PhyslibSearch

QuantumInfo.ForMathlib.HermitianMat.Schatten

14 declarations

definition

Schatten pp-norm of a matrix AA

#schattenNorm

For a matrix AMatd×d(C)A \in \text{Mat}_{d \times d}(\mathbb{C}) and an exponent pRp \in \mathbb{R}, the Schatten pp-norm ASp\Vert A \Vert_{S^p} is defined as ASp=(Tr[(AA)p/2])1/p \Vert A \Vert_{S^p} = \left( \text{Tr}\left[ (A A^*)^{p/2} \right] \right)^{1/p} where AA^* denotes the conjugate transpose of AA, and the power of the positive semi-definite matrix AAA A^* is defined via the continuous functional calculus. The result is returned as the real part of the trace.

theorem

Schatten pp-norm of a positive semidefinite Hermitian matrix equals (Tr(Ap))1/p(\text{Tr}(A^p))^{1/p}

#schattenNorm_hermitian_pow

Let AA be a Hermitian matrix of dimension dd over the complex numbers C\mathbb{C}. If AA is positive semidefinite (0A0 \le A) and pp is a positive real number (p>0p > 0), then the Schatten pp-norm of the underlying matrix of AA (denoted A.matA.mat) is given by Ap=(Tr(Ap))1/p,\|A\|_p = (\text{Tr}(A^p))^{1/p}, where ApA^p is defined via the functional calculus for Hermitian matrices and Tr\text{Tr} denotes the matrix trace.

theorem

The Schatten pp-norm is non-negative

#schattenNorm_nonneg

For any n×nn \times n complex matrix AA and any real number pp, the Schatten pp-norm of AA is non-negative, i.e., ASp0\|A\|_{S^p} \geq 0.

theorem

AkSp=ASkpk\|A^k\|_{S^p} = \|A\|_{S^{kp}}^k for A0A \ge 0

#schattenNorm_pow_eq

Let AA be a positive semidefinite Hermitian matrix of size d×dd \times d over C\mathbb{C}. For any positive real numbers p>0p > 0 and k>0k > 0, the Schatten pp-norm of the matrix power AkA^k is equal to the kk-th power of the Schatten (kp)(kp)-norm of AA. That is, AkSp=(ASkp)k.\|A^k\|_{S^p} = (\|A\|_{S^{kp}})^k.

theorem

tr(Ar)=ASrr\text{tr}(A^r) = \|A\|_{S^r}^r for positive semi-definite AA and r>0r > 0

#trace_eq_schattenNorm_rpow

Let AA be a d×dd \times d complex Hermitian matrix such that AA is positive semi-definite (0A0 \le A). For any real number r>0r > 0, the trace of the rr-th power of AA is equal to the rr-th power of its Schatten rr-norm, denoted as tr(Ar)=ASrr\text{tr}(A^r) = \|A\|_{S^r}^r.

theorem

Trace of (AA)p/2(A^*A)^{p/2} equals the sum of its eigenvalues to the p/2p/2 power

#schattenNorm_trace_as_eigenvalue_sum

Let AA be a d×dd \times d complex matrix and let pRp \in \mathbb{R}. Let M=AAM = A^* A be the Hermitian matrix formed by the product of the conjugate transpose of AA and AA itself. The real part of the trace of the functional calculus of MM applied to the power function xxp/2x \mapsto x^{p/2} is equal to the sum of the p/2p/2-th powers of the eigenvalues of MM: Re(Tr((AA)p/2))=i=1dλi(AA)p/2\text{Re}\left(\text{Tr}\left((A^* A)^{p/2}\right)\right) = \sum_{i=1}^d \lambda_i(A^* A)^{p/2} where λi(AA)\lambda_i(A^* A) denotes the ii-th eigenvalue of the Hermitian matrix AAA^* A.

theorem

ASpp=σi(A)p\|A\|_{S^p}^p = \sum \sigma_i(A)^p for p>0p > 0

#schattenNorm_rpow_eq_sum_singularValues

For any d×dd \times d complex matrix AA and any positive real number p>0p > 0, the pp-th power of the Schatten pp-norm of AA is equal to the sum of the pp-th powers of its singular values, i.e., ASpp=idσi(A)p\|A\|_{S^p}^p = \sum_{i \in d} \sigma_i(A)^p where σi(A)\sigma_i(A) denotes the ii-th singular value of AA.

theorem

Schatten pp-norm equals partial sum of singular values: ASp=(σi(A)p)1/p\|A\|_{S^p} = (\sum \sigma_i(A)^p)^{1/p}

#schattenNorm_eq_sum_singularValues_rpow

Let AA be a square complex matrix of size d×dd \times d and pp be a positive real number (p>0p > 0). The Schatten pp-norm of AA, denoted ASp\|A\|_{S^p}, is equal to the pp-th root of the sum of the pp-th powers of its singular values σi(A)\sigma_i(A): ASp=(idσi(A)p)1/p\|A\|_{S^p} = \left( \sum_{i \in d} \sigma_i(A)^p \right)^{1/p}

theorem

ASpp=(σi(A))p\|A\|_{S^p}^p = \sum (\sigma^\downarrow_i(A))^p

#schattenNorm_rpow_eq_sum_sorted

Let AA be a d×dd \times d complex matrix and pp be a positive real number. Let ASp\|A\|_{S^p} denote the Schatten pp-norm of AA, and let σi(A)\sigma^\downarrow_i(A) denote the ii-th singular value of AA sorted in non-increasing order. Then the pp-th power of the Schatten pp-norm equals the sum of the pp-th powers of the sorted singular values: ASpp=i=0n1(σi(A))p\|A\|_{S^p}^p = \sum_{i=0}^{n-1} (\sigma^\downarrow_i(A))^p where nn is the dimension of the matrix.

