1 Definition and basic idea
The rational canonical form is a standard way to represent a square matrix up to similarity over an arbitrary field. It replaces an arbitrary matrix with a block diagonal matrix built from companion matrices of certain polynomials called invariant factors. This form is canonical in the sense that two matrices are similar if and only if they have the same rational canonical form.
The construction is especially valuable when the base field is not algebraically closed. In that setting, eigenvalues may not exist in the field, so diagonalization and Jordan form may fail to describe the matrix completely. The rational canonical form avoids this difficulty by using polynomial data rather than eigenvectors.
1.1 Similarity and canonical forms
Two matrices are similar if one can be obtained from the other by a change of basis. Similarity preserves the underlying linear transformation, so it is the natural equivalence relation for classification problems in linear algebra. A canonical form gives one distinguished representative from each similarity class.
Rational canonical form is one such representative. It organizes a matrix into blocks that encode its action on invariant subspaces. Because the form is determined uniquely by the matrix up to the order of the blocks, it provides a complete similarity invariant.
1.2 Companion matrices
A companion matrix is a square matrix associated with a monic polynomial. For a polynomial of degree n, the companion matrix is an n by n matrix whose characteristic and minimal polynomial are both closely tied to that polynomial. It has a simple sparse structure and serves as the basic block in rational canonical form.
Companion matrices are important because they represent cyclic linear transformations, those generated by a single vector under repeated application of the operator. Each block in rational canonical form is a companion matrix for one invariant factor.
1.3 Invariant factors
Invariant factors are monic polynomials arranged so that each divides the next. They encode the structure of a matrix as a module over the polynomial ring in one variable. The largest invariant factor is the minimal polynomial, while the product of all invariant factors is the characteristic polynomial.
These polynomials determine the rational canonical form completely. Each invariant factor corresponds to one companion matrix block, and the divisibility condition ensures that the blocks fit together in a canonical sequence.
1.4 Elementary divisors
Elementary divisors provide a finer factorization of the same structural data. They are powers of irreducible polynomials and can be viewed as the prime-power building blocks from which invariant factors are assembled. While invariant factors give the rational canonical form directly, elementary divisors describe the decomposition in a more refined way.
Over fields where polynomials factor into linear terms, elementary divisors are closely related to Jordan blocks. Over general fields, they remain meaningful because they use irreducible polynomials instead of eigenvalues.
2 Existence and uniqueness
The rational canonical form theorem guarantees that every square matrix over a field is similar to a unique block diagonal matrix of companion matrices. The blocks are determined by the invariant factors of the associated linear transformation. This result holds over any field, not only those in which every polynomial splits.
The theorem links matrix similarity to the structure of finitely generated modules over a principal ideal domain. That algebraic viewpoint explains both why the form exists and why it is unique.
2.1 Statement of the rational canonical form theorem
For every square matrix over a field, there is a basis in which the matrix becomes block diagonal, with each block a companion matrix of a monic polynomial. The associated polynomials can be chosen so that each divides the next. The resulting matrix is the rational canonical form.
The theorem also states that this decomposition is unique up to the order of the blocks, provided the invariant factors are listed in the standard divisibility order. Thus the form is a complete invariant of similarity.
2.2 Uniqueness of invariant factors
The invariant factors are uniquely determined by the matrix. Their uniqueness follows from the uniqueness of the decomposition of a module over a principal ideal domain into cyclic submodules. In matrix language, this means that no two different sequences of monic divisibility-ordered polynomials can describe the same similarity class.
This uniqueness makes the rational canonical form computationally useful. Once the invariant factors are found, the canonical representative is fixed.
2.3 Relation to similarity classes
Similarity classes of matrices correspond exactly to rational canonical forms. Two matrices are similar if and only if they have the same invariant factors. Therefore the classification problem for linear operators over a field reduces to computing those polynomials.
This relationship provides a bridge between concrete matrix calculations and abstract algebraic structure. It is particularly helpful when direct eigenvalue analysis is unavailable.
3 Construction of the form
The rational canonical form can be built from cyclic invariant subspaces. One studies the module generated by a basis vector under the action of the matrix and then decomposes the full space into a direct sum of such cyclic pieces. The companion matrices arise from the relations satisfied by these generators.
Another route uses the Smith normal form of the polynomial matrix xI minus A. This approach is often the most systematic and shows clearly how the invariant factors are extracted.
3.1 Cyclic subspaces
A cyclic subspace is generated by a vector v together with its images under repeated application of the linear operator. The vectors v, Av, A squared v, and so on, span a subspace that is stable under the operator. If this subspace has dimension n, then the operator on it is represented by a companion matrix.
