This action might not be possible to undo. Are you sure you want to continue?

Dnes Petz e

Alfrd Rnyi Institute of Mathematics, Hungarian Academy of Sciences, POB 127, e e H-1364 Budapest, Hungary petz@renyi.hu

1 Hilbert spaces

The starting point of the quantum mechanical formalism is the Hilbert space. The Hilbert space is a mathematical concept, it is a space in the sense that it is a complex vector space which is endowed by an inner or scalar product h ¡ Y ¡ i. The linear space Cn of all n-tuples of complex numbers becomes a Hilbert space with the inner product

where z denotes the complex conjugate of the complex number z P C. Another example is the space of square integrable complex-valued function on the real Euclidean space Rn . If f and g are such functions then

yQ Ty U n T U hxY yi = xi yi = [x Y x Y XX X xn ] T X U Y T U R X S i yn

1 2 1 2 =1

P

hfY gi =

gives the inner product. The latter space is denoted by v2 (Rn ) and it is in

nite dimensional contrary to the n-dimensional space Cn . Below we are mostly satis

ed with .

On the other hand. The inner product of the vectors jxi and jy i will be often denoted as hxjy i. is popular in physics.nite dimensional spaces. sometimes called bra and ket. this notation. jxihy j is a linear operator which acts on the vector jz i as Rn f (x) g(x) dx jxihyj jzi = jxi hyjzi hyjzi jxiX ¡ Therefore. .

1 Show that kx + yk kxk + kyk X de. hxY xi ! 0 for every x P r and hxY xi = 0 only for x = 0.2 Dnes Petz e P 1 2 xQ Tx U T U jxihyj = T X U [y Y y Y XX X yn ] T U R X S xn is conjugate linear in jy i. (! P CY xY y P r). hxY yi = hyY xi (xY y P r). A further requirement in the Exercise 1. while hxjy i is linear. hxY yi2 hxY xi hyY yiX kxk := hxY xi p (1) (2) The inner product determines a norm which has the property kxk is interpreted as the length of the vector x.1 Orthogonal expansions in a Hilbert space Let r be a complex vector space. h!xY yi = !hxY yi. 1 2 1. A functional variables is called inner product if (1) (2) (3) (4) h ¡ Y ¡ i : r ¢ r 3 C of two These conditions imply the Schwarz inequality hx + yY zi = hxY zi + hyY zi (xY yYz P r).

then x and y are called orthogonal.nition of a Hilbert space that every Cauchy sequence must be convergent. the space is complete. (3) If hxY y i = 0 for the vectors x and y of a Hilbert space. Example 1.1 Let v2 [Y ] be the set of square integrable (complex-valued) functions on the interval [Y ]. This is a Hilbert space with the inner product hfY gi := b a f (x) g(x) dx . When r & r. that is. kx yk2 + kx + yk2 = 2kxk2 + 2kyk2 which is called parallelogram law. then r c := fx P r : x c h for every h P r g. in notation x c y . For any subset r & r the orthogonal r c is a closed subspace.

2 The in. A maximal orthonormal system is called basis. (The cardinality of any two bases is the same.) A family fxi g of vectors is called orthonormal if hxi Y xi i = 1 and hxi Y xj i = 0 if i T= j .Hilbert space methods for quantum mechanics 3 and with the norm kf k := s b cardinality of a basis is called the dimension of the Hilbert space. The a kf (x)k2 dx X Example 1.

1 Let x1 Y x2 Y XX X be a basis in a Hilbert space vector x P r the expansion r. Then for any x= holds.nite dimensional analogue of Cn is the space 2 (N): (N).2 (Projection theorem) Let w be a closed subspace of a Hilbert space r. where x0 P w and y c w.3 In the space v2 [0Y %] the functions 2 (4) fn (x) = % sin nx form a basis. £ Theorem 1. The convergence is in the v -norm. Any function g P v [0Y %] has an expansion g = n n fn .) £ 2 2 r Theorem 1. (It is known from the theory of Fourier series that for a continuous g the expansion is convergent pointwise as well. = fx = (x Y x Y XX X) : xn P CY 2 1 2 n jxn j2 ` +IgX The inner product is hxY xH i := The canonical basis in this spaces is the sequence n (n = 1Y 2Y XX X): n xn xHn X n = (0Y 0Y XX X Y 1Y 0Y XX X) (1 is at the nth place). The mapping : x U3 x0 de. Any vector x P r can be written in a unique way in the form x = x0 + y. n hxn Y xixn Example 1.

ned in the context of the previous theorem is called orthogonal projection onto the subspace w. This mapping is linear: (!x + "y) = !x + "y .

it is called the matrix of the linear transformation e in the basis e Y e Y XX X Yen . If a basis is . When f : r 3 r is another linear transformation.4 Dnes Petz e eek = k e + k e + X XX + nk en X The numbers ij for an n ¢ n matrix. the the matrix of the composition e f is the usual matrix product of the matrix of e and that of f .

then it induces a 1-1 correspondence betweem linear transformations and n ¢ n matrices.xed. The norm of a linear operator e : r 3 u is de.

2 Show that kef k kek kf k.ned as kek := supfkexk : x P rY kxk = 1g Y Exercise 1. Exercise 1.3 Let f be a continuous function onthe interval [Y ]. De.

2 = .2 The adjoint of a linear operator Let r and u be Hilbert spaces. hxY yiu = h £ xY yir = . (x P r Y y P u ) X (5) is self-adjoint if and only if Example 1. the vector eek is determined by its coordinates: 1.ne a linear operator wf : v [Y ] 3 v [Y ] as wf g = fgX (This is the multiplication by the function f . Furthermore. If : r 3 u is a bounded linear operator. k = 1Y 2Y XX X Yn. Let e : r 3 r be a linear mapping and e1 Y e2 Y XX X Yen be a basis in the Hilbert space r. then its adjoint £ : u 3 r isdetermined by the formula P f (r) is called self-adjoint if £ hxY xi is real for every vector x P r. The mapping e is determined by the vectors eek .4 Let : 2 (N) 3 2 (N) be the right-shift de.) Show that kwf k = supfjf (x)j : x P [Y ]g X 1 1 2 2 1 2 2 2 Moreover.

