{
  "id": "kronecker-product",
  "title": "Kronecker Product",
  "lang": "en",
  "source_lang": "en",
  "summary": "The Kronecker product combines two matrices into a block matrix representing the tensor product of their associated linear maps.",
  "category": "mathematics",
  "subcategory": "algebra",
  "type": "concept",
  "aliases": [
    "Matrix tensor product",
    "Direct product of matrices"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/kronecker-product/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/kronecker-product/",
      "json": "https://aiwiki.page/zh/kronecker-product.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Kronecker product is an operation on two matrices that constructs a larger block matrix by multiplying every entry of the first matrix by the entire second matrix. Denoted \\(A\\otimes B\\), it is the coordinate representation of the tensor product of two linear maps. It is used throughout linear algebra, numerical computation, statistics, and quantum mechanics to express systems with a product structure. The operation is named after the German mathematician Leopold Kronecker."
    },
    {
      "heading": "Definition and example",
      "text": "Let \\(A=(a_{ij})\\) be an \\(m\\times n\\) matrix and \\(B=(b_{k\\ell})\\) a \\(p\\times q\\) matrix over the same field, usually the real or complex numbers. Their Kronecker product is the \\(mp\\times nq\\) matrix\n\n\\[A\\otimes B= \\begin{pmatrix} a_{11}B & a_{12}B & \\cdots & a_{1n}B\\\\ a_{21}B & a_{22}B & \\cdots & a_{2n}B\\\\ \\vdots & \\vdots & \\ddots & \\vdots\\\\ a_{m1}B & a_{m2}B & \\cdots & a_{mn}B \\end{pmatrix}.\\]\n\nEquivalently, its entries satisfy\n\n\\[(A\\otimes B)_{(i-1)p+k,\\,(j-1)q+\\ell} =a_{ij}b_{k\\ell}.\\]\n\nFor example, if\n\n\\[A=\\begin{pmatrix}1&2\\\\3&4\\end{pmatrix}, \\qquad B=\\begin{pmatrix}0&5\\\\6&7\\end{pmatrix},\\]\n\nthen\n\n\\[A\\otimes B= \\begin{pmatrix} 0&5&0&10\\\\ 6&7&12&14\\\\ 0&15&0&20\\\\ 18&21&24&28 \\end{pmatrix}.\\]\n\nUnlike ordinary matrix multiplication, the Kronecker product requires no matching inner dimensions. Every pair of finite matrices over the same field has a Kronecker product."
    },
    {
      "heading": "Relationship to tensor products",
      "text": "Suppose \\(A\\) represents a linear map \\(S:V\\to V'\\), and \\(B\\) represents a linear map \\(T:W\\to W'\\). Their tensor product is the linear map\n\n\\[S\\otimes T:V\\otimes W\\longrightarrow V'\\otimes W'\\]\n\nspecified on elementary tensors by\n\n\\[(S\\otimes T)(v\\otimes w)=S(v)\\otimes T(w).\\]\n\nAfter choosing a basis in each space and ordering the resulting product bases consistently, the matrix of this map is \\(A\\otimes B\\).\n\nThus, the abstract tensor product and the Kronecker product are closely related but not identical notions: the former is a construction involving vector spaces and maps, whereas the latter is an explicit matrix construction in chosen coordinates.\n\nFor column vectors \\(x\\) and \\(y\\), the same notation gives a column vector \\(x\\otimes y\\). The fundamental action rule is\n\n\\[(A\\otimes B)(x\\otimes y)=(Ax)\\otimes(By).\\]\n\nElementary tensors span the tensor-product space, so this identity determines the action on every vector by linearity."
