Physlib

PhyslibAlpha.QuantumMechanics.StinespringDilation

Stinespring dilation

56 declarations

definition

Kraus operator sum application ρiKiρKi\rho \mapsto \sum_i K_i \rho K_i^\dagger

Given a finite family of matrices {Ki}ir\{K_i\}_{i \in r} (known as Kraus operators) and a matrix ρ\rho, the function computes the sum irKiρKi\sum_{i \in r} K_i \rho K_i^\dagger, where KiK_i^\dagger denotes the conjugate transpose of KiK_i. This corresponds to the Kraus representation of a completely positive map acting on an operator ρ\rho.

theorem

The Kraus map preserves positive semi-definiteness

Let RR be a star-ring equipped with a partial order and q,rq, r be finite index sets. Let {Ki}ir\{K_i\}_{i \in r} be a finite collection of q×qq \times q matrices over RR. If ρ\rho is a q×qq \times q positive semi-definite matrix, then the result of the Kraus operator sum irKiρKi\sum_{i \in r} K_i \rho K_i^\dagger is also a positive semi-definite matrix, where KiK_i^\dagger denotes the conjugate transpose of KiK_i.

definition

Quantum channel condition iKiKi=I\sum_i K_i^\dagger K_i = I

A collection of matrices {Ki}ir\{K_i\}_{i \in r} in Mq(R)M_q(R) defines a quantum channel if they satisfy the completeness relation irKiKi=I\sum_{i \in r} K_i^\dagger K_i = I, where II is the q×qq \times q identity matrix and KiK_i^\dagger denotes the conjugate transpose of KiK_i.

definition

Quantum operation condition iKiKiI\sum_i K_i^\dagger K_i \le I

A collection of matrices {Ki}ir\{K_i\}_{i \in r} (Kraus operators) in Matq(R)\text{Mat}_q(R), where RR is a field such as R\mathbb{R} or C\mathbb{C}, defines a **quantum operation** if they satisfy the condition irKiKiI\sum_{i \in r} K_i^\dagger K_i \le I where KiK_i^\dagger denotes the conjugate transpose (adjoint) of KiK_i, and II is the identity matrix. The inequality \le denotes the Loewner order, signifying that the difference IiKiKiI - \sum_i K_i^\dagger K_i is a positive semi-definite matrix. This condition characterizes a trace-non-increasing completely positive map.

definition

Density matrix

Let RR be a ring equipped with a partial order and a star-ring structure, and dd be a finite index set. A density matrix ρ\rho is defined as a d×dd \times d matrix over RR that is positive semidefinite (ρ0\rho \succeq 0) and has a trace equal to 1 (Tr(ρ)=1\text{Tr}(\rho) = 1).

definition

Convex combination of density matrices ρ=tρ0+(1t)ρ1\rho = t \rho_0 + (1 - t) \rho_1

Let RR be a field (specifically an `RCLike` field such as R\mathbb{R} or C\mathbb{C}) and dd be a natural number representing the dimension. Given two d×dd \times d density matrices ρ0\rho_0 and ρ1\rho_1, and a scalar tRt \in R such that 0t0 \le t and 01t0 \le 1 - t (i.e., 0t10 \le t \le 1), their convex combination is a new density matrix defined as: ρ=tρ0+(1t)ρ1 \rho = t \rho_0 + (1 - t) \rho_1 This construction confirms that the set of density matrices is closed under real convex combinations, as the resulting matrix ρ\rho is positive semidefinite and satisfies Tr(ρ)=1\text{Tr}(\rho) = 1.

definition

Partial trace tr2\text{tr}_2 over the second system

Given a ring RR and a matrix ρ\rho whose rows are indexed by the Cartesian product m×nm \times n and columns by m×nm' \times n, the function tr2\text{tr}_2 (the partial trace over the second system) maps ρ\rho to a matrix with rows indexed by mm and columns by mm'. Its (i,j)(i, j)-th entry is defined by the sum over the second index: (tr2ρ)i,j=knρ(i,k),(j,k) (\text{tr}_2 \rho)_{i,j} = \sum_{k \in n} \rho_{(i, k), (j, k)} where nn is a finite index set.

definition

Stinespring operator VV

Let RR be a ring and m,rm, r be finite sets. Given a collection of matrices {Ki}ir\{K_i\}_{i \in r} where each KiMatm(R)K_i \in \text{Mat}_m(R), the Stinespring operator VV is a matrix in Matm×r,m(R)\text{Mat}_{m \times r, m}(R) defined such that for any x,ymx, y \in m and iri \in r, the entry V(x,i),yV_{(x, i), y} is equal to the (x,y)(x, y)-th entry of the matrix KiK_i. In block matrix notation, this is equivalent to V=ireiKiV = \sum_{i \in r} e_i \otimes K_i, where eie_i are the standard basis vectors of RrR^r.

theorem

The ((i,k),j)((i, k), j)-th entry of the Stinespring operator VV is (Kk)i,j(K_k)_{i, j}

Let RR be a ring, and let mm and rr be finite index sets. Given a collection of matrices {Kk}kr\{K_k\}_{k \in r} where each KkMatm(R)K_k \in \text{Mat}_m(R), let VV be the Stinespring operator. For any row index (i,k)m×r(i, k) \in m \times r and column index jmj \in m, the entry of the matrix VV is given by V(i,k),j=(Kk)i,j V_{(i, k), j} = (K_k)_{i, j} where (Kk)i,j(K_k)_{i, j} denotes the (i,j)(i, j)-th entry of the kk-th matrix KkK_k.