Then £ (x1Y x2Y x3Y XX X) = (x2Y x3Y x4Y XX X) In another way.ned as n = n+1 in the canonical basis. Exercise 1. £ is called left-shift. £ 1 = 0Y £ n +1 = n X £ .4 Show that any orthogonal projection is selfadjoint.

3 Tensor product of Hilbert spaces and operators Let r and u be Hilbert spaces. £ P f (r) is called a unitary if 1 = £ .5 Show that the product of any two unitary operators is a unitary. Their algebrac tensor product cosists of the formal . Since hei Y eej i = hej Y e£ ei iY this is the complex conjugate of the jYi element of the matrix of e£ . the operator An invertible operator eA := is a unitary. (3) (e 1 )£ = (e£ ) 1 if e is invertible.7 For any e = e£ P f (r). 1.6 For any e P f (r). I n=0 en n! £ Exercise 1. (4) kek = ke£ k Example 1. the operator e£ e is self-adjoint. (2) (e£ )£ = e.5 Let e : r 3 r be a linear mapping and e1 Y e2 Y XX X Yen be a basis in the Hilbert space r. £ Example 1. The iY j element of the matrix of e is hei Y eej i.Hilbert space methods for quantum mechanics 5 Theorem 1. (!e)£ = !e£ (! P C). (ef )£ = f £ e£ .3 The properties of the adjoint: (1) (e + f )£ = e£ + f £ . Example 1.

j xi yj (xi P rY yi P u)X Computing with these sums. one should use the following rules: (x1 + x2 ) y = x1 y + x2 yY (!x) y = !(x y ) Y x (y1 + y2) = x y1 + x y2Y x (!y) = !(x y) X The inner product is de.nte sums i.

l hxi Y zk ihyj Y wl iX When r and u are .k.j xi yj Y k.j.ned as h (6) i.l z k wl i = i.

nite dimensional spaces. . otherwise the algebraic tensor product must be completed in order to get a Banach space. then we arrived at the tensor product Hilbert space r u.

6 Dnes Petz e Example 1.8 If f P r := v2 (Y ") and g P u := v2 (Y# ). then be interpreted as a function of two variables: f (x)g(y). f g can £ The tensor product of .

ntely many Hilbert spaces is de.

In the space v2 (R) the Hermite functions dim(r u) = dim(r) ¢ dim(r)X nomial.ned similarly. where rn (x) is the apropriately normalized Hermite poly2 (7) £ Exercise 1.9 In the Hilbert space v2 (R2 ) we can get a basis if the space is considered as v2 (R) v2 (R). Therefore.6 Let e P f (r) and f P f (r) be operators on the . This shows that Example 1. If e1 Y e2 Y XX X and f1 Y f2 Y XX X are bases in r and u. then fei ej : iY j g is a basis in the tensor product space. respectively. the two variable Hermite functions 9nm (xY y) := e (x2 +y2 )=2rn (x)rm (y) (nY m = 0Y 1Y XX X)X for a basis in v2 (R2 ). 9n (x) = exp( x a2)rn (x) form a good basis.

(Hint: The determinant is the product of the eigenvalues. Show that where n = dim r and m = dim u.nite dimensional spaces r and u.) Exercise 1.10 Let fe1 Y e2 Y e3 g be a basis in r and ff1 Y f2 g be a basis in u.7 Show that ke f k = kek ¡ kf k. If [eij ] is the matrix of e P f (r1 ) and [fkl ] is the matrix of f P f (r2 ). then eij fkl ei fk X (e f )(ej fl ) = It is useful to order the tensor product bases lexicographically: e1 f1 Y e1 f2Y e2 f1Y e2 f2Y e3 f1Y e3 f2. Example 1. we can write down the matrix of e f and we have P det(e f ) = (det e)m (det f )n Y i. Fixing this ordering.k e T Te T T Te T T Te T Te R e 11 11 21 21 31 31 f f f f f f 11 21 11 21 11 21 e e e e e e 11 11 21 21 31 31 f f f f f f 12 22 12 22 12 22 e e e e e e 12 12 22 22 32 32 f f f f f f 11 21 11 21 11 21 e e e e e e 12 12 22 22 32 32 f f f f f f 12 22 12 22 12 22 e e e e e e 13 13 23 23 33 33 f f f f f f 11 21 11 21 11 21 e e e e e e 13 13 23 23 33 33 f f f f f f U U U U 12 U U U 22 U U U 12 S 22 12 Q X £ 22 .

in notation k r. It follows that the dimension of k r is The subspace spanned by the vectors v1 v2 X XX vk is called the kth antisymmetric tensor power of r. In particularly. rk has two important subspaces. Its restriction is dented by k e which is equivalently de. The terminology \antisymmetric" comes from the property that an antisymmetric tensor changes its sign if two elements are exchanged. the symmetric and the antisymmetric ones. in notation rk . When e P f (r). then the transformation k e leaves the subspace k r invariant. v1 v2 X XX vk if vi = vj for dierent iand j . So k r & k r. If v1 Y v2 Y XX X Yvk P r are vectors then their antisymmetric tensorproduct is the linear combination v1 v2 X XX vk := p1k! ( 1)() v(1) v(2) X XX v(k) (8) !(v v X XX vk ) = v v X XX v` (!v`) v` X XX vk 1 2 1 2 1 +1 and (v1 v2 X XX v` 1 v v`+1 X XX vk ) + + (v1 v2 X XX v` 1 v H v`+1 X XX vk ) = = v1 v2 X XX v` 1 (v + v H ) v`+1 X XX vk X k e(v1 v2 X XX vk ) = ev1 ev2 X XX evk X If e1 Y e2 Y XX X Yen is a basis in r. then fei(1) ei(2) X XX ei(k) : 1 i(1) ` i(2) ` XX X ` i(k)) ng is a basis in k r. then e(1) e(2) X XX e(k) is a linear transformation on rk and it is denited by ek .Hilbert space methods for quantum mechanics 7 where the summation is over all permutations % of the set f1Y 2Y XX X Ykg and '(%) is the number of inversions in %. The computational rules for the antisymmetric tensors are similar to (6): Let r be a Hilbert space. If e P f (r). The k-fold tensor product r X XX r is called the kth tensor power of r.

we have n k ha k nY Exercise 1. (Hint: Show that k (ef ) = (k e)(k f ). Use this to prove that det(ef ) = det e ¢ det f .8 Show that ! = det e in (11). Consequently.) n e = ! ¢ identity (11) . n r has dimension 1 and for any operator e P f (r).ned as (9) (10) otherwise for k b n the power k r has dimension 0.