    },
    {
      "heading": "Algebraic properties",
      "text": "Bilinearity and associativity\n\nThe Kronecker product is linear in each argument:\n\n\\[(A+C)\\otimes B=A\\otimes B+C\\otimes B,\\]\n\n\\[A\\otimes(B+D)=A\\otimes B+A\\otimes D,\\]\n\nand, for any scalar \\(\\alpha\\),\n\n\\[(\\alpha A)\\otimes B =A\\otimes(\\alpha B) =\\alpha(A\\otimes B).\\]\n\nThe sums require matching dimensions. It is also associative:\n\n\\[(A\\otimes B)\\otimes C=A\\otimes(B\\otimes C).\\]\n\nAccordingly, repeated products such as \\(A_1\\otimes\\cdots\\otimes A_d\\) need no parentheses under the standard indexing convention.\n\nMixed-product identity\n\nOne of its most useful properties is\n\n\\[(A\\otimes B)(C\\otimes D)=(AC)\\otimes(BD),\\]\n\nprovided \\(AC\\) and \\(BD\\) are defined. This identity converts multiplication of large structured matrices into multiplication of their smaller factors.\n\nFor square matrices and nonnegative integers \\(k\\), it implies\n\n\\[(A\\otimes B)^k=A^k\\otimes B^k.\\]\n\nIt also gives\n\n\\[I_m\\otimes I_p=I_{mp},\\]\n\nwhere \\(I_r\\) is the \\(r\\times r\\) identity matrix.\n\nTransposes, adjoints, and inverses\n\nThe transpose and conjugate transpose act separately on the factors:\n\n\\[(A\\otimes B)^{\\mathsf T} =A^{\\mathsf T}\\otimes B^{\\mathsf T},\\]\n\n\\[(A\\otimes B)^*=A^*\\otimes B^*.\\]\n\nIf both factors are square and invertible, their inverses satisfy\n\n\\[(A\\otimes B)^{-1}=A^{-1}\\otimes B^{-1}.\\]\n\nConsequently, Kronecker products preserve orthogonality and unitarity when both factors have the corresponding property.\n\nNoncommutativity and reordering\n\nIn general,\n\n\\[A\\otimes B\\ne B\\otimes A.\\]\n\nNevertheless, the two products differ only by prescribed row and column permutations. There are permutation matrices \\(P\\) and \\(Q\\) such that\n\n\\[B\\otimes A=P(A\\otimes B)Q.\\]\n\nWhen \\(A\\) and \\(B\\) are square, these can be chosen so that \\(Q=P^{-1}\\), making the two products similar. The permutation implements the exchange \\(x\\otimes y\\mapsto y\\otimes x\\)."
    },
    {
      "heading": "Rank, determinant, and spectral properties",
      "text": "The rank obeys\n\n\\[\\operatorname{rank}(A\\otimes B) =\\operatorname{rank}(A)\\operatorname{rank}(B).\\]\n\nFor square matrices \\(A\\) of order \\(m\\) and \\(B\\) of order \\(n\\), the determinant is\n\n\\[\\det(A\\otimes B)=\\det(A)^n\\det(B)^m,\\]\n\nand the trace satisfies\n\n\\[\\operatorname{tr}(A\\otimes B) =\\operatorname{tr}(A)\\operatorname{tr}(B).\\]\n\nIf \\(Ax=\\lambda x\\) and \\(By=\\mu y\\), with \\(x\\) and \\(y\\) nonzero, then\n\n\\[(A\\otimes B)(x\\otimes y) =\\lambda\\mu(x\\otimes y).\\]\n\nMore generally, over the complex numbers the eigenvalues of \\(A\\otimes B\\), counted with algebraic multiplicity, are all pairwise products\n\n\\[\\lambda_i(A)\\lambda_j(B).\\]\n\nThis remains true even when the factors are not diagonalizable.\n\nThe nonzero singular values are likewise all pairwise products of the nonzero singular values of the factors, as follows from their singular value decompositions. In particular, the spectral norm satisfies\n\n\\[\\|A\\otimes B\\|_2=\\|A\\|_2\\|B\\|_2.\\]\n\nFor square nonsingular factors, the spectral-norm condition number therefore satisfies\n\n\\[\\kappa_2(A\\otimes B)=\\kappa_2(A)\\kappa_2(B).\\]\n\nThus, a structured representation can reduce computational costs without eliminating ill-conditioning; repeated products can amplify it.\n\nIf both factors are Hermitian positive semidefinite, their Kronecker product is also Hermitian positive semidefinite. If both are positive definite, so is their product."
    },
    {
      "heading": "Vectorization and matrix equations",
      "text": "The vectorization operator \\(\\operatorname{vec}\\) stacks a matrix’s columns into one column vector. Under this convention,\n\n\\[\\operatorname{vec}(AXB) =(B^{\\mathsf T}\\otimes A)\\operatorname{vec}(X)\\]\n\nwhenever the matrix products are defined.\n\nFor example, if \\(A\\) is \\(m\\times n\\), \\(X\\) is \\(n\\times p\\), and \\(B\\) is \\(p\\times q\\), both sides are vectors of length \\(mq\\). The transpose is an ordinary transpose even when entries are complex.\n\nThis identity turns equations involving unknown matrices into ordinary systems of linear equations. The Sylvester equation\n\n\\[AX+XB=C,\\]\n\nwith \\(A\\) of order \\(m\\), \\(B\\) of order \\(n\\), and \\(X,C\\) of size \\(m\\times n\\), becomes\n\n\\[\\left(I_n\\otimes A+B^{\\mathsf T}\\otimes I_m\\right) \\operatorname{vec}(X) =\\operatorname{vec}(C).\\]\n\nThe transformed system is useful for theoretical analysis, although specialized algorithms often solve the original matrix equation without forming the larger coefficient matrix.\n\nVectorization conventions matter: stacking rows instead of columns changes the order of the Kronecker factors and the associated formulas."
