A density operator is a positive self-adjoint trace-class operator with . A finite expectation is . It represents a pure state exactly when for a unit vector, equivalently or .
For a finite-dimensional spectral decomposition , introduce an auxiliary space with orthonormal vectors and use the enlarged space . DefineThen and . This purification shows that subsystem mixedness can arise from entanglement with an unobserved environment even when the total state is pure. Subsystem measurements alone cannot distinguish these two descriptions.
A nonzero pure bipartite vector is entangled when it cannot be written as a tensor product of two vectors; equivalently its Schmidt rank exceeds one. For mixed bipartite states, entanglement means absence of a convex decomposition into product density operators.
For the nonorthogonal expansion in the question, the unnormalized reduced operator isIndeed . Positivity follows because , and . Thus is a density operator; if is already normalized, . The Gram factors cannot be dropped merely because the vectors are called bases. For unnormalized , physical expectations likewise divide its raw quadratic form by .
For the Schmidt decomposition diagonalize in an orthonormal basis with its positive eigenvalues . Expand with . The reduced matrix gives . Therefore are orthonormal andZero-eigenvalue components vanish because their squared norms are zero. This proves the decomposition for an arbitrary nonzero vector; the zero vector uses the empty sum. For a unit vector .
In the numerical example the coefficient matrix in the two product bases is . Taking the first-subsystem trace yieldsThe vector is in its nullspace, so rank is at most two; its leading minor has determinant one, so . This is the rank of and the Schmidt rank, proving entanglement. The displayed source ket has norm squared , and the displayed reduced matrix has trace . The physical density operator is ; its nonzero eigenvalues are , since . The raw matrix is correct for the unnormalized quadratic form actually printed.
A density operator is a positive self-adjoint trace-class operator with trace one. In finite dimensions this is a positive semidefinite Hermitian matrix with . Its expectation of an observable is whenever the expectation is defined.
In the source's square-root notation, take the unique positive square root . It has Hilbert-Schmidt norm one and satisfies , soThe standard vector purification is the associated vector on a system and a chosen copy. Tracing out the copy gives ; measuring gives the same trace formula. Thus the operator square root and the purified vector are two related objects, distinguished by this vectorization.
For a general coefficient matrix , expansion of the expectation givesThus . Positivity is immediate from , and its trace is one by normalization of .
If is nonsingular, it is square and is positive definite. DefineThen , which for a square matrix also implies . Thus is unitary, , and .
In an eigenbasis of , its square root is . Hence . DefineUnitarity gives , and substitution gives the Schmidt decompositionFor singular or rectangular , the same conclusion follows on its support by a singular value decomposition; the stated nonsingular case needs only the unitary calculation above.
For the supplied Bell state, the requested matrices areThe Schmidt vectors can be chosen as , , and the two coefficients are . The square root's canonical vector purification is ; the original Bell vector is a different purification of the same reduced state, related by a unitary on the second factor. This illustrates the unitary freedom of purification.
Use an expansion for the eigenstate continuing the nondegenerate level, with intermediate normalization :Insert these expressions into the eigenvalue equation. At first order,Taking the inner product with eliminates the left side by self-adjointness. ThusFor , projection onto instead givesAt second order, projecting the eigenvalue equation onto gives . Expand in the orthonormal energy basis and use self-adjointness of :Nondegeneracy ensures the denominators for this level do not vanish; convergence and validity of the expansion are among the assumed perturbation-theory hypotheses.
For the cosine perturbation of the ground level in a one-dimensional box, write . The product-to-sum identity givesso, with the given common normalization of the sine eigenfunctions,The diagonal matrix element is , and the only nonzero off-diagonal matrix element out of the ground state is . Let . Since , the ground energy isThe second-order term is negative, as expected for a nondegenerate ground level: every excited-quantum state denominator is negative. The first-order shift is also negative for positive , because the ground-quantum state density is largest near the center, where this perturbing potential is negative.
Use the Dirac notation convention that the inner product is linear in the ket. Orthonormality of the computational basis givesThe finite geometric sum satisfies if in , and zero otherwise: in the second case it equals . ThereforeThere are such orthonormal vectors in an -dimensional space, so they form a basis. The inverse expansion is verified by the same sum:This is the unitary discrete Fourier transform and its inverse.
For the basis-copying isometry, the images are orthonormal, so is a linear isometry, with . Its matrix elements giveApplying the definition of the adjoint operator, the coefficient of in is exactly this expression. ThusEquivalently . The adjoint of basis copying acts as pointwise multiplication of coefficient functions.
For the Fourier transform of a basis-copying isometry, the normalization matters:Hence . Taking adjoint matrix elements, or expanding both output factors in the inverse Fourier basis, yieldsThe printed formula is missing the factor . There are mutually orthogonal terms in its sum, so without that factor its right side has norm , whereas its left side has norm one because is an isometry. The formula above is the normalized identity consistent with all the preceding definitions. For the distinction disappears, but the no-cloning assertion is substantive for .
The normalized output is not . Its coefficient matrix pairs every Fourier label with ; tracing out either factor gives . Thus it is a maximally entangled state, while two clones of a pure Fourier quantum state would have pure reduced quantum states. A copier for the computational basis therefore does not copy the Fourier basis.
More generally, suppose a linear isometry with a fixed blank input could clone every normalized quantum state. Its restriction to would define a linear map with on every computational basis quantum state. Linearity forces , which has just been shown to fail on Fourier quantum states. This proves the no-cloning theorem in the pure-quantum state isometric setting.
The inner-product proof makes the obstruction independent of the chosen basis. If an isometry cloned two quantum states with overlap , preservation of inner products would give , or, allowing irrelevant output phases, . Distinct nonorthogonal rays have , which is impossible. Even allowing an output environment does not help: unitary evolution would give , the same contradiction. Arbitrary unknown quantum states cannot be cloned, although any fixed orthonormal basis can be copied and a one-dimensional quantum state space is trivial.
