{
  "id": "hermitian-matrix",
  "title": "Hermitian Matrix",
  "lang": "en",
  "source_lang": "en",
  "summary": "A Hermitian matrix equals its conjugate transpose; it has real eigenvalues and an orthonormal basis of eigenvectors.",
  "category": "mathematics",
  "subcategory": "algebra",
  "type": "concept",
  "aliases": [
    "Self-adjoint matrix",
    "Hermitian operator matrix"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/hermitian-matrix/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/hermitian-matrix/",
      "json": "https://aiwiki.page/zh/hermitian-matrix.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A Hermitian matrix is a square matrix with complex entries that equals its conjugate transpose. Thus, a matrix \\(A=(a_{ij})\\) is Hermitian if\n\n\\[A=A^*,\\qquad a_{ij}=\\overline{a_{ji}},\\]\n\nwhere \\(A^*\\), also written \\(A^\\dagger\\), is obtained by transposing \\(A\\) and taking the complex conjugate of every entry. Hermitian matrices are the complex counterparts of real symmetric matrices. They are fundamental in linear algebra because their eigenvalues are real and they admit orthonormal eigenvector bases, properties central to optimization, numerical computation, statistics, and quantum mechanics."
    },
    {
      "heading": "Definition and examples",
      "text": "For a complex number \\(z=a+ib\\), its conjugate is \\(\\overline z=a-ib\\). The Hermitian condition requires entries on opposite sides of the main diagonal to be complex conjugates. Every diagonal entry must therefore be a real number, since\n\n\\[a_{ii}=\\overline{a_{ii}}.\\]\n\nFor example,\n\n\\[A= \\begin{pmatrix} 2&1+i\\\\ 1-i&3 \\end{pmatrix}\\]\n\nis Hermitian. By contrast,\n\n\\[B= \\begin{pmatrix} 2&i\\\\ i&3 \\end{pmatrix}\\]\n\nis not Hermitian: its two off-diagonal entries are equal rather than conjugates.\n\nA real matrix is Hermitian exactly when it equals its transpose, so the real case reduces to symmetry. For complex matrices, symmetry and Hermitian symmetry are distinct conditions.\n\nEvery Hermitian matrix can be written uniquely as\n\n\\[A=S+iK,\\]\n\nwhere \\(S\\) is a real symmetric matrix and \\(K\\) is a real skew-symmetric matrix:\n\n\\[S^{\\mathsf T}=S,\\qquad K^{\\mathsf T}=-K.\\]\n\nIn particular, the imaginary part of every diagonal entry is zero.\n\nThe terminology honors the French mathematician Charles Hermite, whose name is also associated with Hermitian forms."
    },
    {
      "heading": "Interpretation as a self-adjoint linear map",
      "text": "On \\(\\mathbb C^n\\), use the standard inner product\n\n\\[\\langle x,y\\rangle=x^*y.\\]\n\nA matrix \\(A\\) is Hermitian exactly when\n\n\\[\\langle Ax,y\\rangle=\\langle x,Ay\\rangle\\]\n\nfor all vectors \\(x,y\\). Thus it represents a self-adjoint operator with respect to this inner product.\n\nThis statement depends on the inner product and the coordinates used. A self-adjoint linear map has a Hermitian matrix in any orthonormal basis, but its matrix in an arbitrary basis need not be Hermitian.\n\nMore generally, if an inner product has the coordinate expression\n\n\\[\\langle x,y\\rangle_G=x^*Gy,\\]\n\nwhere \\(G\\) is Hermitian positive definite, then a coordinate matrix \\(A\\) represents a self-adjoint map precisely when\n\n\\[A^*G=GA.\\]\n\nThe familiar condition \\(A^*=A\\) is the special case \\(G=I\\)."
