_{-1}^2 = \varepsilon_{-1}, \qquad \varepsilon_i \varepsilon_j = 0 \ (i \neq j), \qquad \varepsilon_0 + \varepsilon_1 + \varepsilon_{-1} = I, $$
and
$$ C\varepsilon_0 = 0, \qquad C\varepsilon_1 = \varepsilon_1, \qquad C\varepsilon_{-1} = -\varepsilon_{-1}. $$
Thus $C = 0\cdot\varepsilon_0 + 1\cdot\varepsilon_1 + (-1)\cdot\varepsilon_{-1}$, which is the eigenvalue decomposition of $C$ with eigenvalues $0, 1, -1$. The two decompositions, the Chinese remainder one and the eigenvalue one, coincide here because all factors are linear and distinct.
The idempotents are not central in $M_3(\mathbb{Q})$. For instance, $\varepsilon_0 E_{12} \neq E_{12}\varepsilon_0$: the left product is supported on the second column and the right product vanishes. Consequently $M_3(\mathbb{Q})$ does not split as a product of three matrix algebras, even though $\mathbb{Q}[C]$ splits as a product of three copies of $\mathbb{Q}$. Only the subalgebra generated by $C$ splits; the ambient algebra keeps its Peirce decomposition with nonzero off-diagonal parts.
The split-complex and the dual numbers
The split-complex algebra $\mathbb{C}_{\mathrm{spl}} = \mathbb{R}[j]/(j^2-1)$ is $\mathbb{R}[x]/(x^2-1)$ with $j$ the class of $x$. The minimal polynomial is $(x-1)(x+1)$, so
$$ \mathbb{C}_{\mathrm{spl}} \cong \mathbb{R} \times \mathbb{R}, \qquad \Pi_1 = \frac{1+j}{2}, \quad \Pi_2 = \frac{1-j}{2}, $$
with $\Pi_1 + \Pi_2 = 1$ and $\Pi_1\Pi_2 = 0$. Since $\mathbb{C}_{\mathrm{spl}}$ is commutative, the idempotents are central, and the Chinese remainder decomposition is a decomposition of the whole algebra. This is the smallest case in which the decomposition is visible as a product of algebras.
The dual numbers $\mathbb{C}_{\mathrm{dln}} = \mathbb{R}[\varepsilon]/(\varepsilon^2)$ have minimal polynomial $x^2$, a single factor with exponent two. There is no nontrivial idempotent and no product decomposition; the algebra is local. The contrast between $\mathbb{C}_{\mathrm{spl}}$ and $\mathbb{C}_{\mathrm{dln}}$ is the contrast between a split squarefree polynomial and a repeated factor, and it is the reason the eigenvalue decomposition needs distinct roots while the Jordan–Chevalley decomposition does not.
The complex numbers as a real algebra
The complex numbers $\mathbb{C} = \mathbb{R}[i]/(i^2+1)$ have minimal polynomial $x^2+1$, irreducible over $\mathbb{R}$. The algebra is a field, so it has no nontrivial idempotents and no Chinese remainder decomposition over $\mathbb{R}$. Over $\mathbb{C}$ the polynomial splits as $(x-i)(x+i)$, and the algebra $\mathbb{C}[x]/(x^2+1)$ is isomorphic to $\mathbb{C} \times \mathbb{C}$, with idempotents
$$ \frac{1 + ix}{2}, \qquad \frac{1 - ix}{2}. $$
They involve $i$ and do not lie in $\mathbb{R}[x]/(x^2+1)$. The example shows that the Chinese remainder decomposition depends on the base field: the same polynomial is irreducible over $\mathbb{R}$ and split over $\mathbb{C}$.
The Case of a Commutative Ring
Over a commutative ring $R$ in place of the field $k$, several statements require care. The polynomial ring $R[x]$ is not a principal ideal domain, so the kernel of the evaluation map $R[x] \to A$ need not be principal, and a minimal polynomial in the strict sense need not exist. What survives is the Chinese remainder theorem for comaximal ideals: if $m = m_1 \cdots m_r$ with the ideals $(m_i)$ pairwise comaximal, then
$$ R[x]/(m) \cong \prod_i R[x]/(m_i), $$
with the same Bezout construction of the idempotents. The idempotents exist as soon as the factors are comaximal, and the internal decomposition $R[a] = \bigoplus_i R[a]\varepsilon_i$ holds. The eigenvalue decomposition needs more: the differences $\lambda_i - \lambda_j$ must be units of $R$, not merely nonzero, because the Lagrange formula divides by them. This is the same distinction that separates a field from a general ring in Eigenvalues and Diagonalisation and in The Lagrange–Sylvester Eigenprojector Formula. The polynomial algebra and its ideals are treated in Polynomial Algebras.