Theorem 1. is positive if and only if for any 1 k n the determinant of the k ¢ k matrix (hei Y ej i)k =1 ij is positive.4 Let P f (r) be a self-adjoint operator and e1 Y e2 Y XX X Yen be a basis in the Hilbert space r.10 Show that an orthogonal projection is positive. Conversely. Exercise 1. A positive operator is self-adjoint. : 1 (12) P f (r) is called positive if hxY xi ! 0 for every vector x P r.8 Dnes Petz e The symmetric tensor product of the vectors v1 Y v2 Y XX X Yvk P r is v v X XX vk := p1k! 1 2 where the summation is over all permutations % of the set f1Y 2Y XX X Ykg again. The linear span of the symmetric tensors is the symmetric tensor power k r. in particular the eigenvalues of a positive operator.9 1. in notation ! 0. lies in acting on a .4 Positive operators i(1) i(2) X XX i(k) ngX Give the dimension of k r if dim(r) = n. R+ . It has the basis v v X XX v k Y (1) (2) ( ) fei(1) ei(2) X XX ei(k) Exercise 1. if all the eigenvalues are positive for a self-adjoint operator The spectrum.

nite dimensional space. Positive matrices are also called positive semide. then it is positive.

e f if f e is positive. d hxY f (e + t )xi ! 0 dt holds for every vector x which means that d dt f (e + t ) ! 0X . Example 1.11 Let monotone if 0 f : R 3 R be a smooth function.nite. Let eY f P f (r) be self-adjoint operators. f + is called matrix e f implies that f ( e) f ( f ) X f is matrix monotone if and only for every positive operator e and and for the real parameter t ! 0.

1 i k. !j must be positive as well. the matrix (or the corresponding Hilbert space operator) can be written in the form k i=1 !i i i Y where !1 Y !2 Y XX X Y!k are the dierent eigenvalues and ii is the orthogonal projection onto the subspace spanned by the igenvectors correspondg to the eigenvalue !i . we have for the trace 2!j Tr ij p (t)ij = Tr ij ij X Since both traces are positive. then !i(iip (t) + p (t)ii) = and after multiplication by ij from the left and from the right. we get p H(t)p (t) + p (t)p H(t) = X If p H (t) = i !i ii is the spectral decomposition.11 Show that that the square function is not matrix monotone.5 The spectral theorem 11 11 ! X Use the argument of the previous example for 2 ¢ 2 matrices. The spectral theorem extends this to arbitrary selfadjoint operator e. Let p (t) := e + t X It is enough to see that the eigenvalues of p H (t) are positive.Hilbert space methods for quantum mechanics 9 We want to show that the squareroot function is matrix monotone. Dierentiating the equality p (t)p (t) = e + t .) The eigenvalues of a self-adjoint matrix are real and the eigenvectors corresponding to dierent eigenvalues are orthogonal. £ i p Exercise 1. (Hint: Choose e to be diagonal and = 1. Then the spectrum is not necessary discrete and the . Therefore.

i (Y) = 0. . i (C) = s .nite sum is replaced by an integral. Assume that for each Borel set f & a positive operator i (f ) P f (r) is given such that (1) 0 i (f ) s . Let be a complete separable metric space and r be a Hilbert space.

i=1 then I i (f ) e = or the unit circle T.10 Dnes Petz e (2) If (fi ) is a sequence of pairwise disjoint Borel subset of and f = I fi . We want to integrate a function f : When is a .

In the most important examples is a .nite set. shortly POVM. then In this case i is called a positive operator-valued measure.

nite set. the real line R for every vector e P r. i=1 i (fi)e 3 C with respect an POVM on . f (x) di (x) = xP f (x)i (fxg) is a .

the de.nite sum. In the general case.

f (x) d"e (x) Exercise 1. then i (f ) = 0 and f (e) = f (!) di (!) for every continuos function de. if f & R and the spectrun of e are disjoint. Then there exists a unique projectionvalued measure on the real line such that The next theorem is the spectral theorem for a bounded self-adjoint operator.nition of the integral can be reduced to many integrals with respect to common measures. "e (f ) = heY i (f )ei gives us a positive measure on the Borel sets of f (x) di (x) = P f (r). Show that if a vector e is in the range of i (f1 ). then i (f2)e = 0. if . Given a vector e P r. We say that the integral heY ei = holds for every e P .5 Let e = e£ P f (r). (Therefore. i (f1) and i (f2) are orthogonal. e = ! di (!)X Moreover. that is i (f ) = i (f )2 .12 Let i be a projection-valued measure and let f1 Y f2 be disjoint Borel set.) Theorem 1. A POVM i is called projection-valued measure if i (f ) is a projection operator for every Borel set f .

.ned on the spectrum of e.

Hilbert space methods for quantum mechanics 11 erator e but in this case e and f (e) are not everywhere de.

ned operators. The projection-valued measure in the theorem is called the spectral measure of the operator e. then the spectral measure is on the unit circle. Similar theorem holds for unitary operators. Similar result holds for unbounded self-adjoint op- 2 Postulates of quantum mechanics The .