definition

Stinespring dilation VρVV \rho V^\dagger

Let RR be a star ring and m,rm, r be finite sets. Given a collection of matrices {Ki}ir\{K_i\}_{i \in r} in Matm(R)\text{Mat}_m(R) and a matrix ρMatm(R)\rho \in \text{Mat}_m(R), the Stinespring dilation of ρ\rho with respect to {Ki}\{K_i\} is the matrix in Matm×r(R)\text{Mat}_{m \times r}(R) defined as VρVV \rho V^\dagger, where VV is the Stinespring operator associated with the collection {Ki}\{K_i\}, and VV^\dagger denotes the conjugate transpose of VV.

definition

Stinespring form ρiKiρKi\rho \mapsto \sum_i K_i \rho K_i^\dagger

Let RR be a star ring and m,rm, r be finite sets. For a collection of matrices {Ki}ir\{K_i\}_{i \in r} in Matm(R)\text{Mat}_m(R), the Stinespring form is the operator that maps a matrix ρMatm(R)\rho \in \text{Mat}_m(R) to the partial trace over the second system of its Stinespring dilation. Mathematically, it is the map: ρtr2(VρV)=irKiρKi \rho \mapsto \text{tr}_2(V \rho V^\dagger) = \sum_{i \in r} K_i \rho K_i^\dagger where VV is the Stinespring operator associated with the collection {Ki}\{K_i\}, and tr2\text{tr}_2 is the partial trace over the index set rr.

theorem

iKiKi=VV\sum_i K_i^\dagger K_i = V^\dagger V for the Stinespring operator VV

Let RR be a star-ring and m,rm, r be finite sets. Given a collection of m×mm \times m matrices {Ki}ir\{K_i\}_{i \in r} over RR, let VV be the Stinespring operator (the matrix in Mat(m×r),m(R)\text{Mat}_{(m \times r), m}(R) formed by stacking the matrices KiK_i vertically). Then the sum of the products of the adjoints of KiK_i with KiK_i is equal to the product of the adjoint of VV with VV: irKiKi=VV\sum_{i \in r} K_i^\dagger K_i = V^\dagger V where MM^\dagger denotes the conjugate transpose (adjoint) of a matrix MM.

theorem

iKiKiI\sum_i K_i^\dagger K_i \le I implies the Stinespring operator VV satisfies VVIV^\dagger V \le I

Let RR be either the real or complex field, and let mm and rr be finite sets. Given a collection of matrices {Ki}ir\{K_i\}_{i \in r} in Matm(R)\text{Mat}_m(R) such that irKiKiI\sum_{i \in r} K_i^\dagger K_i \le I, where II is the identity matrix, let VMatm×r,m(R)V \in \text{Mat}_{m \times r, m}(R) be the Stinespring operator defined by the entries V(x,i),y=(Ki)x,yV_{(x, i), y} = (K_i)_{x, y} for x,ymx, y \in m and iri \in r. Then VVIV^\dagger V \le I.

theorem

The Stinespring operator of a CPTP map is an isometry

Let RR be a star-ring and m,rm, r be finite sets. Given a collection of matrices {Ki}ir\{K_i\}_{i \in r} in Matm(R)\text{Mat}_m(R) that satisfy the condition irKiKi=I\sum_{i \in r} K_i^\dagger K_i = I (where KiK_i^\dagger denotes the conjugate transpose and II is the identity matrix), the Stinespring operator VMatm×r,m(R)V \in \text{Mat}_{m \times r, m}(R) is an isometry, satisfying VV=IV^\dagger V = I.

theorem

The columns of the Stinespring operator VV are orthonormal if iKiKi=I\sum_i K_i^\dagger K_i = I

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be finite sets, and let {Ki}ir\{K_i\}_{i \in r} be a collection of matrices in Matm(R)\text{Mat}_m(R) satisfying the identity irKiKi=I\sum_{i \in r} K_i^\dagger K_i = I. Let VV be the Stinespring operator in Matm×r,m(R)\text{Mat}_{m \times r, m}(R) defined such that its entries are V(x,i),y=(Ki)x,yV_{(x, i), y} = (K_i)_{x, y} for iri \in r and x,ymx, y \in m. Then the columns of VV form an orthonormal set in Rm×rR^{m \times r}.

theorem

The dimension n|n| equals the cardinality of an extended orthonormal basis

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let nn and mm be finite types, and let ss be an n×mn \times m matrix with entries in RR. If the columns of ss form an orthonormal set in the Hilbert space RnR^n, then the cardinality of nn is equal to the cardinality of the orthonormal basis obtained by extending this set to the whole space.

theorem

The cardinality of the extended Stinespring orthonormal basis equals m×r|m \times r|

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be finite sets, and let {Ki}ir\{K_i\}_{i \in r} be a collection of matrices in Matm(R)\text{Mat}_m(R) satisfying the identity irKiKi=Im\sum_{i \in r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose. The columns of the Stinespring operator VMatm×r,m(R)V \in \text{Mat}_{m \times r, m}(R) form an orthonormal set in Rm×rR^{m \times r}. The cardinality of the orthonormal basis for Rm×rR^{m \times r} obtained by extending this orthonormal set is equal to the cardinality of the product set m×rm \times r.

theorem

The dimension of the orthogonal complement of the Stinespring operator's range is m(r1)m(r-1)

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers. Consider a collection of m×mm \times m matrices {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} that satisfies the condition iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose. Let VV be the Stinespring operator (an mr×mmr \times m matrix) whose entries are defined by V(x,i),y=(Ki)x,yV_{(x, i), y} = (K_i)_{x, y}. Let Range(V)\text{Range}(V) be the subspace of Rm×rR^{m \times r} spanned by the columns of VV, and let Range(V)\text{Range}(V)^\perp be its orthogonal complement. If uu' is an orthonormal basis for Range(V)\text{Range}(V)^\perp, then the cardinality of uu' is equal to the cardinality of the set Fin m×Fin (r1)\text{Fin } m \times \text{Fin } (r - 1), which is m(r1)m(r-1).