Summary
The algebra generated by an element $a$ of a finite-dimensional unital associative $k$-algebra is the commutative subalgebra $k[a] = \{f(a) : f \in k[x]\}$. Evaluation at $a$ gives an isomorphism $k[a] \cong k[x]/(m_a)$, where $m_a$ is the monic minimal polynomial; the dimension is $\deg m_a$, and $1, a, \dots, a^{d-1}$ is a basis. The minimal polynomial divides every polynomial that vanishes at $a$, and for a matrix algebra its degree is at most the matrix size, by Cayley–Hamilton.
If $m_a = m_1 \cdots m_r$ with the $m_i$ pairwise coprime, the Chinese remainder theorem gives $k[a] \cong \prod_i k[x]/(m_i)$ and, internally, $k[a] = \bigoplus_i k[a]\varepsilon_i$ with canonical idempotents $\varepsilon_i$ satisfying $\varepsilon_i^2 = \varepsilon_i$, $\varepsilon_i\varepsilon_j = 0$ for $i \neq j$, and $\sum_i \varepsilon_i = 1$. The element splits as $a = \sum_i a_i$ with $a_i = a\varepsilon_i$ and minimal polynomial $m_i$. The decomposition is internal to $k[a]$; it is a decomposition of $A$ when the idempotents $\varepsilon_i$ are central, which is automatic when $a$ is central.
The two extreme cases are the irreducible minimal polynomial, where $k[a]$ is a field with no nontrivial idempotents, and a power of a linear polynomial, where $k[a]$ is local with no nontrivial idempotents and $a$ is a scalar plus a nilpotent. The squarefree case gives a product of fields, and the split squarefree case is the eigenvalue decomposition. The companion matrix of $x^3-x$ exhibits the full split squarefree case with three rational idempotents that are not central in $M_3(\mathbb{Q})$.
Summary of Notation
| Symbol | Meaning |
|---|---|
| $k$ | A field, the base of the article |
| $A$ | A finite-dimensional unital associative $k$-algebra |
| $a$ | An element of $A$ |
| $k[a]$ | The unital subalgebra generated by $a$, equal to $\{f(a) : f \in k[x]\}$ |
| $\varphi_a$ | The evaluation map $k[x] \to A$, $f \mapsto f(a)$ |
| $m_a$ | The minimal polynomial of $a$, the monic generator of $\ker \varphi_a$ |
| $d = \deg m_a$ | The dimension of $k[a]$; $1, a, \dots, a^{d-1}$ is a basis |
| $m = m_1 \cdots m_r$ | A factorisation into pairwise coprime monic factors |
| $\varepsilon_i = g_i(a)$ | The canonical idempotents, with $g_i \equiv 1 \pmod{m_i}$ and $g_i \equiv 0 \pmod{m_j}$ for $j \neq i$ |
| $a_i = a\varepsilon_i$ | The component of $a$ in the $i$-th factor; its minimal polynomial is $m_i$ |
| $f_i^{e_i}$ | The primary factors of $m_a$, with $f_i$ irreducible and distinct and $e_i \ge 1$ |
| $C$ | The companion matrix of $x^3-x$ in §Worked Examples |
Further Reading
- Nicolas Bourbaki, Algèbre, Chapitres I–III (Hermann, then Springer). The polynomial algebra, the minimal polynomial and the Chinese remainder theorem in the generality used here.
- Serge Lang, Algebra (Springer, 3rd ed. 2002), Chapters II, IV and V, for the evaluation map, the minimal polynomial and the Chinese remainder theorem.
- Thomas W. Hungerford, Algebra (Springer, 1974), Chapter III, for the isomorphism theorems and the principal ideal domain structure of $k[x]$.
- Richard S. Pierce, Associative Algebras (Springer, 1982), for the Peirce decomposition that follows the Chinese remainder decomposition when the idempotents are not central.
- Roger A. Horn and Charles R. Johnson, Matrix Analysis (Cambridge University Press, 2nd ed. 2013), Chapter 3, for the minimal polynomial, the companion matrix and the cyclic subalgebra of a matrix.
- M. F. Atiyah and I. G. Macdonald, Introduction to Commutative Algebra (Addison-Wesley, 1969), Chapter 1, for the Chinese remainder theorem for comaximal ideals.