    },
    {
      "heading": "Spectral properties",
      "text": "The finite-dimensional spectral theorem states that \\(A\\) is Hermitian if and only if\n\n\\[A=U\\Lambda U^*,\\]\n\nwhere \\(U\\) is a unitary matrix and\n\n\\[\\Lambda=\\operatorname{diag}(\\lambda_1,\\ldots,\\lambda_n)\\]\n\nhas real diagonal entries.\n\nThe columns of \\(U\\) form an orthonormal basis of eigenvectors. Consequently:\n\n- Every eigenvalue is real.\n- Eigenvectors associated with distinct eigenvalues are orthogonal.\n- Each eigenvalue’s algebraic multiplicity equals the dimension of its eigenspace.\n- Every Hermitian matrix is diagonalizable, including when eigenvalues are repeated.\n\nWhy the eigenvalues are real\n\nSuppose \\(Av=\\lambda v\\) with \\(v\\ne0\\). Then\n\n\\[v^*Av=\\lambda v^*v.\\]\n\nHermitian symmetry gives\n\n\\[\\overline{v^*Av}=v^*A^*v=v^*Av,\\]\n\nso \\(v^*Av\\) is real. Since \\(v^*v>0\\),\n\n\\[\\lambda=\\frac{v^*Av}{v^*v}\\in\\mathbb R.\\]\n\nThis argument explains why Hermitian symmetry constrains eigenvalues and eigenvectors so strongly. Merely having real eigenvalues, however, does not make a matrix Hermitian: the unitary diagonalization condition is essential.\n\nSpectral projections\n\nLet \\(\\mu_1,\\ldots,\\mu_r\\) be the distinct eigenvalues of \\(A\\). Then\n\n\\[A=\\sum_{j=1}^r\\mu_jP_j,\\]\n\nwhere \\(P_j\\) is the orthogonal projection onto the eigenspace associated with \\(\\mu_j\\). These projections satisfy\n\n\\[P_j^*=P_j,\\qquad P_j^2=P_j,\\qquad P_jP_k=0\\quad(j\\ne k),\\qquad \\sum_{j=1}^rP_j=I.\\]\n\nThis representation separates the action of \\(A\\) into independent real scalings on mutually orthogonal subspaces."
    },
    {
      "heading": "Quadratic forms and definiteness",
      "text": "For a Hermitian matrix \\(A\\), the expression\n\n\\[q_A(x)=x^*Ax\\]\n\nis real for every \\(x\\in\\mathbb C^n\\). It is the complex analogue of a real quadratic form, often called a Hermitian form when expressed through its associated sesquilinear form \\(x^*Ay\\).\n\nConversely, if a complex square matrix satisfies \\(x^*Ax\\in\\mathbb R\\) for every complex vector \\(x\\), then it is Hermitian. Requiring the condition for every complex vector is important; testing only real vectors does not suffice.\n\nHermitian matrices are classified by the signs of these values:\n\nType | Condition for every nonzero \\(x\\) | Eigenvalue condition\nPositive definite | \\(x^*Ax>0\\) | All eigenvalues are positive\nPositive semidefinite | \\(x^*Ax\\ge0\\) | All eigenvalues are nonnegative\nNegative definite | \\(x^*Ax<0\\) | All eigenvalues are negative\nNegative semidefinite | \\(x^*Ax\\le0\\) | All eigenvalues are nonpositive\nIndefinite | Takes both positive and negative values | Has both positive and negative eigenvalues\n\nA positive definite matrix is necessarily invertible. A positive semidefinite matrix may be singular.\n\nFor any complex matrix \\(B\\), the Gram matrix\n\n\\[B^*B\\]\n\nis Hermitian positive semidefinite because\n\n\\[x^*B^*Bx=\\|Bx\\|^2\\ge0.\\]\n\nIt is positive definite exactly when the columns of \\(B\\) are linearly independent. Conversely, every Hermitian positive semidefinite matrix can be factored as \\(B^*B\\).\n\nFor example,\n\n\\[A= \\begin{pmatrix} a&z\\\\ \\overline z&b \\end{pmatrix}, \\qquad a,b\\in\\mathbb R,\\]\n\nhas eigenvalues\n\n\\[\\lambda_\\pm =\\frac{a+b\\pm\\sqrt{(a-b)^2+4|z|^2}}{2}.\\]\n\nIt is positive definite exactly when\n\n\\[a>0,\\qquad ab-|z|^2>0.\\]"