definition

Predecessor of iFin ni \in \text{Fin } n excluding kk

Given a natural number nn and indices k,ik, i in the set {0,,n1}\{0, \dots, n-1\} such that iki \neq k, the function returns an index in the set {0,,n2}\{0, \dots, n-2\} defined by shifting indices greater than kk down by one: f(i)={i1if i>kiif i<k f(i) = \begin{cases} i - 1 & \text{if } i > k \\ i & \text{if } i < k \end{cases}

theorem

Injectivity of `Fin.predAboveOfNe`

For any natural number nn and elements k,x,y{0,,n1}k, x, y \in \{0, \dots, n-1\}, let f:{iFin nik}{0,,n2}f: \{i \in \text{Fin } n \mid i \neq k\} \to \{0, \dots, n-2\} be the function defined by f(i)={iif i<ki1if i>k f(i) = \begin{cases} i & \text{if } i < k \\ i - 1 & \text{if } i > k \end{cases} If xkx \neq k, yky \neq k, and f(x)=f(y)f(x) = f(y), then x=yx = y. In other words, the function that maps an index ii (excluding kk) to its predecessor above kk is injective.

definition

Orthonormal basis vectors of (Range V)(\text{Range } V)^\perp

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers. Consider a collection of m×mm \times m matrices {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose. Let VV be the Stinespring operator, defined as the mr×mmr \times m matrix where V(a,i),b=(Ki)a,bV_{(a, i), b} = (K_i)_{a, b}. Let Range(V)\text{Range}(V)^\perp be the orthogonal complement of the subspace spanned by the columns of VV in Rm×rR^{m \times r}. Given a fixed index zFin rz \in \text{Fin } r, and an index x=(j,i)Fin m×Fin rx = (j, i) \in \text{Fin } m \times \text{Fin } r such that izi \neq z, `onbPart` returns the vector in Rm×rR^{m \times r} corresponding to the basis element u(j,fz(i))u_{(j, f_z(i))} of a chosen orthonormal basis {uw}\{u_w\} of Range(V)\text{Range}(V)^\perp. Here, fz(i)f_z(i) is the index in Fin (r1)\text{Fin } (r-1) obtained by shifting the index ii to skip zz.

theorem

Distinct `onbPart` vectors are orthogonal

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers, and let {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} be a collection of m×mm \times m matrices satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m. For a fixed index zFin rz \in \text{Fin } r, let x,yFin m×Fin rx, y \in \text{Fin } m \times \text{Fin } r be indices such that their second components are not equal to zz (i.e., x2zx_2 \neq z and y2zy_2 \neq z). If xyx \neq y, then the inner product of the vectors defined by `onbPart` for xx and yy is zero: onbPart(y),onbPart(x)=0 \langle \text{onbPart}(y), \text{onbPart}(x) \rangle = 0

theorem

onbPart=1\|\text{onbPart}\| = 1

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers. Consider a collection of m×mm \times m matrices {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose. Let VV be the Stinespring operator, defined as the mr×mmr \times m matrix where V(a,i),b=(Ki)a,bV_{(a, i), b} = (K_i)_{a, b}. For a fixed index zFin rz \in \text{Fin } r and an index x=(j,i)Fin m×Fin rx = (j, i) \in \text{Fin } m \times \text{Fin } r such that izi \neq z, the vector `onbPart` (which represents a basis element of the orthogonal complement (Range V)(\text{Range } V)^\perp) has a norm of 1, i.e., onbPart2=1\|\text{onbPart}\|_2 = 1.

definition

Unitary Stinespring dilation UU of {Ki}\{K_i\} at zz

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers. Consider a collection of m×mm \times m matrices {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose. For a fixed index zFin rz \in \text{Fin } r, the Stinespring dilation UU (also known as the unitary dilation) is an (mr)×(mr)(mr) \times (mr) matrix constructed as follows: For any row index x=(x1,x2)Fin m×Fin rx = (x_1, x_2) \in \text{Fin } m \times \text{Fin } r and column index y=(y1,y2)Fin m×Fin ry = (y_1, y_2) \in \text{Fin } m \times \text{Fin } r, the entries of UU are: Ux,(y1,y2)={(Kx2)x1,y1if y2=z(onbPart y)xif y2zU_{x, (y_1, y_2)} = \begin{cases} (K_{x_2})_{x_1, y_1} & \text{if } y_2 = z \\ (\text{onbPart } y)_x & \text{if } y_2 \neq z \end{cases} Here, the columns where y2=zy_2 = z form the Stinespring operator VV, and the columns where y2zy_2 \neq z are formed by the components of an orthonormal basis (onbPart y)(\text{onbPart } y) for the orthogonal complement of the range of VV. This construction results in a unitary (or orthogonal) matrix UU.

definition

General dilation matrix DD from SS and MM at zz

Let RR be a ring and m,rm, r be index sets. Given a fixed index zrz \in r, a matrix SS of size (m×r)×m(m \times r) \times m, and a matrix MM of size (m×r)×(m×r)(m \times r) \times (m \times r), the general dilation is the (m×r)×(m×r)(m \times r) \times (m \times r) matrix whose entries are defined as follows: for any row index xm×rx \in m \times r and column index y=(y1,y2)m×ry = (y_1, y_2) \in m \times r, the entry is Sx,y1S_{x, y_1} if y2=zy_2 = z, and Mx,yM_{x, y} if y2zy_2 \neq z. Effectively, this construction replaces the zz-th block-column of MM with the matrix SS.