Cyclic subspaces are the basic building blocks of the rational canonical form. The full space is assembled from a direct sum of such pieces.
3.2 Decomposition into invariant subspaces
The vector space can be decomposed into a direct sum of invariant subspaces, each of which is cyclic. On each subspace, the operator acts like multiplication by x modulo a polynomial relation. The corresponding companion matrix records that relation in matrix form.
The decomposition is arranged so that the associated polynomials divide one another in a nested chain. This ordering produces the invariant factors.
3.3 Smith normal form approach
The Smith normal form is applied to the polynomial matrix xI minus A over the polynomial ring in one variable. By using elementary row and column operations, one obtains a diagonal form whose diagonal entries are the invariant factors. These entries determine the rational canonical form directly.
This method is conceptually powerful because it converts a similarity problem into a module-theoretic normal form problem. It also makes the relationship with the characteristic and minimal polynomials transparent.
3.4 Algorithmic computation
In practice, the rational canonical form can be computed by determining the invariant factors from xI minus A. This may involve polynomial factorization, greatest common divisors, and Smith normal form calculations. Once the invariant factors are known, the companion blocks are written down immediately.
The method is well suited to symbolic computation and to matrices over finite fields or other nonalgebraically closed fields. It provides a deterministic procedure rather than a search for eigenvectors.
4 Matrix blocks in rational canonical form
The rational canonical form is block diagonal, and each block is the companion matrix of one invariant factor. The blocks are ordered so that each polynomial divides the next. This arrangement gives a canonical and highly structured matrix representative.
The block structure reflects the decomposition of the underlying module into cyclic components. It also makes many similarity computations straightforward.
4.1 Companion matrix of a polynomial
For a monic polynomial of degree n, the companion matrix is an n by n matrix with ones on the subdiagonal and coefficients from the polynomial in the last column. It is designed so that its characteristic polynomial is the given polynomial. The matrix is cyclic, meaning one vector generates the entire space under repeated application.
Companion matrices are especially convenient because they translate polynomial relations into linear transformations. They are the elementary blocks used throughout the rational canonical form.
4.2 Block diagonal structure
The rational canonical form consists of a block diagonal matrix whose blocks are companion matrices. Each block corresponds to one invariant factor. Since the blocks act independently on separate invariant subspaces, the whole matrix is a direct sum of simpler pieces.
This structure makes the canonical form easy to read: the matrix is completely determined by listing the companion blocks in order. The decomposition is unique up to rearrangement consistent with the divisibility condition.
4.3 Ordering of invariant factors
The invariant factors are ordered so that each divides the next. This convention is standard and ensures uniqueness. Different orderings of the same set of polynomials do not change the similarity class, but the canonical form uses a fixed ordering to remove ambiguity.
The ordering also aligns the rational canonical form with the module decomposition underlying it. Smaller cyclic pieces are absorbed into larger ones in a controlled way.
4.4 Special cases
If the matrix is already cyclic, there is only one invariant factor, and the rational canonical form is a single companion matrix. If the matrix is diagonalizable over the field and its eigenvalues lie in the field, the rational canonical form may further simplify into 1 by 1 blocks. When the field is not algebraically closed, however, companion blocks of higher degree may remain unavoidable.
These special cases show how the rational canonical form adapts to different algebraic settings. It is flexible enough to handle both splitting and non-splitting polynomials.
5 Relation to other canonical forms
The rational canonical form is closely related to several other matrix normal forms. It generalizes ideas from the Jordan canonical form and shares a module-theoretic foundation with the Smith normal form. It is also connected to the characteristic and minimal polynomials.
These relationships help place the form within the broader structure of linear algebra and algebraic theory.
5.1 Jordan canonical form
The Jordan canonical form applies when the base field is algebraically closed or at least contains the relevant eigenvalues. It decomposes a matrix into Jordan blocks associated with individual eigenvalues. The rational canonical form instead uses irreducible polynomials, so it remains valid even when eigenvalues are absent from the field.
When the field does split all relevant polynomials, the rational canonical form can be refined into Jordan form. In that sense, Jordan form is a more detailed version over suitable fields.
5.2 Frobenius normal form
The rational canonical form is often called the Frobenius normal form. In many texts, the two names refer to the same construction. The terminology emphasizes the role of cyclic modules and invariant factors.
This alternative name reflects the historical development of the theory. It also underscores the form's algebraic rather than geometric character.