    },
    {
      "heading": "Rayleigh quotient and variational characterization",
      "text": "For \\(x\\ne0\\), the Rayleigh quotient is\n\n\\[R_A(x)=\\frac{x^*Ax}{x^*x}.\\]\n\nIf the eigenvalues are ordered as\n\n\\[\\lambda_1\\le\\cdots\\le\\lambda_n,\\]\n\nthen\n\n\\[\\lambda_1\\le R_A(x)\\le\\lambda_n.\\]\n\nThe extreme values are attained at eigenvectors:\n\n\\[\\lambda_1=\\min_{\\|x\\|=1}x^*Ax,\\qquad \\lambda_n=\\max_{\\|x\\|=1}x^*Ax.\\]\n\nMore generally, intermediate eigenvalues have minimax characterizations over subspaces. These connect spectral theory to mathematical optimization and underpin many eigenvalue approximation methods.\n\nThe spectral norm satisfies\n\n\\[\\|A\\|_2=\\max_j|\\lambda_j|.\\]\n\nIf \\(A\\) is invertible, its spectral condition number is\n\n\\[\\kappa_2(A)= \\frac{\\max_j|\\lambda_j|}{\\min_j|\\lambda_j|}.\\]\n\nThus Hermitian symmetry does not guarantee good conditioning: an eigenvalue very close to zero can make a matrix highly ill-conditioned."
    },
    {
      "heading": "Algebraic structure",
      "text": "Hermitian matrices are closed under addition and multiplication by real scalars. The \\(n\\times n\\) Hermitian matrices therefore form a real vector space of dimension \\(n^2\\): there are \\(n\\) real diagonal parameters and two real parameters for each of the \\(n(n-1)/2\\) entries above the diagonal.\n\nThey do not form a complex vector space. If \\(A\\ne0\\) is Hermitian, then \\(iA\\) is skew-Hermitian:\n\n\\[(iA)^*=-iA.\\]\n\nSeveral useful closure properties follow directly:\n\n- If \\(A\\) is invertible and Hermitian, its inverse is Hermitian.\n- If \\(A\\) and \\(B\\) are Hermitian, then \\(AB\\) is Hermitian exactly when \\(AB=BA\\).\n- If \\(A\\) is Hermitian, then \\(C^*AC\\) is Hermitian for every compatible matrix \\(C\\).\n- Every principal submatrix of a Hermitian matrix is Hermitian.\n- Every polynomial in \\(A\\) with real coefficients is Hermitian.\n\nEvery complex square matrix \\(M\\) has a unique decomposition\n\n\\[M=H+iK, \\qquad H=\\frac{M+M^*}{2},\\qquad K=\\frac{M-M^*}{2i},\\]\n\nwhere both \\(H\\) and \\(K\\) are Hermitian.\n\nThe trace and determinant of a Hermitian matrix are real:\n\n\\[\\operatorname{tr}(A)=\\sum_j\\lambda_j,\\qquad \\det(A)=\\prod_j\\lambda_j.\\]\n\nNeither is necessarily positive."
    },
    {
      "heading": "Functions of a Hermitian matrix",
      "text": "The spectral decomposition defines matrix functions through\n\n\\[f(A)=U\\operatorname{diag} \\bigl(f(\\lambda_1),\\ldots,f(\\lambda_n)\\bigr)U^*.\\]\n\nWhenever \\(f\\) is real-valued on the eigenvalues, \\(f(A)\\) is Hermitian.\n\nImportant cases include:\n\n- A positive semidefinite matrix has a unique positive semidefinite square root.\n- A positive definite matrix has a unique Hermitian logarithm.\n- For Hermitian \\(A\\), the matrix exponential \\(e^A\\) is positive definite.\n- For real \\(t\\), the matrix \\(e^{itA}\\) is unitary.\n\nThe last property connects Hermitian generators with unitary evolution."