definition

General dilation matrix DD from {Ki}\{K_i\} and MM at zz

Let RR be a ring and m,rm, r be index sets where rr is finite. Given a collection of matrices {Ki}ir\{K_i\}_{i \in r} where each KiMatm(R)K_i \in \text{Mat}_m(R), a fixed index zrz \in r, and a matrix MMatm×r(R)M \in \text{Mat}_{m \times r}(R), the dilation is the (m×r)×(m×r)(m \times r) \times (m \times r) matrix DD obtained by replacing the zz-th block-column of MM with the Stinespring operator VV associated with {Ki}\{K_i\}. Specifically, for any row index x=(x1,x2)m×rx = (x_1, x_2) \in m \times r and column index y=(y1,y2)m×ry = (y_1, y_2) \in m \times r, the entries of the matrix are defined as: D(x1,x2),(y1,y2)={(Kx2)x1,y1if y2=zM(x1,x2),(y1,y2)if y2zD_{(x_1, x_2), (y_1, y_2)} = \begin{cases} (K_{x_2})_{x_1, y_1} & \text{if } y_2 = z \\ M_{(x_1, x_2), (y_1, y_2)} & \text{if } y_2 \neq z \end{cases} This construction represents a general, not necessarily unitary, dilation of the operators {Ki}\{K_i\}.

theorem

The columns of the Stinespring dilation matrix are orthonormal

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let m,rm, r be natural numbers and {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} be a collection of m×mm \times m matrices over RR satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m. Let VV be the Stinespring operator, defined as the (m×r)×m(m \times r) \times m matrix such that V(a,i),b=(Ki)a,bV_{(a, i), b} = (K_i)_{a, b}. For a fixed index zFin rz \in \text{Fin } r, define a collection of vectors {vy}yFin m×Fin r\{v_y\}_{y \in \text{Fin } m \times \text{Fin } r} in Rm×rR^{m \times r} as follows: - If y=(j,z)y = (j, z) for some jFin mj \in \text{Fin } m, then vyv_y is the jj-th column of the Stinespring operator VV. - If y=(j,i)y = (j, i) for izi \neq z, then vyv_y is the vector provided by `onbPart`, which belongs to a chosen orthonormal basis of the orthogonal complement (Range V)(\text{Range } V)^\perp. Then the collection of vectors {vy}yFin m×Fin r\{v_y\}_{y \in \text{Fin } m \times \text{Fin } r} is orthonormal.

theorem

Stinespring Dilation Columns Form an Orthonormal Basis

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers. Suppose {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} is a collection of m×mm \times m matrices over RR that satisfy the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where ImI_m is the m×mm \times m identity matrix and KiK_i^\dagger is the conjugate transpose of KiK_i. Let VV be the Stinespring operator, defined as the mr×mmr \times m matrix with entries V(a,i),b=(Ki)a,bV_{(a, i), b} = (K_i)_{a, b} for a,bFin ma, b \in \text{Fin } m and iFin ri \in \text{Fin } r. For a fixed index zFin rz \in \text{Fin } r, we define a set of vectors {vy}yFin m×Fin r\{v_y\}_{y \in \text{Fin } m \times \text{Fin } r} in RmrR^{mr} such that: - If y.2=zy.2 = z, vyv_y is the y.1y.1-th column of the Stinespring operator VV. - If y.2zy.2 \neq z, vyv_y is the basis vector of the orthogonal complement (Range V)(\text{Range } V)^\perp defined by `onbPart`. Then the collection of vectors {vy}\{v_y\} is orthonormal with respect to the standard L2L^2 inner product.

theorem

β2=1    xβ(x)β(x)=1\|\beta\|_2 = 1 \implies \sum_x \overline{\beta(x)} \beta(x) = 1

Let RR be a field that is either the real numbers R\mathbb{R} or the complex numbers C\mathbb{C} (specifically, an `RCLike` field). For a finite set tt and a function β:tR\beta : t \to R, if the L2L^2-norm of β\beta is equal to 1, then the sum of the products of the conjugates of its components with the components themselves is 1. That is, if β2=1\|\beta\|_2 = 1, then xtβ(x)β(x)=1\sum_{x \in t} \overline{\beta(x)} \beta(x) = 1 where β(x)\overline{\beta(x)} denotes the conjugate of β(x)\beta(x) in RR.

theorem

Orthonormality of rows implies αα=I\alpha \alpha^* = I

Let RR be a field that is either the real or complex numbers (R\mathbb{R} or C\mathbb{C}). Let mm and rr be finite index sets, and let α\alpha be a square matrix with entries in RR indexed by (m×r)×(m×r)(m \times r) \times (m \times r). If the rows of α\alpha (viewed as vectors with the standard L2L^2 inner product) form an orthonormal set, then αα=I\alpha \alpha^* = I, where α\alpha^* denotes the conjugate transpose of α\alpha and II is the identity matrix.

theorem

The transpose UTU^T of the Stinespring dilation is unitary

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers and {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} be a collection of m×mm \times m matrices over RR satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose. For any fixed index zFin rz \in \text{Fin } r, let UU be the Stinespring dilation (also known as the unitary dilation) associated with {Ki}\{K_i\} at zz. Then the transpose of UU, denoted UTU^T, is a unitary matrix.

theorem

The Stinespring dilation UU is unitary

Let RR be a field that is either the real numbers R\mathbb{R} or the complex numbers C\mathbb{C}. Let mm and rr be natural numbers. Suppose {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} is a collection of m×mm \times m matrices over RR that satisfies the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, where KiK_i^\dagger denotes the conjugate transpose and ImI_m is the m×mm \times m identity matrix. For any fixed index zFin rz \in \text{Fin } r, the Stinespring dilation matrix UU (constructed from {Ki}\{K_i\} and zz) is a unitary matrix. That is, UU=UU=ImrU U^\dagger = U^\dagger U = I_{mr}.