£ x j 4i + z j 5i X (Indeed. then the corresponding states identical if f1 = zf2 for a complex number z of modulus 1. the state j 4i may have the interpretation that the \polarization is vertical") and j 5i means that the \polarization is horizontal".rst postulate of quantum mechanics tells that to each quantum mechanical system a Hilbert space r is associated. Such z is often called phase. When f1 and f2 are unit vectors. we have the constraint x2 + x2 + x2 = 1 for the parameters (x1 Y x2 Y x3 ) P R3 .) Splitting z into real and imaginary parts as z = x2 + ix3 . the space of all pure states of a qubit is conveniently visualized as the sphere in the three dimensional Euclidean space.1 n-level quantum systems The pure physical state of the system determines a corresponding state vector up to a phase. This means that the space has a basis consisting of eigenvectors of the mixed states. Mixed states are described by density matrices.12 The 2 dimensional Hilbert space C2 is used to describe a 2level quantum system called qubit. Example 1. Then the state vector is 1 can make the coecient of j 4i real and the corresponding state remains the same. This correspondance is not 1-1. 1 2 3 Therefore. The (pure) physical states of the system correspond to unit vectors of the Hilbert space. it is called the Bloch sphere. respectively. To specify a state of a qubit we need to give a real number x1 and a complex number z such that x2 + jz j2 = 1. we 1 1 2 matrix or statistical operator is a positive operator of trace 1 on the Hilbert space. A density Traditional quantum mechanics distinguishes between pure states and . 2. The canonical basis vectors (1Y 0) and (0Y 1) are usually denoted by j 4i and j 5i. (An alternative notation is j1i for (0Y 1) and j0i for (1Y 0). multiplying a unit vector z j 4i + z j 5i by an appropriate phase.) Since the polarization of a photon is an important example of a qubit.

12 Dnes Petz e statistical operator and the sum of eigenvalues is 1. (In the .

nite dimensional case the .

rst condition is automatically ful.

lled.) The pure states represented by unit vectors of the Hilbert space are among the density matrices under an appropriate identi.

£ (A2) The observables of a quantum mechanical system are described by self-adjoint operators acting on the Hilbert space. The physical states of a quantum mechanical system are described by statistical operators acting on the Hilbert space. If x = jxi is a unit vector.13 A state of the spin (of 1a2) can be represented by the 2 ¢ 2 matrix ! 1 1 + x3 x1 ix2 X (13) 2 x1 + ix2 1 x3 This is a density matrix if and only if x2 + x2 + x2 1 2 3 The second axiom is about observables.cation. then jxihxj is a density matrix. (A1) Example 1. Geometrically jxihxj is the orthogonal projection onto the linear subspace generated by x. 1. A self-adjoint operator r 3 r which satis. Note that jxihxj = jy ihy j if the vectors x and y dier in a phase.

es for xY y P r. Self-adjoint operators on a .

) They are called Pauli matrices. e on a Hilbert space r is a linear operator hexY yi = hxY eyi Example 1. Multiplicity of !i is exactly the rank of ii .nite dimensional Hilbert space Cn are n ¢ n self-adjoint matrices. Any 2 ¢ 2 self-adjoint matrix is of the form e(x0 . A self-adjoint matrix admits a spectral decomposition e = i !i ii .14 In case of a quantum spin (of 1a2) the matrices ' 1 = 01 10 ! Y ' 2 = 0 i i 0 ! Y ' 3 = 1 0 0 1 ! are used to describe the spin of direction xY yYz (with respect to a coordinate system. where !i are the dierent eigenvalues of e and ii is the orthogonal projection onto the subspace spanned by the eigenvectors corresponding to the eigenvalue !i . We also use the shorthand notation x0 '0 + x ¡ '. The density matrix (13) can be written as 1 2 ( '0 + x ¡ ' ) Y (14) .x) := x0'0 + x1'1 + x2'2 + x3'3 if '0 stands for the unit matrix s .

The extreme points of the ball correspond to pure state and any mixed state is the convex combination of pure states in in.Hilbert space methods for quantum mechanics 13 where kxk 1. Formula (14) makes an ane correspondence between 2 ¢ 2 density matrices and the unit ball in the Euclidean 3-space.

then the result of the measurement may not be the same. If we prepare two identical systems in the same state. This indeterminism or stochastic feature is fundamental. there is no multiple eigenvalue. In higher dimension the situation is much more complicated. and we measure the same observable on each.nitely many dierent ways. i !i = 1. It is unique if the spectrum of & is non-degenerate.2 Measurements Quantum mechanics is not deterministic. Since & is self-adjoint such a decomposition is deduced from the spectral theorem and the vectors jxi i may be chosen pairwise orthogonal eigenvectors and !i are the corresponding eigenvalues. £ Any density matrix can be written in the form &= by means of unit vectors jxi i and coecients !i ! 0. i !ijxiihxij (15) 2. Under this condition (15) is called Schmidt decomposition. (A3) Let be a . that is.

nite set and for x P an operator x P f (r) be given such that x x£ x = s . then the outcome x P appears with probability Tr x &x£ and after the measurement the state of the system is x &x£ X Tr x &x£ A particular case is the measurement of an observable described by a self adjoint operator e with spectral decomposition i !i ii . One compute easily that the expectation of the random outcome is Tr &e.Tr &s = 1. Such an indexed family of operators is a model of a measurement with values in . if 9(e) ! 0 (16) (17) . then Tr &e ! 0. In this case = f!i g is the set of eigenvalues and i = ii . These properties allow to see quantum states in a dierent way. The functional e U3 Tr &e is linear and has two important properties: 1. If 9 : f (r) 3 C is a linear functional such that e ! 0 and 9(s ) = 1Y then there exists a density matrix &' such that 9(e) = Tr &' eX The functional 9 associates the expectation value to the observables e. If the measurement is performed in a state &. 2. If e ! 0.

the vectors h2Y (e s2 )2i Y where s2 is the identity operator on r2 . (Each subsystem could be a particle or a spin.15 The Hilbert space of a composite system of two spins (of 1a2) is C2 C2 . for example.3 Composite systems According to axiom (A1).) Then we have (A4) Example 1.14 Dnes Petz e 2. a Hilbert space is associated to any quantum mechanical system. When e = e£ is an observable of the . In this space. Assume that a composite system consists of the subsystems (1) and (2). they are described by the Hilbert spaces r1 and r2 .