theorem

Trace Young Inequality for PSD Matrices: A,BTr(Ap)p+Tr(Bq)q\langle A, B \rangle \le \frac{\text{Tr}(A^p)}{p} + \frac{\text{Tr}(B^q)}{q}

#trace_young

Let AA and BB be complex d×dd \times d Hermitian matrices that are positive semidefinite (A0,B0A \ge 0, B \ge 0). For any real numbers p,q>1p, q > 1 satisfying the conjugacy condition 1p+1q=1\frac{1}{p} + \frac{1}{q} = 1, the real inner product of AA and BB satisfies the Young inequality: A,BRTr(Ap)p+Tr(Bq)q\langle A, B \rangle_{\mathbb{R}} \leq \frac{\text{Tr}(A^p)}{p} + \frac{\text{Tr}(B^q)}{q} where A,BR=Tr(AB)\langle A, B \rangle_{\mathbb{R}} = \text{Tr}(AB) and Ap,BqA^p, B^q are defined via functional calculus.

theorem

(A1/2B)(A1/2B)=BAB(A^{1/2} B)^* (A^{1/2} B) = B A B for A0A \ge 0

#conjTranspose_half_mul_eq_conj

Let AA and BB be Hermitian matrices of size d×dd \times d over C\mathbb{C}. If AA is positive semidefinite (0A0 \le A), then the product of the conjugate transpose of (A1/2B)(A^{1/2} B) and the matrix (A1/2B)(A^{1/2} B) is equal to the matrix BABB A B. Specifically, let C=A1/2BC = A^{1/2} B, then CC=BABC^* C = B A B.

theorem

A1/2BS2α2α=tr((A.conj B)α)\|A^{1/2} B\|_{S^{2\alpha}}^{2\alpha} = \text{tr}((A.conj \ B)^\alpha)

#schattenNorm_half_mul_rpow_eq_trace_conj

Let AA and BB be d×dd \times d Hermitian matrices over C\mathbb{C} such that AA is positive semidefinite (A0A \ge 0). For any real number α>0\alpha > 0, the Schatten 2α2\alpha-norm of the product A1/2BA^{1/2}B raised to the power 2α2\alpha is equal to the trace of the α\alpha-th power of the matrix ABAA B A, denoted here as (A.conj B.mat)α(A.conj \ B.mat)^\alpha. Specifically: A1/2BS2α2α=tr((A1/2BBA1/2)α)\|A^{1/2} B\|_{S^{2\alpha}}^{2\alpha} = \text{tr}((A^{1/2} B^* B A^{1/2})^\alpha) where A1/2A^{1/2} is defined via functional calculus for Hermitian matrices.

theorem

Schatten–Hölder Inequality for Matrix Products: ABSrASpBSq\|AB\|_{S^r} \le \|A\|_{S^p} \|B\|_{S^q}

#schattenNorm_mul_le

Let AA and BB be complex d×dd \times d matrices. For any positive real numbers r,p,q>0r, p, q > 0 satisfying the relation 1r=1p+1q\frac{1}{r} = \frac{1}{p} + \frac{1}{q}, the Schatten rr-norm of the product ABAB is bounded by the product of the Schatten pp-norm of AA and the Schatten qq-norm of BB: ABSrASpBSq\|AB\|_{S^r} \le \|A\|_{S^p} \|B\|_{S^q} where the Schatten pp-norm is defined as MSp=(iσi(M)p)1/p\|M\|_{S^p} = \left( \sum_{i} \sigma_i(M)^p \right)^{1/p} for singular values σi(M)\sigma_i(M). This inequality holds for both the norm (p,q,r1)(p, q, r \ge 1) and quasi-norm (0<p,q,r<1)(0 < p, q, r < 1) cases.

theorem

Tr((A1/2BA1/2)α)(Tr(Ap/2)1/pTr(Bq)1/q)2α\text{Tr}((A^{1/2} B A^{1/2})^\alpha) \le (\text{Tr}(A^{p/2})^{1/p} \text{Tr}(B^q)^{1/q})^{2\alpha} for 0A,B0 \le A, B and 12α=1p+1q\frac{1}{2\alpha} = \frac{1}{p} + \frac{1}{q}

#trace_rpow_conj_le

Let AA and BB be positive semidefinite Hermitian matrices of size d×dd \times d over C\mathbb{C}. Let α,p,q\alpha, p, q be positive real numbers such that 12α=1p+1q\frac{1}{2\alpha} = \frac{1}{p} + \frac{1}{q}. Then the trace of the power of the matrix A1/2BA1/2A^{1/2}BA^{1/2} satisfies the inequality Tr((A1/2BA1/2)α)(Tr(Ap/2)1/pTr(Bq)1/q)2α,\text{Tr}((A^{1/2} B A^{1/2})^\alpha) \le \left( \text{Tr}(A^{p/2})^{1/p} \cdot \text{Tr}(B^q)^{1/q} \right)^{2\alpha}, where ArA^r denotes the rr-th power of a Hermitian matrix defined via functional calculus and A.conj B.matA. \text{conj } B. \text{mat} denotes the congruence action A1/2BA1/2A^{1/2} B A^{1/2}.