theorem

tr2(ρe)=ρ\text{tr}_2(\rho \otimes e) = \rho for tr(e)=1\text{tr}(e) = 1

Let RR be a field of real or complex numbers. Let ee be a square matrix indexed by a finite set ww, and let ρ\rho be a square matrix indexed by a set mm. If the trace of ee is equal to 1, then the partial trace over the second system of the Kronecker product ρe\rho \otimes e is equal to ρ\rho. That is, if tr(e)=1\text{tr}(e) = 1, then: tr2(ρe)=ρ \text{tr}_2(\rho \otimes e) = \rho

definition

Unitary Stinespring form tr2(U(ρEz,z)U)\text{tr}_2(U(\rho \otimes E_{z,z})U^\dagger)

Let RR be a field that is either R\mathbb{R} or C\mathbb{C} and m,rm, r be natural numbers. Given a collection of m×mm \times m matrices {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} satisfying the completeness relation iFin rKiKi=I\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I, and a fixed index zFin rz \in \text{Fin } r, the Stinespring unitary form of an m×mm \times m matrix ρ\rho is defined as: tr2(U(ρEz,z)U) \text{tr}_2 \left( U (\rho \otimes E_{z,z}) U^\dagger \right) where UU is the unitary Stinespring dilation of the Kraus operators {Ki}\{K_i\} at the index zz, Ez,zE_{z,z} is the r×rr \times r matrix with the entry 11 at (z,z)(z, z) and 00 elsewhere, and tr2\text{tr}_2 is the partial trace over the second system.

definition

General Stinespring Unitary Form tr2(U(ρe)U)\text{tr}_2(U (\rho \otimes e) U^\dagger)

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers. Given a collection of m×mm \times m matrices {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, and a fixed reference index zFin rz \in \text{Fin } r, let UU be the unitary Stinespring dilation matrix (as defined by `Ud`). For any m×mm \times m matrix ρ\rho and r×rr \times r matrix ee, the general Stinespring unitary form is the m×mm \times m matrix defined by: tr2(U(ρe)U) \text{tr}_2(U (\rho \otimes e) U^\dagger) where \otimes denotes the Kronecker product, UU^\dagger is the conjugate transpose of UU, and tr2\text{tr}_2 is the partial trace over the second system.

theorem

Trace-free Stinespring Dilation: U(ρIr)U=iKiρKiU^\dagger (\rho \otimes I_r) U = \sum_i K_i \rho K_i^\dagger

Let RR be a field (either R\mathbb{R} or C\mathbb{C}), and let mm and rr be finite index sets. Given a collection of m×mm \times m matrices {Ki}irMatm(R)\{K_i\}_{i \in r} \subseteq \text{Mat}_m(R), define the Stinespring operator U=ireiKiU = \sum_{i \in r} e_i \otimes K_i^\dagger, where KiK_i^\dagger is the conjugate transpose of KiK_i and eie_i are the standard basis vectors. For any m×mm \times m matrix ρ\rho, the following equality holds: U(ρIr)U=irKiρKi U^\dagger (\rho \otimes I_r) U = \sum_{i \in r} K_i \rho K_i^\dagger where IrI_r is the r×rr \times r identity matrix and UU^\dagger denotes the conjugate transpose of UU.

theorem

Schrödinger-Heisenberg Equivalence for Stinespring Dilation: tr2(VρV)=U(ρIr)U\text{tr}_2(V \rho V^\dagger) = U^\dagger (\rho \otimes I_r) U

Let RR be a field (either R\mathbb{R} or C\mathbb{C}), and let mm and rr be finite index sets. Consider a collection of m×mm \times m matrices {Ki}irMatm(R)\{K_i\}_{i \in r} \subseteq \text{Mat}_m(R). Define the Stinespring operator V=ireiKiV = \sum_{i \in r} e_i \otimes K_i and the dual Stinespring operator U=ireiKiU = \sum_{i \in r} e_i \otimes K_i^\dagger, where KiK_i^\dagger denotes the conjugate transpose of KiK_i. For any m×mm \times m matrix ρ\rho, the Schrödinger picture representation, given by the partial trace tr2(VρV)\text{tr}_2(V \rho V^\dagger), is equal to the Heisenberg picture representation U(ρIr)UU^\dagger (\rho \otimes I_r) U, where IrI_r is the r×rr \times r identity matrix.

definition

General Stinespring dilation tr2(U(ρEz,z)U)\text{tr}_2(U (\rho \otimes E_{z,z}) U^\dagger)

For a field RR (either R\mathbb{R} or C\mathbb{C}) and finite index sets mm and rr, let zrz \in r be a reference index, SS be a matrix of size (m×r)×m(m \times r) \times m, and MM be a matrix of size (m×r)×(m×r)(m \times r) \times (m \times r). This definition represents a map that transforms an m×mm \times m matrix ρ\rho as: Φ(ρ)=tr2(U(ρEz,z)U) \Phi(\rho) = \text{tr}_2 \left( U (\rho \otimes E_{z,z}) U^\dagger \right) where UU is the dilation matrix obtained by replacing the zz-th block-column of MM with SS, Ez,zE_{z,z} is the r×rr \times r matrix unit with 11 at the (z,z)(z, z) entry and 00 elsewhere, and tr2\text{tr}_2 denotes the partial trace over the second subsystem rr.

definition

General Stinespring dilation tr2(U(ρEz,z)U)\text{tr}_2(U (\rho \otimes E_{z,z}) U^\dagger) from {Ki}\{K_i\} and MM