Measurement of this observable yields the values 1Y 2Y 3Y 4 with probabilities 0Y 1a2Y 1a2 and 0. If ei P f (ri ) (i = 1Y 2). then fej fj : j P tY i P s g is a basis of r1 r2 .rst system. then the action of the tensor product operator e1 e2 is determined by (e1 e2 )(1 2 ) = e1 1 e2 2 since the vectors 1 2 span r1 r2 . then its expectation value in the vector state 2 P r1 r2 . respectively. Therefore in the vector state 0 the spins are anti-correlated. The composite system is described by the tensor product Hilbert space r1 r2 . Therefore. is When fej : j P t g is a basis in r1 and ffi : i P s g is a basis in r2 . The vector state 1 0 = p2 (j 4i j 5i j 5i j 4i) has a surprising property. the dimension of r1 r2 is dim r1 ¢ dim r2 . £ We consider now the composite system r1 r2 in a state 0 P r1 r2 . Consider the observable (18) e := 4 which has eigenvalues 1Y 2Y 3Y 4 and the corresponding eigenvectors are just the basis vectors. e 1 := j 4i j 4iY e 2 := j 4i j 5iY e 3 := j 5i j 4iY e 4 := j 5i j 5i form a basis. The 0 probability occurs when both spins are up or both are down. Let e P f (r1 ) be an observable which is localized at the .

rst subsystem. we have to de. If we want to consider e as an observable of the total system.

i=1 ijeiiheijY . The tensor product operator e s will do. s is the identity operator of r2 .ne an extension to the space r1 r2 .

1 Assume that r1 and r2 are .Hilbert space methods for quantum mechanics 15 Lemma 1.

ikl is called matrix unit. Let fej : j P t g be a basis in r1 and ffi : i P s g be a basis in r2 .j t.j wij ej fiY (ikl s ) wij wtu lujk it = i. Let ikl be an operator on r1 which is determined by the relations ikl ej = lj ek (kYl P s ). taking linear combinations of the previous equations. £ This lemma shows a natural way from state vectors to density matrices.nite dimensional Hilbert spaces.u i wik wil X Since any linear operator e P f (r1 ) is of the form e = kl ikl (kl P C). (They are the quantum analogue of marginal distributions.u wtu eu ft C = i. As a matrix. Then £ is a density matrix and h0Y (e s )0i = Tr e £ X Proof. it is a matrix such that (kYl) entry is 1. Our computation may be summarized as h0Y (ikl s )0i = Tr ikl ( £ ) (kYl P s )X h0Y (e s )0i = Tr e( £ ) X £ is obviously positive and Tr £ = jwij j i.u wij wtuhej Y ikl euihfiY fti = t. we have Then we arrived at the (kYl) entry of £ . Set for the matrix which is determined by the entries wkl .j t. Assume that 0= is the expansion of a unit vector 0 P r1 r2 . Then i. all others are 0.j wij ej fi h0Y (ikl s )0i = = = B i. Given a density matrix & on r1 r2 there are density matrices &i P f (ri ) such that Tr (e s )& = Tr e&1 (e P f (r1 )) (19) and & 1 and &2 are called reduced density matrices.) Tr (s f )& = Tr f&2 (f P f (r2 ))X (20) .j 2 = k0k2 = 1 X Therefore it is a density matrix.

16 Dnes Petz e The proof of Lemma 1.1 contains the reduced density of j0ih0j on the .

For example. it is ( £ )t . then Tr (e s )& = & = (&ij )m i.rst system. Let r1 and r2 be Hilbert spaces and let dim r1 = m and dim r2 = n.j =1 is a density matrix of the composite system. It is well-known that the matrix of a linear operator on r1 r2 has a block-matrix form m = ( ij )m i.j =1 where where =1 ij = eij sn ij = ij fX s f = (ij )m Y i.j =1 = relative to the lexicographically ordered product basis. where t denotes the transpose of the matrix .j eij Tr sn &ij = i. One computes similarly that the reduced density on the second subsystem. and Assume that i. it is £ . Since £ and ( £ )t have the same non-zero eigenvalues.j =1 iij ij Y n¢n e s = (ij )m Y i.j i. the two subsystems are very strongly connected if the total system is in a pure state. where ij are matrices.j eij Tr &ij (21) and this gives that for the .

rst reduced density matrix we have We can compute similarly the second reduced density &2 . Since Tr (s f )& = we obtain (&1 )ij = Tr &ij X i Tr f&ii & 2 = m i=1 &iiX (22) The reduced density matrices might be expressed by the partial taces. Tr2 : f (r1 ) f (r2 ) 3 f (r1 ) and Tr1 : f (r1 ) f (r2 ) 3 f (r2 ) are de.

ned as Tr2 (e f ) = eTr fY We have Tr1 (e f ) = Tref X and (23) (24) & 1 = Tr2 & & 2 = Tr1 & X .

It may happen that the quantum system under study has a classical and a quantum component. In case of more quantum components.Hilbert space methods for quantum mechanics 17 Axiom (A4) tells about a composite quantum system consisting of two quantum components. the formalism is similar. more tensor factors appear. assume that the .

Then the description by tensor product Hilbert space is still possible.rst component is classical. A basis (jei i)i of r1 can be .

xed and the possible density matrices of the joint system are of the form i pijeiiheij &i Y (2) (25) where (pi )i is a probability distribution and &(2) are densities on r2 . Then the i reduced state on the .

then (A5) where the unitary propagator (tY s) is a family of unitary operators such that The . Another postulate of quantum mechanics tells about the time development of a closed quantum system. If the system is not subject to any measurement in the time interval s & R and &t denotes the statistical operator at time t.rst component is the probability density (pi )i (which may be regarded as a diagonal density matrix) and i pi &(2) is the second i reduced density.

When diso crete time development is considered. this is the so-called Schrdinger picture of quantum mechanics. &t = (tY s)&s (tY s)£ (tY s P s ). If the Hamiltonian is time independent. the term gate is used instead of unitary. (ii) (sY t) U3 (sY t) P f (r) is strongly continuous. a single unitary gives the transformation of the vector state in the form 2 U3 2 . then i (sY t) = exp (s t)r X ~ In the approach followed here the density matrices are transformed in time. or in the density matrix formalism & U3 & £ . . where r (t) is the Hamiltonian at time t. When the unitary time development is viewed as a quantum algorithm in connection with quantum computation. So the gates constitute an algorithm are simply unitary operators.rst order approximation of the unitary (sY t) is the Hamiltonian: i (t + ¡tY t) = s ~ r (t)¡tY (i) (tY s) (sY r) = (tY r).

The dynamical change caused by the interaction is implemented by a unitary and (&e &) £ is the new statistical operator and the reduced density & is the new statistical operator of the quantum system ~ we are interested in. The ane change & U3 & is typical for quantum mechanics ~ and called state transformation.18 Dnes Petz e 2. In this way the map & U3 & is de. When the quantum systems is not closed. it is coupled to another system. The environment has a Hilbert space re and statistical operator &e .4 State transformations Assume that r is the Hilbert space of our quantum system which initially has a statistical operator & (acting on r). called environment. Before interaction the total system has density &e &.