    },
    {
      "heading": "Kronecker sums",
      "text": "For square matrices \\(A\\) of order \\(m\\) and \\(B\\) of order \\(n\\), the Kronecker sum is commonly defined as\n\n\\[A\\oplus B=A\\otimes I_n+I_m\\otimes B.\\]\n\nIts eigenvalues are all pairwise sums\n\n\\[\\lambda_i(A)+\\lambda_j(B).\\]\n\nBecause its two summands commute, its matrix exponential factors as\n\n\\[e^{t(A\\oplus B)}=e^{tA}\\otimes e^{tB}.\\]\n\nKronecker sums occur when a multidimensional operator is the sum of operators acting separately on individual coordinates. Some texts use a reversed ordering of the factors; the convention must therefore be checked."
    },
    {
      "heading": "Applications",
      "text": "Numerical differential equations\n\nDiscretizations of partial differential equations on tensor-product grids often produce Kronecker structure. For example, a two-dimensional discrete Laplace operator can be written, with a suitable ordering of grid values, as\n\n\\[L=L_x\\otimes I_y+I_x\\otimes L_y,\\]\n\nwhere \\(L_x\\) and \\(L_y\\) are one-dimensional difference matrices.\n\nThis representation separates coordinate directions and supports efficient storage, spectral analysis, and solution methods. Curved domains, irregular grids, and nonseparable coefficients can weaken or destroy the exact structure.\n\nStatistics and covariance modeling\n\nA separable covariance matrix for observations indexed by two factors, such as location and time, has the form\n\n\\[\\Sigma=\\Sigma_{\\mathrm{time}}\\otimes\\Sigma_{\\mathrm{space}},\\]\n\nunder a compatible ordering. Individual covariances then factor into a temporal term and a spatial term.\n\nSuch models can greatly reduce storage and simplify likelihood calculations. Separability is a modeling restriction, however: a general space–time covariance need not admit a single Kronecker factorization.\n\nQuantum mechanics\n\nIn quantum mechanics, the state space of a composite finite-dimensional system is a tensor product. If an operator \\(A\\) acts only on the first subsystem, its action on the composite system is represented by\n\n\\[A\\otimes I.\\]\n\nOperators acting independently on two subsystems are represented by \\(A\\otimes B\\). General states need not be single tensor products, and general operators need not be single Kronecker products; sums of such products are required to represent arbitrary operators."
    },
    {
      "heading": "Computational considerations and distinctions",
      "text": "If \\(A\\) is \\(m\\times n\\) and \\(B\\) is \\(p\\times q\\), explicitly storing a dense \\(A\\otimes B\\) requires \\(mnpq\\) entries. Keeping only the factors requires \\(mn+pq\\) entries.\n\nMultiplication can often be performed without constructing the product. If \\(X\\) is \\(q\\times n\\), then\n\n\\[(A\\otimes B)\\operatorname{vec}(X) =\\operatorname{vec}(BXA^{\\mathsf T}).\\]\n\nFor two dense \\(n\\times n\\) factors, this reduces a matrix–vector multiplication from order \\(n^4\\) arithmetic operations to order \\(n^3\\), using ordinary matrix multiplication. If the factors are sparse, the number of nonzero entries satisfies\n\n\\[\\operatorname{nnz}(A\\otimes B) =\\operatorname{nnz}(A)\\operatorname{nnz}(B).\\]\n\nThese savings apply when the operator is available in exact or useful approximate Kronecker form; an arbitrary large matrix need not have such a compact representation.\n\nThe Kronecker product should also be distinguished from the Hadamard product, which multiplies corresponding entries of equally sized matrices and preserves their dimensions. Ordinary matrix multiplication contracts a shared index; the Kronecker product instead retains both sets of indices, producing a larger matrix."
    }
  ],
  "links": [
    "matrix",
    "tensor-product",
    "linear-algebra",
    "field-mathematics",
    "matrix-multiplication",
    "linear-map",
    "basis-linear-algebra",
    "identity-matrix",
    "matrix-transpose",
    "conjugate-transpose",
    "inverse-matrix",
    "matrix-similarity",
    "matrix-rank",
    "determinant",
    "matrix-trace",
    "eigenvalues-and-eigenvectors",
    "singular-value-decomposition",
    "condition-number",
    "positive-definite-matrix",
    "system-of-linear-equations",
    "matrix-exponential",
    "partial-differential-equation",
    "covariance-matrix",
    "quantum-mechanics"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}