Let RR be a field (either R\mathbb{R} or C\mathbb{C}), and let mm and rr be finite index sets. Given a collection of m×mm \times m matrices {Ki}ir\{K_i\}_{i \in r}, a reference index zrz \in r, and a base matrix MM of size (m×r)×(m×r)(m \times r) \times (m \times r), the Stinespring general form is a map that transforms an m×mm \times m matrix ρ\rho as: Φ(ρ)=tr2(U(ρEz,z)U) \Phi(\rho) = \text{tr}_2 \left( U (\rho \otimes E_{z,z}) U^\dagger \right) where: - UU is the dilation matrix of size (m×r)×(m×r)(m \times r) \times (m \times r) obtained by replacing the zz-th block-column of MM with the Stinespring operator associated with {Ki}\{K_i\} (specifically, the entry U(x1,x2),(y1,z)U_{(x_1, x_2), (y_1, z)} is given by (Kx2)x1,y1(K_{x_2})_{x_1, y_1}). - Ez,zE_{z,z} is the r×rr \times r matrix unit with 11 at the (z,z)(z, z) entry and 00 elsewhere. - \otimes denotes the Kronecker product. - tr2\text{tr}_2 denotes the partial trace over the second subsystem rr. - UU^\dagger is the conjugate transpose of UU.

definition

General Stinespring map ρtr2(U(ρe)U)\rho \mapsto \text{tr}_2(U (\rho \otimes e) U^\dagger)

Let RR be either the real numbers R\mathbb{R} or the complex numbers C\mathbb{C}, and let mm and rr be finite index sets. Given a collection of matrices {Ki}ir\{K_i\}_{i \in r} where each KiMatm(R)K_i \in \text{Mat}_m(R), a fixed reference index zrz \in r, an "environment" matrix eMatr(R)e \in \text{Mat}_r(R), and a background matrix MMatm×r(R)M \in \text{Mat}_{m \times r}(R), this definition represents a map Φ:Matm(R)Matm(R)\Phi: \text{Mat}_m(R) \to \text{Mat}_m(R) that transforms an input matrix ρ\rho as: Φ(ρ)=tr2(U(ρe)U) \Phi(\rho) = \text{tr}_2 \left( U (\rho \otimes e) U^\dagger \right) where: - UU is the (m×r)×(m×r)(m \times r) \times (m \times r) dilation matrix obtained by replacing the zz-th block-column of MM with the Stinespring operator associated with {Ki}\{K_i\}. - \otimes denotes the Kronecker product of matrices. - UU^\dagger is the conjugate transpose of UU. - tr2\text{tr}_2 is the partial trace over the second subsystem (indexed by rr). This is a general form of the Stinespring dilation where the environment state is represented by an arbitrary matrix ee.

theorem

`stinespringGeneralForm` with M=UM = U equals `stinespringUnitaryForm`

Let RR be a field (either R\mathbb{R} or C\mathbb{C}), and let m,rm, r be natural numbers. Let {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} be a collection of m×mm \times m matrices satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m. For a fixed index zFin rz \in \text{Fin } r, let UU be the unitary Stinespring dilation of the Kraus operators {Ki}\{K_i\} at zz (denoted by `Ud`). Then the general Stinespring form evaluated with the background matrix M=UM = U is equal to the unitary Stinespring form: stinespringGeneralForm(K,z,U)=stinespringUnitaryForm(hK,z) \text{stinespringGeneralForm}(K, z, U) = \text{stinespringUnitaryForm}(hK, z) where both sides represent the map ρtr2(U(ρEz,z)U)\rho \mapsto \text{tr}_2 \left( U (\rho \otimes E_{z,z}) U^\dagger \right).

theorem

General Stinespring form with environment ee equals the unitary form for U=UdU = \text{Ud}

Let RR be a field that is either the real numbers R\mathbb{R} or the complex numbers C\mathbb{C}, and let mm and rr be natural numbers. Suppose {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} is a collection of m×mm \times m matrices satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m. For a fixed reference index zFin rz \in \text{Fin } r and an r×rr \times r environment matrix ee, the general Stinespring map (which transforms a matrix ρ\rho via the environment ee and a dilation matrix UU) evaluated at the specific unitary dilation U=UdU = \text{Ud} is equal to the unitary form of the Stinespring dilation. That is, stinespringGeneralFormE(K,z,e,Ud)=stinespringUnitaryFormE(K,z,e), \text{stinespringGeneralFormE}(K, z, e, \text{Ud}) = \text{stinespringUnitaryFormE}(K, z, e), where both sides represent the operation ρtr2(U(ρe)U)\rho \mapsto \text{tr}_2(U (\rho \otimes e) U^\dagger).

theorem

The general Stinespring dilation form equals the Kraus map iKiρKi\sum_i K_i \rho K_i^\dagger

Let RR be a field (such as R\mathbb{R} or C\mathbb{C}), and let mm and rr be natural numbers. Given a collection of m×mm \times m matrices {Ki}i=0r1\{K_i\}_{i=0}^{r-1} over RR, a reference index z{0,,r1}z \in \{0, \dots, r-1\}, and an arbitrary matrix MM of size mr×mrmr \times mr, let UU be the dilation matrix obtained by replacing the zz-th block-column of MM with the operators KiK_i. The Stinespring general form map Φ\Phi, defined by Φ(ρ)=tr2(U(ρEz,z)U) \Phi(\rho) = \text{tr}_2 \left( U (\rho \otimes E_{z,z}) U^\dagger \right) where tr2\text{tr}_2 is the partial trace over the second subsystem and Ez,zE_{z,z} is the r×rr \times r matrix unit with 11 at the (z,z)(z, z) entry, is equal to the Kraus map: Φ(ρ)=i=0r1KiρKi \Phi(\rho) = \sum_{i=0}^{r-1} K_i \rho K_i^\dagger This identity holds for any matrix MM and any collection {Ki}\{K_i\}, without requiring MM to be unitary or the map to be trace-preserving.