The above de. In this way we obtain a trace preserving and positivity preserving linear transformation.ned on ~ density matrices but it can be extended by linearity to all matrices.

If is i.l i k zk ik & l zl il £ i where the operators ei := z p e£ ep = s p (27) due to (26) and Theorem 1.ned state transformation can be described in several other forms. reference to the environment could be omitted completely.k. = ( ij )m =1 and ij P wn .j a unitary.k. they are m ¢ m matrices with n ¢ n matrix entries.l ik (&e &)kl ( £ )li ik (zk zl &)( il )£ ei&e£ i k k ik satisfy i. Assume that & is an n ¢ n matrix and &e is of the form (zk zl )kl where (z1Y z2Y XX X Yzm ) is a unit vector in the m dimensional space re .6 Any state transformation & U3 i (&) can be written in the form 2 k j k j = 1. In particular.) All operators acting on re r are written in a block matrix form. z . (&e is pure state. then £ is the identity and this implies that i £ ik il = kl sn (26) Formula (22) for the reduced density matrix gives &= ~ = = = = i ( (&e &) £ )ii i.

The . Conversely.Hilbert space methods for quantum mechanics 19 i (&) = p ep&e£ Y p where the operator coecients satisfy (27). all linear mappings of this form are state transformations.

we need to solve the equations ei := The coecients ep in the operator-sum representation are called the operation elements of the state transformation. To prove the converse part. The terms quantum (state) Choose simply z1 = 1 and z2 = z3 = X XX = zm = 0 and the equations reduce to p1 = ep .rst part of the theorem was obtained above. This means that the .

Thanks to the condition (27) this is possible.rst column is given from the block matrix and we need to determine the other columns such a way that should be a unitary. Condition (27) tells us that the .

) i is called completely positive if i idn is positivity preserving for the identical mapping idn : wn (C) 3 wn (C) on any matrix algebra. This result was .rst column of our block matrix determines an isometry which extends to a unitary. Theorem 1. (It is not known what the extreme points of this set are.7 Let i : wn (C) 3 wk (C) be a linear mapping. The state transformations form a convex subset of the set of all positive trace preserving linear transformations. Then completely positive if and only if it admits a representation i is i ( e) = u ueu£ (28) by means of some linear operators u : Cn 3 Ck . £ k zk ik (i = 1Y 2Y XX X Ym)X operation and channeling transformation are also often used instead of state transformation.

(29) (Here iij denote the matrix units. Note that this representation is not unique.j k .) This is the block-matrix representa- . where ij = i (iij ) tion of i .rst proven by Kraus. Let i : wn (C) 3 wk (C) be a linear mapping. It follows that stochastic mappings are completely positive and the operator-sum representation is also called Kraus representation. i is determined by the block-matrix (ij )1 i.

Example 1. it induces the coecient of k as a linear functional on wn (C).20 Dnes Petz e Theorem 1.8 Let i : wn (C) 3 wk (C) be a linear mapping. We want to show the positivity of the representing block-matrix: ij iij k k Tr (pk iij ) = k ij iij k ij iij Tr (pk iij ) s Y where denotes the Hadamard (or entry-wise product) of nm ¢ nm matrices. £ Example 1.17 Consider a positive trace-preserving transformation i : wn (C) 3 wm (C) such that its range consists of commuting operators. The . so the transpose mapping is not completely positive. Recall that according to Schur's theorem the Hadamard product of positive matrices is positive.j k P wk (C) wn (C) is positive. We show that i is automatically a state transformation.16 Consider the transpose mapping e U3 et on 2 ¢ 2 matrices: x y ! U3 x z ! X z w y w 1 T0 =R 0 0 P The representing block-matrix is 0 0 1 0 0 1 0 0 0Q 0U X 0S 1 This is not positive. one can see that i has the form i (e) = k Tr pk eY (30) k where pk is a positive operator in wn (C). Then i is completely positive if and only if the representing block-matrix (ij )1 i. Since a commutative subalgebra of wm (C) is the linear span of some pairwise orthogonal projections k .

Such a family of pk 's describe a measurement which associates the r-tuple (Tr &p1 Y Tr &p2 Y XX X Y Tr &pr ) to the density matrix &. both are positive. Therefore a measurement can be formulated as a state transformation with diagonal outputs. £ .rst factor is [k Y k Y XX X Yk ]£ [k Y k Y XX X Yk ] and the second factor is pk s . r Consider the particular case of (30) where each k is of rank one and k=1 pk = s .

Hilbert space methods for quantum mechanics 21 The Kraus representation and the block-matrix representation are convenient ways to describe a state transformation in any .

Any trace preserving mapping i : w2 (C) 3 w2 (C) has a matrix = t with respect to this basis.18 (Pauli channels) t = 0 and 3 = Diag (YY ). In the 2 ¢ 2 case we have the possibility to expand the mappings in the basis '0Y '1Y '2Y '3.nite dimension. It is not dicult to compute the representing block-matrix. Density matrices are sent to density matrices if and only if for the real parameters YY . we have P 1+ +. where P w and i (w ' + w ¡ ') = w ' + (t + w) ¡ 'X 3 3 3 0 0 0 0 3 1 0 ! (31) -representation: The following examples of state transformations are given in term of the Example 1.

Q 0 0 2 2 .

1 T 0 0 U 2 2 T = R 0 .

1 0 U X (32) S 2 2 +.

1+ 0 0 2 2 is unitarily equivalent to the matrix P 1+ +.

0 0 Q 2 2 T +.

1+ 0 0 U 2 T 2 .

U X 1 R 0 S 0 2 2 .