    },
    {
      "heading": "Applications",
      "text": "Quantum mechanics\n\nIn finite-dimensional quantum mechanics, observables are represented by Hermitian operators. Their eigenvalues are possible measurement outcomes, and their orthogonal spectral projections specify the associated measurement probabilities.\n\nA quantum state represented by a density matrix is Hermitian positive semidefinite and has trace one. A Hermitian Hamiltonian \\(H\\) generates unitary time evolution, for a time-independent system, through\n\n\\[U(t)=e^{-itH/\\hbar},\\]\n\nwhere \\(\\hbar\\) is the reduced Planck constant.\n\nIn infinite-dimensional Hilbert spaces, the corresponding theory requires care with operator domains. For unbounded operators, symmetry alone is weaker than self-adjointness; the finite-dimensional matrix equivalence does not transfer without qualification.\n\nStatistics and signal processing\n\nFor a complex random vector \\(X\\) with mean \\(\\mu\\), its covariance matrix\n\n\\[C=\\mathbb E\\bigl[(X-\\mu)(X-\\mu)^*\\bigr]\\]\n\nis Hermitian positive semidefinite. Its eigenvalues measure variation along orthogonal directions, supporting principal component analysis for complex data.\n\nHermitian covariance and spectral-density matrices also occur in multichannel signal processing, array processing, and frequency-domain analysis.\n\nOptimization\n\nPositive semidefinite Hermitian matrices form a convex cone. Constraints of the form\n\n\\[A(x)=A_0+\\sum_{j=1}^m x_jA_j\\succeq0, \\qquad x_j\\in\\mathbb R,\\]\n\nwhere each \\(A_j\\) is Hermitian, are linear matrix inequalities. They are central to semidefinite programming, a branch of convex optimization.\n\nNumerical computation\n\nAlgorithms in numerical linear algebra exploit Hermitian structure to reduce storage and computation. Only one triangular part and the real diagonal need to be stored.\n\nDense Hermitian eigenvalue algorithms commonly reduce the matrix to real symmetric tridiagonal form using unitary transformations, then solve the tridiagonal eigenproblem. For large sparse matrices, iterative methods can estimate selected eigenvalues without computing the full spectrum.\n\nPositive definite Hermitian systems admit Cholesky decomposition,\n\n\\[A=LL^*,\\]\n\nwhere \\(L\\) is lower triangular with positive real diagonal entries. Indefinite Hermitian matrices generally require other factorizations, often with pivoting."
    },
    {
      "heading": "Perturbations and computational limitations",
      "text": "Hermitian eigenvalues obey strong perturbation bounds. If \\(A\\) and \\(A+E\\) are Hermitian and their eigenvalues are arranged in increasing order, then\n\n\\[|\\lambda_j(A+E)-\\lambda_j(A)|\\le\\|E\\|_2.\\]\n\nThus a small Hermitian perturbation cannot move an eigenvalue by more than the perturbation’s spectral norm.\n\nEigenvectors can be much more sensitive. When eigenvalues are close together, small perturbations may substantially rotate individual eigenvectors, even though the invariant subspace associated with the cluster is more stable.\n\nIn floating-point computation, an intended Hermitian matrix may acquire a small non-Hermitian component. Replacing \\(M\\) by\n\n\\[\\frac{M+M^*}{2}\\]\n\nremoves that component and is the orthogonal projection onto the Hermitian matrices under the real Frobenius inner product. This operation enforces Hermitian symmetry but does not ensure positive semidefiniteness or correct errors in the underlying data."
    }
  ],
  "links": [
    "matrix",
    "conjugate-transpose",
    "linear-algebra",
    "complex-number",
    "real-number",
    "matrix-transpose",
    "inner-product",
    "self-adjoint-operator",
    "linear-map",
    "spectral-theorem",
    "unitary-matrix",
    "orthonormal-basis",
    "eigenvalues-and-eigenvectors",
    "orthogonal-projection",
    "quadratic-form",
    "positive-definite-matrix",
    "gram-matrix",
    "mathematical-optimization",
    "condition-number",
    "vector-space",
    "inverse-matrix",
    "matrix-trace",
    "determinant",
    "exponential-function",
    "quantum-mechanics",
    "covariance-matrix",
    "principal-component-analysis",
    "convex-optimization",
    "numerical-linear-algebra",
    "cholesky-decomposition"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}