theorem

The unitary Stinespring dilation map equals the Kraus map iKiρKi\sum_i K_i \rho K_i^\dagger

Let RR be a field that is either R\mathbb{R} or C\mathbb{C}. Let mm and rr be natural numbers, and let {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} be a collection of m×mm \times m matrices over RR satisfying the completeness relation iFin rKiKi=I\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I. For any fixed index zFin rz \in \text{Fin } r representing the ancilla coordinate, the unitary Stinespring dilation map (defined using a unitary dilation UU of the Kraus operators) is equal to the Kraus map. That is, for any m×mm \times m matrix ρ\rho: tr2(U(ρEz,z)U)=iFin rKiρKi \text{tr}_2 \left( U (\rho \otimes E_{z,z}) U^\dagger \right) = \sum_{i \in \text{Fin } r} K_i \rho K_i^\dagger where Ez,zE_{z,z} is the r×rr \times r matrix with 11 at the (z,z)(z, z) entry and 00 elsewhere, and tr2\text{tr}_2 is the partial trace over the second subsystem.

definition

Kraus completion V~\tilde{V} of a CPTNI map

Let RR be a field such as R\mathbb{R} or C\mathbb{C} (specifically an `RCLike` field). Given a collection of rr matrices {Ki}i=0r1\{K_i\}_{i=0}^{r-1} where each KiMatm(R)K_i \in \text{Mat}_m(R), the Kraus completion V~\tilde{V} is a matrix in Matm(r+1),m(R)\text{Mat}_{m(r+1), m}(R). It is defined as a block matrix where the first rr blocks are the matrices KiK_i and the final block is the matrix square root of the difference between the identity and the sum of KiKiK_i^\dagger K_i: V~=(K0Kr1Ii=0r1KiKi) \tilde{V} = \begin{pmatrix} K_0 \\ \vdots \\ K_{r-1} \\ \sqrt{I - \sum_{i=0}^{r-1} K_i^\dagger K_i} \end{pmatrix} where II is the m×mm \times m identity matrix and KiK_i^\dagger denotes the conjugate transpose. This matrix provides the "orthogonal" completely positive trace preserving (CPTP) completion of the completely positive trace non-increasing (CPTNI) map defined by the Kraus operators KiK_i.

theorem

VV=iKiKiV^* V = \sum_i K_i^* K_i for the Stinespring operator VV

Let RR be either the real numbers R\mathbb{R} or the complex numbers C\mathbb{C}. For a collection of m×mm \times m matrices {Ki}ir\{K_i\}_{i \in r} over RR, let VV be the Stinespring operator defined by the block matrix V=ireiKiV = \sum_{i \in r} e_i \otimes K_i. Then the Gram matrix of VV satisfies the identity: VV=irKiKiV^* V = \sum_{i \in r} K_i^* K_i where VV^* and KiK_i^* denote the conjugate transpose (adjoint) of the matrices VV and KiK_i, respectively.

theorem

The Kraus Completion V~\tilde{V} of a Trace Non-Increasing Map is an Isometry (V~V~=I\tilde{V}^\dagger \tilde{V} = I)

Let RR be an `RCLike` field (such as R\mathbb{R} or C\mathbb{C}). Consider a collection of rr matrices {Ki}i=0r1Matm(R)\{K_i\}_{i=0}^{r-1} \in \text{Mat}_m(R) that satisfy the trace non-increasing (TNI) condition i=0r1KiKiI\sum_{i=0}^{r-1} K_i^\dagger K_i \le I, where II is the m×mm \times m identity matrix. Then the Kraus completion V~\tilde{V} of these matrices—defined as the block matrix formed by stacking K0,,Kr1K_0, \dots, K_{r-1} and the matrix square root IKiKi\sqrt{I - \sum K_i^\dagger K_i}—is an isometry, satisfying V~V~=I\tilde{V}^\dagger \tilde{V} = I.

definition

Unital condition iKiKi=I\sum_i K_i K_i^* = I

For a collection of m×mm \times m matrices {Ki}i=0r1\{K_i\}_{i=0}^{r-1} over R\mathbb{R} or C\mathbb{C}, the property of being unital is defined by the condition that the sum of the products of each matrix and its conjugate transpose equals the identity matrix: i=0r1KiKi=I\sum_{i=0}^{r-1} K_i K_i^* = I where KiK_i^* denotes the conjugate transpose (adjoint) of KiK_i and II is the m×mm \times m identity matrix.

definition

Subunital condition iKiKiI\sum_i K_i K_i^\dagger \le I

A family of matrices {Ki}i{0,,r1}\{K_i\}_{i \in \{0, \dots, r-1\}} of size m×mm \times m over a field RR (where RR is R\mathbb{R} or C\mathbb{C}) is **subunital** if the sum of the products of each matrix and its conjugate transpose is less than or equal to the identity matrix, i.e., iKiKiI\sum_i K_i K_i^\dagger \le I.

theorem

tr2(AB)=(Tr B)A\text{tr}_2 (A \otimes B) = (\text{Tr } B) A

For any matrices AMatm(R)A \in \text{Mat}_m(R) and BMatn(R)B \in \text{Mat}_n(R), where RR is the field of real or complex numbers, the partial trace over the second system of their Kronecker product is equal to the trace of BB multiplied by AA. That is, tr2(AB)=(Tr B)A \text{tr}_2 (A \otimes B) = (\text{Tr } B) A where \otimes denotes the Kronecker product and tr2\text{tr}_2 is the partial trace over the second subsystem.