5.3 Smith normal form
The Smith normal form is a diagonal form for matrices over a principal ideal domain. It is central to the derivation of the rational canonical form because xI minus A is treated as a matrix over a polynomial ring. The invariant factors appearing in the Smith normal form become the blocks of the rational canonical form.
Thus the rational canonical form can be viewed as a matrix analogue of the Smith normal form specialized to similarity questions. The connection explains why divisibility properties govern the classification.
5.4 Minimal polynomial and characteristic polynomial
The characteristic polynomial is the product of the invariant factors, while the minimal polynomial is the largest invariant factor. These two polynomials summarize much of the structure of the form. In particular, the minimal polynomial indicates the strongest polynomial relation satisfied by the matrix.
Knowing only these two polynomials is not always enough to recover the rational canonical form, because the intermediate invariant factors may vary. Nevertheless, they provide important constraints on the possible blocks.
6 Applications
The rational canonical form is used to classify operators, compute structural invariants, and simplify polynomial expressions in matrices. It is especially effective over finite fields and other settings where direct eigenvalue methods are limited. The form also supports module-theoretic interpretations that connect linear algebra to abstract algebra.
Its applications are both theoretical and computational, making it a standard tool in advanced linear algebra.
6.1 Classification of linear transformations
The form gives a complete classification of linear transformations up to change of basis. Once the invariant factors are known, the transformation is determined up to similarity. This turns classification into a problem of polynomial decomposition.
Such a classification is useful in studying endomorphisms of finite-dimensional vector spaces. It identifies precisely which transformations have the same algebraic structure.
6.2 Computations over finite fields
Over finite fields, eigenvalues may lie in extension fields rather than in the base field. Rational canonical form avoids that complication by working with irreducible polynomials over the finite field itself. This makes it a practical tool for explicit matrix computations in coding theory, cryptography, and finite linear algebra.
Because the blocks are companion matrices, arithmetic can often be carried out efficiently. The representation is compact and well adapted to symbolic manipulation.
6.3 Module-theoretic interpretation
A matrix can be interpreted as defining a module over the polynomial ring in one variable. Under this viewpoint, the rational canonical form corresponds to the decomposition of that module into a direct sum of cyclic modules. The invariant factors record the sizes and relations among these cyclic components.
This perspective explains the theorem in a conceptual way. It shows that the canonical form is not merely a matrix trick, but an expression of a deeper algebraic structure.
6.4 Determining matrix powers and functions
Once a matrix is in rational canonical form, computing powers or polynomial functions of the matrix becomes simpler. Since each block is a companion matrix, one can reduce high powers using the polynomial relation defining the block. The same idea extends to evaluating polynomials in the matrix.
This is useful when studying recurrence relations and linear dynamical systems. The canonical form turns matrix expressions into calculations modulo invariant factors.
7 Examples
Examples illustrate how the rational canonical form packages information about a matrix. The exact blocks depend on the field as well as on the matrix entries. A matrix that appears complicated in one basis may become transparent after conversion to canonical form.
These examples also show that the same matrix can have different interpretations depending on the base field.
7.1 Matrices over the real numbers
Over the real numbers, a matrix may have irreducible quadratic factors when its eigenvalues are complex. In such cases, the rational canonical form can contain companion matrices of quadratic polynomials rather than 1 by 1 real blocks. This gives a real canonical representation without passing to complex numbers.
For example, a rotation matrix in the plane is represented naturally by a companion matrix of a quadratic polynomial. The form captures its behavior using real coefficients only.
7.2 Matrices over finite fields
Over a finite field, the characteristic polynomial may factor into irreducible polynomials of various degrees. The rational canonical form then uses companion blocks for those factors. This is particularly useful because the field may not contain roots of the characteristic polynomial.
In computations, this form provides a compact and exact description of the similarity class. It is widely used in finite field linear algebra.
7.3 Repeated invariant factors
A matrix may have several invariant factors, some repeated or related by divisibility. In that situation, the rational canonical form consists of multiple companion blocks, each corresponding to one factor in the chain. Repetition does not imply redundancy; rather, it reflects the internal decomposition of the module.
Such examples show how the form encodes more than just the overall minimal and characteristic polynomials. The intermediate factors reveal the layered structure of the transformation.
7.4 Non-diagonalizable matrices
Non-diagonalizable matrices still have rational canonical forms, even when they lack a basis of eigenvectors over the field. The blocks then record the failure of diagonalization through higher-degree or repeated polynomial factors. In fields where Jordan form applies, these blocks correspond to chains of generalized eigenvectors.
The rational canonical form thus serves as a uniform substitute for diagonalization. It provides a canonical description even in cases where simpler normal forms are unavailable.