1 0 0 2 2 This matrix is obviously positive if and only if j1 ¦ j ! j ¦ jX (33) This positivity condition holds when = = = p b 0. £ 1 YY 1 3 Some applications In the traditional approach to quantum mechanics. a physical system is described in a Hilbert space: Observables correspond to self-adjoint operators and statistical operators are associated with the states. Von Neumann associated an entropy quantity to a statistical operator in 1927 [15] and the discussion was extended in his book [16]. Hence the next example gives a channeling transformation. .

we can easily move to an arbitrary density matrix & = i !i j9i ih9i j and we have (&) = !i (j9iih9ij) !i log !iX (35) i i p (& ) + (1 p) (& ) = (&) + p log p + (1 p) log(1 p) Y (34) where is a certain thermodynamical entropy quantity. Then 1 2 temperature and molecule density.1 Von Neumann entropy Von Neumann's argument was a gedanken experiment on the ground of phenomenological thermodynamics which is not repeated here. relative to the . only his conclusion. (Remember that the ortogonality of states has a particular meaning in quantum mechanics. Assume that the density & is the mixture of orthogonal densities &1 and &2.22 Dnes Petz e 3.) From the two-component mixture. & = p&1 + (1 p)&2.

see the original paper [15] or [23]. (Von Neumann's argument was somewhat dierent.) If the entropy of pure states is de. If we expect the entropy (&) to be independent of the Schatten decomposition. then the corresponding eigenvectors can be chosen in many ways. that is. any vector of the Hilbert space can be a state vector. then we are led to the conclusion that (j9ih9j) must be independent of the state vector j9i. The so-called Schatten decomposition i !i j9i ih9i j of a statistical operator is not unique although h9i Y 9j i = 0 is assumed for i T= j .xed This formula reduces the determination of the (thermodynamical) entropy of a mixed state to that of pure states. When !i is an eigenvalue with multiplicity. This argument assumes that there are no super-selection sectors.

A detailed axiomatic characterization of the von Neumann entropy is Theorem 2.ned to be 0 as a kind of normalization. then (34) seems to be natural. The following inequalities hold: holds for an orthogonal mixture and Shannon's classical information measure is involved.1 in [19]. . then we have the von Neumann entropy formula: (&) = if !i are the eigenvalues of & and (t) = t log t.9 Let &1 and &2 be density matrices and 0 ` p ` 1. The mixing property (37) essentially determines the von Neumann entropy and tells us that the relation of orthogonal quantum states is classical. For the sake of simplicity the multiplicative constant will mostly be omitted. i !i log !i = Tr (&) (36) (p& 1 + (1 p)&2 ) = p (&1 ) + (1 p) (&2 ) + r (pY 1 p) Y (37) Theorem 1. It is worthwhile to note that if (&) is interpreted as the uncertainty carried by the statistical operator &.

The .Hilbert space methods for quantum mechanics 23 p (& ) + (1 p) (& ) (p& + (1 p)& ) p (& ) + (1 p) (& ) + r (pY 1 p)X 1 2 1 2 1 2 Proof.

rst inequality is an immediate consequence of the concavity of the function (t) = t log t. In order to obtain the second inequality we bene.

t from the formula Tr e log(e + f ) log e = and infer and Tr (1 p)&2 log(p&1 + (1 p)&2 ) ! Tr (1 p)&2 log(1 p)&2 X Adding the latter two inequalities we obtain the second inequality of the theorem. Note that on an in. The proof is found in [7] or [19].10 Let &1 and &2 be densities on a d-dimensional Hilbert space. If k&1 &2 k1 ` 1 . ¡ I 0 Tr e(e + t) 1 f (e + f + t) 1 dt ! 0 (eY f ! 0) Tr p&1 log(p &1 + (1 p)&2 ) ! Tr p&1 log p&1 Theorem 1. £ The von Neumann entropy is the trace of a continuous function of the density matrix. (k k1 := Tr ( £ )1=2 ). However. a more precise estimate for the continuity will be required in approximations. Such an estimate is due to Fannes. hence it is an obviously continuous functional on the states. then the inequality 3 j (&1 ) (&2 )j k&1 &2 k1 log d + (k&1 &2 k1 ) holds.

j9i equals to j2i up to a phase.nite dimensional Hilbert space the von Neumann entropy is not continuous (but it is such restricted to a set f& : (&) g). This quantity is phase invariant. . see [19]. It equals to 1 if and only if the two states coincide that is. it lies between 0 and 1. 3.2 Fidelity How close are two quantum states? There are many possible answers to this question. Quantum mechanics has used the concept of transition probability jh9 j 9ij2 for a long time. j9i and j2 i. Restricting ourselves to pure states. we have to consider two unit vectors. Most properties of the von Neumann entropy will be deduced from the behavior of the relative entropy.

24 Dnes Petz e Under a quantum operation pure states could be transformed into mixed states. hence we need extension of the .

delity: We call the square root of the transition probability .

p (j9ih9jY &) = h9 j & j 9i Y q 1 2 1 2 2 1 1 2 1 2 p (38) or in full generality dierent context [25] and he proved a variational formula: p (& Y & ) = Tr & = & & = (39) for positive matrices & and & . Shannon used a nonnegative distortion measure.11 1 From Theorem 1.11 the symmetry of p (&1 Y &2 ) is obvious and we can easily deduce the monotonicity of the . This quantity was studied by Uhlmann in a 1 Theorem 1. and we may regard 1 p (j9iY j2 i) as a distortion function on quantum states.delity: p (j9iY j2 i) := jh9 j 2ij.

12 For a state transformation i the inequality p (i (& )Y i (& )) ! p (& Y & ) 1 2 holds. 4 b 0 is arbitrary and q is chosen to be appropriate. Tr &1 i £ (q) Tr &2 i £ (q 1 ) ! Tr &1 i £ (q) Tr &2 i £ (q) 1 ! p (&1 Y &2 )2 X p (i (& )Y i (& )) ! Tr i (& )q Tr i (& )q 4 ! Tr & i £ (q) Tr & i £ (q ) 4 Y 2 2 1 2 1 2 1 In this way the monotonicity is concluded: 1 2 Theorem 1. hence i £ (q) 1 ! i £ (q 1 ). Another remarkable operational formula is p (&1Y &2) = max fjh21j22ij : i (j21ih21j) = &1Y (40) i (j22 ih22 j) = &2 for some state transformation igX This variational expression reduces the understanding of the .delity under state transformation: p (& Y & ) = inf 2 1 np Tr (&1 q) Tr (&2 q 1 ) : 0 q is invertible 1 o where i £ is the adjoint of i with respect to the Hilbert-Schmidt inner product. It is well-known that i £ is unital and positive.

delity of arbitrary states to the case of pure states. The monotonicity property is implied by this formula easily. Convergence in .