theorem

U(ρEzz)U=αβ    iKiρKi=αU (\rho \otimes E_{zz}) U^\dagger = \alpha \otimes \beta \implies \sum_i K_i \rho K_i^\dagger = \alpha

Let RR be the field of real or complex numbers (R\mathbb{R} or C\mathbb{C}). Let {Ki}iFin r\{K_i\}_{i \in \text{Fin } r} be a collection of m×mm \times m matrices satisfying the completeness relation iFin rKiKi=Im\sum_{i \in \text{Fin } r} K_i^\dagger K_i = I_m, and let UU be the unitary Stinespring dilation of this collection at a fixed index zFin rz \in \text{Fin } r. Let ρ\rho and α\alpha be m×mm \times m matrices, and let β\beta be an r×rr \times r matrix with unit trace (Tr(β)=1\text{Tr}(\beta) = 1). If the unitary evolution of the system ρ\rho coupled with an ancilla in state EzzE_{zz} (the matrix with 11 at (z,z)(z, z) and 00 elsewhere) results in a product state αβ\alpha \otimes \beta, such that U(ρEzz)U=αβ,U (\rho \otimes E_{zz}) U^\dagger = \alpha \otimes \beta, then α\alpha is the result of the Kraus map applied to ρ\rho: iFin rKiρKi=α.\sum_{i \in \text{Fin } r} K_i \rho K_i^\dagger = \alpha.

theorem

Tr(ρ)=Tr(tr2ρ)\text{Tr}(\rho) = \text{Tr}(\text{tr}_2 \rho)

Let RR be a field such as R\mathbb{R} or C\mathbb{C}. For any matrix ρ\rho whose rows and columns are indexed by the Cartesian product Fin m×Fin n\text{Fin } m \times \text{Fin } n, the trace of the matrix ρ\rho is equal to the trace of its partial trace over the second system tr2ρ\text{tr}_2 \rho: Tr(ρ)=Tr(tr2ρ) \text{Tr}(\rho) = \text{Tr}(\text{tr}_2 \rho)

definition

Kraus completion map from quantum operations to channels

Let RR be a field such as R\mathbb{R} or C\mathbb{C} (specifically an `RCLike` field). Given a collection of rr matrices {Ki}i=0r1\{K_i\}_{i=0}^{r-1} in Matq(R)\text{Mat}_q(R) that satisfy the quantum operation condition i=0r1KiKiI\sum_{i=0}^{r-1} K_i^\dagger K_i \le I (where II is the q×qq \times q identity matrix and \dagger denotes the conjugate transpose), this function constructs an augmented collection of r+1r+1 matrices {Kj}j=0r\{K'_j\}_{j=0}^{r} that satisfy the quantum channel condition j=0r(Kj)Kj=I\sum_{j=0}^{r} (K'_j)^\dagger K'_j = I. This is achieved by setting Kj=KjK'_j = K_j for j{0,,r1}j \in \{0, \dots, r-1\} and defining the final operator KrK'_r as the matrix square root of the deficit Ii=0r1KiKiI - \sum_{i=0}^{r-1} K_i^\dagger K_i.

theorem

Completion of a CPTNI Map into a CPTP Map

Let RR be a field such as R\mathbb{R} or C\mathbb{C}. Suppose a collection of rr matrices {Ki}i=0r1\{K_i\}_{i=0}^{r-1} in Matq(R)\text{Mat}_q(R) defines a quantum operation, meaning they satisfy the condition i=0r1KiKiI\sum_{i=0}^{r-1} K_i^\dagger K_i \le I. Then there exists an extended collection of r+1r+1 matrices {Kj}j=0r\{K'_j\}_{j=0}^r that defines a quantum channel, satisfying j=0r(Kj)Kj=I\sum_{j=0}^{r} (K'_j)^\dagger K'_j = I, such that the original matrices are preserved for all indices i<ri < r, i.e., Ki=KiK'_i = K_i.

definition

Left partial trace Trm(ρ)\text{Tr}_m(\rho)

The partial trace on the left is a function that takes a matrix ρ\rho of size (m×n)×(m×n)(m \times n) \times (m \times n) over the field RR (where RR is R\mathbb{R} or C\mathbb{C}) and returns a matrix of size n×nn \times n. For any indices i,jni, j \in n, the (i,j)(i, j)-th entry of the resulting matrix is defined by the sum over the first index component: (Trmρ)i,j=kmρ(k,i),(k,j) (\text{Tr}_m \rho)_{i,j} = \sum_{k \in m} \rho_{(k, i), (k, j)} This operation corresponds to "tracing out" the first subsystem (indexed by mm) of a composite system.

theorem

Partial Trace of Stinespring Dilation Equals Kraus Map Application

Let RR be a field such as R\mathbb{R} or C\mathbb{C}. For any natural numbers mm and rr, let ρMatm(R)\rho \in \text{Mat}_m(R) be a matrix and {Ki}i=0r1\{K_i\}_{i=0}^{r-1} be a collection of matrices in Matm(R)\text{Mat}_m(R). The partial trace over the second system of the Stinespring dilation of ρ\rho with respect to {Ki}\{K_i\} is equal to the Kraus representation application of {Ki}\{K_i\} to ρ\rho. That is, tr2(VρV)=i=0r1KiρKi \text{tr}_2(V \rho V^\dagger) = \sum_{i=0}^{r-1} K_i \rho K_i^\dagger where VV is the Stinespring operator associated with the collection {Ki}\{K_i\}, and VV^\dagger is its conjugate transpose.