This property of the .delity is equivalent with convergence in trace norm: p (&n Y &Hn ) 3 1 if and only if Tr j&n &Hn j 3 0.

delity is a consequence of the inequalities p 1 Tr j&1 &2 j 1 p (&1 Y &2 ) X (41) 1 p (&1 Y &2 ) 2 .

VEB Deutscher Verlag Wiss. Robinson. American Institute of Physics. J. Hilbert space operators in quantum physics. Fannes. L. New York. 34(2004). A simple operational interpretation of . Alicki and M. 1981. Operator Algebras and Quantum Statistical Mechanics II. 5. Berlin. 2. Matrix Analysis. R. O. 1996. Dodd and M. Doubly stochastic maps and unitary mixing.W. A. New York-Heidelberg-Berlin. 1981 6. Nielsen. P. Havlicek. Springer-Verlag. Springer-Verlag. J.Hilbert space methods for quantum mechanics 25 References 1. Alberti and A. l55{L57.. Stochasticity and partial order. Bratteli and D. Exner and M. Uhlmann. 4. Continuity of the quantum conditional information. J. Phys A: Math. Gen. Bhatia. R. P. Blank. 1994. 3.M.

Watanabe. 291{294. The proper formula for relative entropy and its asymptotics in quantum probability. 143(1991). Separable states are more disordered globally than locally. Fannes. von Neumann.. 11. Ohya and D. Petz. A. Springer. 1993 20. Ohya. Math. Inequalities: Theory of majorization and its applications. 12. 1976. Phys. Petz. A. J. 99{114. 258(1982). Helstrom Quantum detection and estimation theory. New York. On capacities of quantum channels. 40(1975). 179{196.S. Phys. Statistical structure of quantum theory. 17. Amsterdam. Rev. 1979. Lett. Math. Springer. Petz. 17(1997). Commun. M. NorthHolland. New York. Phys. F. Math. 229{241 9. A. Probabilistic and statistical aspects of quantum theory. Quantum Entropy and Its Use. 1932. Jensen's inequality for operator and Lwner's o theorem. Kempe. Nielsen and J. D. A continuity property of the entropy density for spin lattice systems. Gtt. 31(1973). 147{151. Springer. Academic Press. Commun. Completely positive maps and entropy inequalities. 2001. Quasi-entropies for . von Neumann. 10. Prob. 19. Petz. 507{513. 86. J. 273-291. Berlin. Math.W. A. Comp. Pedersen. 5(2005). 15. 18. M. Stat.. Olkin. Mathematische Grundlagen der Quantenmechanik. 13. arXiv e-print quant-ph/0111053 7. Academic Press. 1982. Hiai and D. Anal. Marshall and I. 14. Hansen and G. C.W. 16. M. 21. 1(1927).delity. Math. Lindblad. G. F. Phys. Quantum Inf. Thermodynamik quantummechanischer Gesamheiten.K. M. N. M.S. Petz. D. 5184-7 (2001). Holevo. Holevo. 8. Nielsen and D. o Nach. A simple proof of the strong subadditivity. Commun. A.

Rdei and M. Mathematical Phys. Cambridge Philos. D. Uhlmann. Soc. Kluwer. Petz. Schrodinger. 22. Entropy. Phys. Math. 31(1936). . Algebra of the canonical commutation relation. o 24.nite quantum systems. Proc. Rep. in John von Neumann and the Foundations of Quantum Physics. D. 25. E. 446{452. 1990. von Neumann and the von Neumann entropy. 21(1986). 9(1976). The "transition probability" in the state space of a *-algebra. Petz. Leuven University Press. Rep. 23. 57{65. Probability relations between separated systems. 273{279. 2001. e Stltzner. M. eds. A.

Rev. 197{234. Modern Phys. Commun. Relative entropy and the Wigner-Yanase-Dyson-Lieb concavity in an interpolation theory. The role of relative entropy in quantum information theory. 27. 74(2002). Phys. V. . Math. Uhlmann.26 Dnes Petz e 26. Vedral. 21{32. 54(1977). A.

- Note02(18.01.2009).pdf
- Note02(18.01.2009).docx
- Note01(17.01.2009).pdf
- Note01(17.01.2009).docx
- Assignment(Double,Surface & Triple Integrals).pdf
- Assignment(Double,Surface & Triple Integrals)
- Assignment(Double,Surface & Triple Integrals).pdf
- Lecture03.pdf
- Lecture 01
- Assignment(Line Integrals)
- Assignment(Double,Surface & Triple Integrals).pdf
- Lecture 01
- Assignment(Line Integrals)
- Assignment(Double,Surface & Triple Integrals)
- Lecture 01
- Assignment(Line Integrals)
- Assignment I - Vibrations of Physical Systems
- Note03(05.08.2009)

Sign up to vote on this title

UsefulNot useful- Quantum Mechanics Lecture
- 4604 Exam1 Solutions Fa05
- Basic Quantum Mechanics 2
- Quantum Mechanics PyEd 342
- Quantum Mechanics
- gauge transformations
- Advanced Quantum Mechanics
- path-integral
- 3063 Exam2 Solutions Sp06
- The Dirac formalism of quantum mechanics
- Gauge Theory
- Quantum mechanics
- Commutators, eigenvalues and quantum mechanics on surfaces...
- Barut a.O. - Formulation of Quantum Dynamics in Terms of Generalized Symmetries
- Sakurai Solutions 5-1 5-2
- Quantum Mechanics for Engineers
- Sakurai quantum mechanics solutions 4
- Quantum Mechanics
- Notes on Quantum Mechanics
- Griffith's Quantum Mechanics Problem 2.51
- 4604 Solutions Set2 Fa07
- David Bohm
- 51891292 Quantum Mechanics I
- Super Symmetry and Quantum Mechanics
- Advanced Quantum Mechanics
- Quantum-Mechanics
- Quantum Mechanics - Concepts and Applications
- Relativistic Quantum Mechanics
- Relativistic Quantum Mechanics
- Merzbacher 3rd Ed
- Hilbert Space Methods for Quantum Mechanics

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd