Arxiv:Quant-Ph/0504091V1 12 Apr 2005

Home , Quantum operation

arXiv:quant-ph/0504091v1 12 Apr 2005 h pc flna prtr ntesaesae utas just space, state the on for operators basis orthonormal linear the of provid- i.e., by space [16], form the basis matrix operator super- in an the represented ing Alternatively, be can up 15]. only operators [2, but unique equivalence is distinct, unitary to projectors of orthogonal, sum a convex necessarily of to into not decomposition up operator opera- the only density like Kraus unique given just of is [2], description set equivalence decomposed this a unitary be However, by can form map [1]. itself Kraus CP tors operators to any so-called density that space the known of Hilbert into is It set extended 2]. the trivially [1, the map be on should to positive acting it completely has is i.e., operation that quantum (CP); operators superoperator as physical a linear Any it by of described consider 14]. space to 13, the is [3, on which of acting one superoperators operation, quantum a 12]. 11, 10, as process 9, 8, “quantum known 7, or tomography”[6, data identification” experimental channel of engineered “quantum from or estimation natural processes statistical in quantum the occurring on operations quantum efforts re- quantum have extensive There a reliable interest. been is recent ef- considerable of and of on-going subject communications, capable search quantum of devices and component develop computing essential to an 5]. forts quantum [4, is of system analysis operations quantum and me- isolated characterization quantum the initially Experimental describes by an It for allowed chanics transformation [3]. us- general noise, etc. most and subroutines, decoherence probabilistic computation, ing the It the in measurements of 2]. issues: [1, middle central evolution several with non-unitary deal as sys- can uni- well quantum including as of evolution computation, tary quantum dynamics in the occurring describing tems for tool erful hr r eea aso nrdcn h oinof notion the introducing of ways several are There pow- a us offers operations quantum of formalism The pc r osdrdfragnrloeainatn nasing a on acting operation fo general basis a for orthonormal Th considered the are applications. is, space their that and basis, operation operator quantum arbitrary a of — matrices ASnmes 36.a 03.65.Wj 03.67.-a, numbers: are PACS technology applicati and Several science trivial. information quantum is in case formulas qudits more assoc establish to are the Extention matrices of two products the between dyadic formulas and conversion basis prod operator tensor relevant different having th the basis on operator operator coeffic induced positive expansion of bijective, types the associated by the given as are well matrices as two the that show We hspprcniestofeunl sdmti representa matrix used frequently two considers paper This .INTRODUCTION I. nteMti ersnaino unu Operations Quantum of Representation Matrix the On E,3 iuiak,Tuua brk 0-51 Japan 305-8501, Ibaraki Tsukuba, Miyukigaoka, 34 NEC, udmna n niomna eerhLaboratories, Research Environmental and Fundamental ohhr ab n au Nakamura Kazuo and Nambu Yoshihiro Dtd eray1 2008) 1, February (Dated: eea uhr 2 ,1,1,1,1,2] e scon- us Let 20]. 19, 17, 12, by 10, matrix a dynamical 7, sider or [2, process authors the several called also is which il pc ( space ville h ieaue h rtoei htw althe call we what is in one first found The matrix frequently literature. different the superoperators two considers for paper representations This resentation. operator the once fixed. be matrix been also has the the can basis choosing using operation in quantum described freedom a uniquely is Similarly, a basis have basis. state still state the we once although state the unique fixed, quantum to a the provides respect of matrix with density description repre- The defined be Hilbert basis. matrix state can the density chosen operator a for density by basis the sented state example, in For a represented providing space. be by can space form Hilbert matrix the on operators the s 1,2,2] nti omls,telna operators linear the formalism, formal- this Liouville In the 23]. in of 22, terms [13, in ism given is operation tum [21]. operation quantum physical oma form etdb h iayfr ftesuperoperator the of form binary the by sented h pc flna prtr cigon acting operators linear of space the product { pc ( space nodnr prtratn on acting operator symbol ordinary substitution an the (1), double-hat Eq. In ⊙ one. another into cigon acting E ˆ n a osrc aydffrn ye fmti rep- matrix of types different many construct can One lentvl,aohrmti ersnaino quan- a of representation matrix another Alternatively, α hudb elcdb rnfre prtr and operator, transformed a by replaced be should HS } obe-pc endwt epc otwo to respect with defined doubled-space e h pc flna prtr ntestate the on operators linear of space the r α d da optbemti multiplication. matrix computable a as ed n fteemtie n h conversion the and matrices these of ons c tutrs .. rnce rdcsof products Kronecker i.e., structures, uct eo two or le 2 d =0 HS − d ae iatt tt ai.Teexplicit The basis. state bipartite iated d r dnie ihtespretr naLiou- a in supervectors the with identified are 1 2 h presented. dmninlHletsae( space Hilbert -dimensional A, arcsdfie ihrsett an to respect with defined matrices e in htw althe call we what — tions × ˆ -space) in HS i sdt itnus h ueoeao from superoperator the distinguish to used îs ˆ et fteLovlesuperoperator Liouville the of ients L B d ˆ HS -space) 2 d ≡ i d S hc asalna prtrin operator linear a maps which , oiiematrix positive ˆ ˆ lvlqatmsse (qudit). system quantum -level ( d ⊙ h ieroperation linear the , HS Tr = ) A L d ˆ fw hoetefie ai set basis fixed the choose we If . d † α,β d 2 B X ˆ 2 nrdcn obebra-ket double a Introducing . − hti,teHilbert-Schmidt the is, that , =0 1 χ αβ χ H E ˆ d ≡ α χ h coefficients The . and - ⊙ [ χ H αβ E ˆ H -space) β S † ] α,β d d S 2 a erepre- be can - − ihascalar a with =0 1 if , χ H -matrix, d S and , HS χ sa is (1) αβ d 2

⊗2 ˆ notation for the elements of d2 [23], we associate every on , and ˆ ( ) = is the identity superoperator L d operator Aˆ with an L-ket Aˆ and its Hermitian con- actingH on the IHS⊙-space⊙ of the second system. This re- † | ii jugate operator Aˆ with an L-bra Aˆ . The space 2 ˆ ˆ hh | Ld sult implies that there is a bijection between and χˆ, is furnished with an inner product Aˆ Bˆ = TrAˆ†Bˆ, from which we can deduce the conversion formulaS be- and constitutes a d2-dimensional Hilberthh | space.ii Then by 2 tween χ- and -matrices as a computable matrix alge- − S chosing an arbitrary fixed set of operator basis Eˆ d 1 bra. Although Nielsen and Chuang have considered such { α}α=0 in d, any linear operation S can be written as the a formula [2, 17], their method requires finding matrix superoperatorHS inverses to convert from the -matrix to the χ-matrix. S 2 Here, we show a conversion formula without matrix in- d −1 version. We then extend the formula to two-qudit op- ˆ ˆ ˆ = X αβ Eα Eβ (2) S S | iihh | erations. We also briefly review the requirement for the α,β=0 χ-matrix to represent physical operations. In Sec. IV, we illustrate the applications of the present formulation. acting on d2 , which maps a L-space supervector into L 2 2 First, we discuss how χ- and -matrices can be obtained another one. The coefficients αβ form a d d complex 2 S d −1 S × experimentally. We describe a typical procedure to ob- matrix [ αβ ]α,β=0. They represent the amplitudes tain the χ- and -matrices defined with respect to an S ≡ S S of the operator components Eˆα contained in the state arbitrary operator basis set. Next, we discuss how these after applying the quantum operation on the operator matrices and the present conversion formulas are useful component Eˆβ. We call it the -matrix by analogy with for the analysis and design of the quantum operations, the S-matrix appearing in time-independentS scattering quantum circuits, as well as quantum algorithms. In Sec. theory [24, 25]. The -matrix has been actually used V, we summarize the results. to describe quantum operationsS by several researchers on quantum information science [8, 9, 26]. The χ- and -matrices offer us the most general description of the II. OPERATOR BASIS dynamicsS of initially isolated quantum systems allowed in quantum mechanics, just as the density matrix offers us It has been noted by many researchers that the su- the most general description of the quantum mechanical pervectors defined in the L-space d2 can be identi- state. ⊗ fied with the vectors in the doubledL Hilbert space 2 The choice of matrix representation type is a matter of d [27, 28, 29, 30]. Let us start by reviewing thisH fact, convenience, depending on the application. We will dis- briefly. Consider an arbitrary chosen set of an orthonor- cuss later how the χ- and -matrices are useful for the d−1 S mal basis i for d (denoted as standard state ba- analysis and design of quantum operations. Although {| i}i=1 H sis). Any linear operator Aˆ in can be expanded these matrices indeed have their own useful applications, HSd as Aˆ = d−1 A i j where the dyadic products i j their mutual relation is non-trivial and has not been clar- Pi,j=0 ij | ih | | ih | ified. The main purpose of this paper is to clarify the form a basis for d. The L-space supervectors corresponding to the dyadicHS operator i j is denoted by the underlying relation between two different matrix repre- | ih | sentations of quantum operations, and to provide the way double ket ij , with which the L-space supervector as- | ii d−1 for building bridges across the different classes of appli- sociated with Aˆ is written as Aˆ = Aij ij . The | ii Pi,j=0 | ii cations. We here consider a quantum operation acting scalar product of two L-space supervectors Aˆ and Bˆ on the state of a single d-level quantum system (abbre- is defined as | ii | ii viated as single-qudit operation) or a two d-level quantum systems (two-qudit operation). We start in Sec. II Aˆ Bˆ = TrAˆ†B,ˆ (3) by recalling the notion of operator basis and its proper- hh | ii ties, which is helpful for the subsequent discussions. We which introduces a metric of an L-space. The vectors ij note the equivalence between the supervectors in the L- form a basis for a d2-dimensional Hilbert space. Thus,| weii space d2 and the vectors in the doubled Hilbert space can safely identify ij with the product state i j . ⊗2 L | ii | i ⊗ | i d = d d. This equivalence implies that for any Then, the L-space vector associated with Aˆ can be identi- H H ⊗ H 2 ˆ d −1 ˆ ˆ ˆ ˆ operator-basis set Eα α=0 for d, there is an isomor- fied with vector A (A I) I in the doubled space { } 2 HS | ii ≡ ⊗ | ii d−1 phic state-basis set Eˆ d −1 for ⊗2. We review sev- ⊗2[21, 31], where d−1/2 Iˆ d−1/2 i i is {| αii}α=0 Hd Hd | ii ≡ Pi=0 | i ⊗ | i eral properties of the operator basis for later discussion. the isotropic state in ⊗2 [32, 33], and Iˆ is the identity Hd In Sec. III, we first consider the single-qudit operations. operator in d. It may be helpful to recall the mathe- We show that χ- and -matrices are given by the ex- matical representationsHS of Aˆ in the space Cd×d of d d 2 S ˆ d × pansion coefficients of the L-space superoperator ˆ and complex matrices and the vector Aˆ in C . Consider a S | ii d ˆ ˆ ˆ ⊗2 representation where i is a column vector in C with a the associated operator χˆ (dρÎ ) acting on d | i ≡ S ⊗ I H unit element in the j th row and zeros elsewhere. Then, Aˆ defined with respect to two types of induced operator d−1 ⊗2 is identified with the d d complex matrix A [Aij ] basis on d having different tensor product structures. × ≡ i,j=0 Here,ρ ˆ His the density operator of the isotropic state in Cd×d, and Aˆ is obtained by placing the entries of a I | ii 3 d d matrix into a column vector of size d2 row-by-row, This is a trivial consequence of Eq. (3). i.e.,× 2 ˆ d −1 Lemma 2 A set of the operators Eα α=0 in d is T {2 } HS Aˆ [A11, , A1d, A21, , A2d, , Ad1, , Add] . ˆ d −1 ⊗2 | ii ≡ · · · · · · · · · · · · complete iff a set of states Eα α=0 in d is com- (4) plete. {| ii} H Therefore, Aˆ contains the same elements as Aˆ but in | ii different positions. This and Eq. (3) indicate that Aˆ and To prove this, the following Theorem is helpful [16]: Aˆ are isometrically isomorphic. Accordingly, we may | ii Theorem 1 (D’Ariano, Presti, and Sacchi (2000)) A identify the L-space with the doubled Hilbert space, i.e., 2 ⊗2 ⊗2 ˆ d −1 2 = H . Hereafter, we use a common symbol to set of the operators Eα in d is complete iff it d d d { }α=0 HS Ldenote both these spaces without loss of clarity. H satisfies one of the following equivalent statements: It will be useful for the later discussion to note the 1. For any linear operator Aˆ on , we have following relations hold: Hd 2 ˆ ˆ ˆ ˆ ˆ ˆT d −1 A B C = ACB , (5) ˆ ˆ† ˆ ˆ ⊗ | ii | ii A = X (TrEαA)Eα. (10) α=0

ˆ ˆ ˆ ˆ† (1) ˆ T r2[ A 1212 B ] = (AB ) , (6) 2. Let ˆ ( ) be the superoperator on the space of | ii hh | Edepol · · · linear operators in d describing completely de- polarizing operation.HS For any linear operator Aˆ on ˆ ˆ ˆT ˆ∗ (2) d, we have T r1[ A 1212 B ] = (A B ) , (7) H | ii hh | 2 d −1 ⊗2 ˆ 1 1 † where the indices refer to the factors in d in which the ˆ (Aˆ)= (TrAˆ)Iˆ = Eˆ AÊˆ . (11) H Edepol d d X α α corresponding operators have a nontrivial action, and the α=0 transposition and conjugation are referred to the chosen standard state basis [16]. 3. For chosen any state basis i d−1 for , we have Now, let us consider the operator basis, that is, the {| i}i=1 Hd 2 complete basis for d. Consider an arbitrary set of d - ⊗2 HS 2 vectors in d . From the above isomorphism, this set d −1 H 2 2 ˆ d −1 ˆ d −1 n Eˆ† m l Eˆ k = δ δ . (12) can be written as Eα α=0 , where Eα α=0 is the X α α nk ml 2 h | | ih | | i {| ii} ˆ {d −1} α=0 associated set in d. The set Eα α=0 is the state ⊗2 HS {| ii} basis for d iff it is orthonormal, i.e., 2 H Now, let us prove Lemma 2. If Eˆ d −1 is complete, {| αii}α=0 Eˆα Eˆβ = δαβ, (8) Eq. (9) must be satisfied. Then, for any Aˆ in d, we hh | ii have HS and it is complete, i.e., 2 d −1 2 ˆ ˆ ˆ ˆ ˆ ˆ ˆ d −1 (A I) I = A = X Eα Eα A ˆ ˆ ˆ ˆ ⊗ | ii | ii | iihh | ii X Eα Eα = I I. (9) α=0 | iihh | ⊗ 2 α=0 d −1 ˆ† ˆ ˆ ˆ ˆ = X (TrEαA)Eα I I . ˆ ⊗ | ii The previous discussion shows that the vector A in α=0 ⊗2 | ii d is identified with L-space supervectors associated H ˆ with the operator Aˆ in d. This implies the following Since this holds for any A, Eq. (10) must be satisfied. 2 2 proposition. HS ˆ d −1 ˆ d −1 Hence, Eα α=0 is complete. Conversely, if Eα α=0 is { } ⊗2 { } 2 complete, then for any Aˆ in , we have ˆ d −1 | ii Hd Proposition 1 A set of the operators Eα α=0 is a ba- { 2 } 2 ˆ d −1 d −1 sis set for d iff a set of states Eα α=0 is the basis ⊗HS2 {| ii} ˆ ˆ ˆ ˆ ˆ† ˆ ˆ ˆ ˆ set for . A = (A I) I = X (TrEαA)Eα I I d | ii ⊗ | ii ⊗ | ii H α=0 2 We argue below that this is true. First, we note the d −1 following lemmmas. ˆ ˆ ˆ = X Eα Eα A . | iihh | ii 2 α=0 d −1 Lemma 1 A set of the operators Eˆα in d is α2=0 { }d −1 HS⊗2 ˆ orthonormal iff a set of states Eˆ in is Since this holds for any A , Eq. (9) must be satisfied. α α=0 d 2 | ii orthonormal. {| ii} H Hence, Eˆ d −1 is complete. {| αii}α=0 4

⊗2 From these two lemmas, we obtain Proposition 1. This d are unitarily related, i.e., ⊗2 H indicates that the state basis Eˆα for has bijective d 2 2 | ii H d −1 d −1 correspondence to the operator basis Eˆα for d. The HS Fˆ = Eˆ Eˆ Fˆ = Eˆ (15) following corollary is a consequence of Theorem 1, which | βii X | αiihh α| βii X | αiiU αβ will be useful for the later discussions. α=0 α=0

2 ˆ ˆ ˆ d −1 iﬀ the operator bases Eα and Fα in d are unitarily Corollary 1 Let Eα α=0 be an arbitrary chosen set of HS { } ⊗2 related, i.e., an operator basis for d. The isotropic state in d is HS H 2 written as d −1 ˆ ˆ 2 Fβ = X Eα αβ. (16) d −1 U 1 ˆ ˆ 1 ˆ ˆ∗ α=0 ρˆI = I I = X Eα Eα, (13) d| iihh | d ⊗ α=0 ˆ ˆ ˆ† ˆ In Eqs. (15) and (16), αβ = Eα Fβ = TrEαFβ is a 2 2 U hh | ii ⊗2 αβ-entry of the d d unitary matrix . If we consider ˆ ⊗ and the swap operator V on d is written as a unitary superoperator× acting on the vectorsU in 2, H Hd 2 2 d −1 d −1 Vˆ = Eˆ Eˆ† . (14) ˆ ˆ ˆ X α α = X αβ Eα Eβ , (17) ⊗ U U | iihh | α=0 α,β=0

Equation (13) can be proven by explicit evaluation of Equations. (15) and (16) are the unitary transformation the matrix elements and using Eq. (12). Equation (14) of the operators. Note that Eq. (16) does not imply uni- is obtained by performing the partial transpose on both † tary equivalence of Eˆα and Fˆα, i.e., Fˆβ = Wˆ EˆαWˆ for sides of Eq. (13) with respect to the second system. some unitary operator Wˆ in d, although the unitary equivalence of Eˆ and Fˆ impliesHS Eq. (16). In general, Examples of the operator basis α α Fˆ = Wˆ Eˆ Wˆ † for any unitary operator Wˆ in , even We show three illustrative examples of the operator β α d if Eq.6 (16) holds. For example,π ˆ is the operatorHS ba- basis for frequently found in the literature. The (i,j) Hd sis of all the elements of which have rank one, whereas first example is a set of transition operatorsπ ˆ(i,j) := ˆ ˆ i j with i, j = 0, , d 1, and (i, j) := di + j Uα and λα are those operator bases of all the elements 2 | ih | · · · − d −1 of which have rank exceeding one. Therefore,π ˆ(i,j) is in- [34]. The associated states form a basis set πˆα α=0 d−1 {| ii} duced by a standard state basis i for d, whereas whose elements are the tensor product of the standard {| i}i=0 H Uˆα and λˆα are not. This clearly indicates that either the state basis, i.e., πˆ(i,j) = i j . The next ex- 2 2 | ii | i ⊗ | i ˆ d −1 ˆ d −1 ample is a set of unitary irreducible representations of set Uα or λα is unitarily related to the set 2 α=0 α=0 {d −}1 { } the group SU (d) or the discrete displacement operators πˆα α=0 , but is not unitarily equivalent to this set. ˆ { } on the phase-space torus, whose elements are U(m,n) = mn/2 d−1 mk √ ω Pk=0 ω πˆ(k⊕n,k)/ d with m,n = 0, , d 1, 1/d i2π/d · · · − III. MATRIX REPRESENTATION OF where ω = 1 = e and denotes addition mod- QUANTUM OPERATIONS ulo d [34, 35, 36]. The associated⊕ states form a basis set 2 ˆ d −1 2 Uα α=0 of d -orthogonal maximally entangled states {| ⊗ii}2 2 Let us turn our attention to a single-qudit operation. in d . The last example is a set of d 1 traceless Her- As shown in Sec. I, this quantum operation can be rep- mitianH generators of the group SU (d) supplemented− with 2 resented by either an HS-space superoperator or an L- ˆ d −1 the normalized identity operator given by λα α=0 = space superoperator, in which the χ- and -matrices are { } S I/ˆ √d, uˆ0,1, uˆ0,2, , uˆd−2,d−1, vˆ0,1, vˆ0,2, , vˆd−2,d−1, introduced with respect to the arbitrary, but associated 2 2 { · · · · · · d −1 d −1 ⊗2 wˆ1, wˆ2, , wˆd−1 where d(d 1) off-diagonal genera- sets of basis Eˆ for and Eˆ for , · · · } − { α}α=0 HSd {| αii}α=0 Hd tors are given byu î,j = (ˆπ(i,j) +ˆπ(j,i))/√2,v î,j = respectively. In this section, the underlying relationship i(ˆπ πˆ )/√2 with 0 i < j d 1, between these two matrix representations is discussed. (i,j) − (j,i) ≤ ≤ − and d 1 diagonal generators are given byw ˆk = Let us first consider the L-space superoperator. In the k−−1 ( πˆ + kπˆ )/ k(k +1) with 1 k d 1 Liouville formalism, the operators in d are identified Pi=0 (i,i) (k,k) p ⊗ − ≤ ≤ − with the vectors in 2. Any operationHS S is identified [34]. The choice of basis is of course a matter of conve- Hd nience, depending on the application. ˆ ⊗2 2 2 with the one-sided operator acting on d , which can Since the associated sets πˆ d −1, Uˆ d −1, and S ˆ H ⊗2 α α=0 α α=0 be expanded using the state basis Eα for d as shown 2 {| ii} {| ii} 2 | ii2 H λˆ d −1 are state bases for ⊗2, they should be uni- in Eq. (2). The elements of a d d matrix are for- α α=0 d × S tarily{| ii} related. This implies thatH the operator basesπ ˆ , ˆ α mally written as αβ = Eˆα ˆ Eˆβ . Alternatively, the Uˆα, and λˆα should also be unitarily related. In general, SS hh |S| ii 2 2 same operation is written as a two-sided superoperator two sets of state basis Eˆ d −1 and Fˆ d −1 in acting on the operator in , which can be expanded {| αii}α=0 {| αii}α=0 HSd 5 using the operator basis Eˆα for d as shown in Eq. (1). Equations (19) and (20) are one of the main results of this ˆ ˆ HS ˆ Since ˆ Eˆβ = ˆ(Eˆβ ) , we find the matrix element αβ paper. From these equations, we find that ˆ and χˆ are can alsoS| beii written|S as ii S complementary to each other in the sense thatS they can

2 be interchanged if we exchange the two operator bases d −1 Eˆ Eˆ and Eˆ Eˆ∗ on ⊗2 in their expressions. ˆ ˆ ˆ ˆ ˆ ˆ∗ ˆ | αiihh β | α ⊗ β Hd αβ = Eα (Eβ ) = X χγδ Eα Eγ Eδ Eβ , These equations show that the -matrix (χ-matrix) is S hh |S ii hh | ⊗ | ii S γ,δ=0 ˆ ˆ (18) given by the expansion coefficients of (χˆ) with respect Sˆ where we used Eq. (5). Substituting the right-hand side to Eˆα Eˆβ as well as those of χˆ ( ˆ) with respect to ˆ ˆ | îihh∗ | S of Eq. (18) for in Eq. (2), we find that ˆ can be Eα E , which is explicitly written as αβ ⊗ β written in termsS of either the matrix or χ asS S ˆ ˆ † ˆ ˆ ˆ∗ † ˆ 2 2 χαβ = Tr( Eα Eβ ) χˆ = Tr(Eα Eβ) , (21) d −1 d −1 | iihh | ⊗ S ˆ ˆ ˆ ˆ ˆ∗ = X αβ Eα Eβ = X χαβEα Eβ. (19) S S | iihh | ⊗ α,β=0 α,β=0 ˆ ˆ † ˆ ˆ ˆ∗ † ˆ αβ = Tr( Eα Eβ ) = Tr(Eα Eβ) χ.ˆ (22) In Eq. (19), we find two types of induced operator ba- S | iihh | S ⊗ ⊗2 From Eqs. (19) and (20), we can explore the mutual sis on d having different tensor product structures, H ∗ conversion formulas between χ- and -matrices. To this that is, Kronecker products Eˆα Eˆ and dyadic products ⊗ β end, let us define a bijection betweenS the two operators Eˆα Eˆβ of the state basis associated with the operator ⊗2 2 on originally found by Havel [40]. | iihh |ˆ d −1 Hd basis set Eα α=0 . Note that both types of basis set do 2 { } ⊗2 d −1 not cover all the possible basis sets on d . Obviously, the former type of basis set covers onlyH those sets that Λ( )= (Iˆ πˆ ) (ˆπ Iˆ), (23) ⊙ X ⊗ γ ⊙ γ ⊗ are factorable with respect to the original and extended γ=0 system spaces. For example, the set of d4-dyadic products of the d2-maximally-entangled states in ⊗2 is not which is also considered to be the super-superoperator Hd acting on . Then, we have the following Theorem [40]. a factorable basis set, and can not be covered by the for- Hd mer type. Similarly, only an operator basis on ⊗2 with d Theorem 2 ˆ all elements of which have rank one can be reducedH to (Havel (2003)) For arbitrary operators X and Yˆ in , we have the latter type of the basis set. Therefore, each type of HSd basis set can describe its own particular subset of all the ∗ ⊗2 Xˆ Yˆ = Λ(Xˆ Yˆ ), (24) possible basis sets on d . | iihh | ⊗ H ˆ ⊗2 Let us next consider another operator χˆ on d , which we call the Choi operator [37]. We will showH that this operator has bijective correspondence to the L-space su- Xˆ Yˆ ∗ = Λ( Xˆ Yˆ ). (25) ˆ ⊗ | iihh | peroperator ˆ. It is known that the isomorphism be- ⊗2 S Theorem 2 connects two relevant operators on d hav- tween the operator in d and the bipartite vector in H ⊗2 HS ing different tensor product structures, i.e., the Kro- can be straightforwardly extended to the isomor- ˆ ˆ ∗ ˆ ˆ Hd necker product X Y and the dyadic product X Y . phism between the superoperator acting on d and the ⊗ | iihh | ⊗2 HS To prove the Theorem 2, we first note the following operator acting on d . Jamio lkowski first showed that lemma. H ˆ the map between the HS-space superoperator ˆ( ) and ⊗2 ˆ ˆ ⊗ S ⊙ Lemma 3 The identity operator on and the (un- the operator χˆ ˆ ˆ(dρˆ ) acting on 2 is an iso- Hd ≡ S ⊗ I I Hd normalized) density operator of the isotropic state on ˆ ˆ ⊗2 morphism, where ( ) and ( ) act on the HS-space d are related as follows. of the first and secondS ⊙ systems,I ⊙ respectively [38]. If we H note Eq. (13) and the following equivalent relation that Iˆ Iˆ = Λ(Iˆ Iˆ), (26) follows from Eq. (19) | iihh | ⊗

ˆ ˆ ∗ ˆ ˆ ˆ ˆ ˆ ˆ ∗ ′ ′ † ′′ ′′ ′′ ′′ ′ ′ X Y = (X I)(I I)(I Y ) Mα β ,αβ and Pα′′,β′′ (M )αβ,α β Mα β ;α β = ⊗ ⊗ ⊗ ⊗ † ˆ ˆ ˆ ˆ ˆ ˆ ∗ ˆ ˆ ′′ ′′ ′′ ′′ ′′ ′′ ′ ′ ′ ′ = Λ((X I) I I (I Y )) = Λ( X Y ), Pα ,β Mαβ;α β (M )α β ,α β = δαα δββ . By us- ⊗ | iihh | ⊗ | ih | ing these results, we obtain a bijection between the χ- where we used Eq. (5). Accordingly, Theorem 2 is and -matrices: S 2 proved. At this point, we note that the action of the d −1 bijection Λ ( ) corresponds to reshuffling of the matrix χ = Qγ Rγ , (33) ⊙ ˙ X S introduced by Zyczkowski and Bengtsson: if we con- γ=0 sider the matrix for the operator on ⊗2 defined with Hd respect to the standard state basis, the mapped opera- 2 d −1 tor by Λ ( ) has a reshuffled one of the original matrix γ γ ⊙ = X Q χR , (34) [20]. The bijection Λ( ) is also closely related to the ma- S trix realignment introduced⊙ by Chen and Wu to discuss γ=0 the separability criterion for the bipartite density matrix which are given by the sum of the multiplication of three [39]. It is easy to confirm that Λ( ) is involutory, that is, known matrices Qγ, χ, and Rγ, and evidently com- Λ(Λ( )) = . It should be also noted⊙ that Λ( ) does not putable. preserve⊙ Hermiticity⊙ and the rank of the transformed⊙ op- The above formulation can be straightforwardly ex- erator, so its spectrum is not preserved [20]. This means tended to describe two-qudit operations. In this case, that Λ( ) represents a non-physical operation. It follows different choices of basis sets are allowed for two sys- 2 ⊙ d −1 ˆ ˆ ˆ ˆ ˆ tems. Let us choose the set Eˆα acting on the first from Theorem 2 that χˆ = Λ( ) and = Λ(χˆ), i.e., and 2 α=0 S S S ˆ { d −}1 χˆ are bijective. qudit space and the set Fα acting on the second { }α=0 From this bijective relation, we can explore the bijec- qudit space. The general two-qudit superoperator can be tion between χ- and -matrices. To this end, we expand written as ˆ ˆ S ˆ ˆ∗ 2 Eα Eβ in terms of Eα Eβ, and vice versa. Since d −1 |theyiihh are bijective,| it follows⊗ that ˆ ˆ ˆ ˆ ˆ = X αβ,γδ Eα Fβ Eγ Fδ (35) S S | ii| iihh |hh | ˆ ˆ∗ ˆ ˆ α,β,γ,δ=0 Eα Eβ = Λ( Eα Eβ ) ⊗ | iihh | 4 2 acting on the d -dimensional L-space 4 as well as d −1 Ld ˆ ′ ˆ ′ ′ ′ 2 = X Eα Eβ Mα β ,αβ, (28) d −1 ′ ′ | iihh | ˆ † † α ,β ,γ=0 ˆ( )= χ Eˆ Fˆ Eˆ Fˆ (36) S ⊙ X αβ,γδ α ⊗ β ⊙ γ ⊗ δ α,β,γ,δ=0

ˆ ˆ ˆ ˆ∗ 2 Eα Eβ = Λ(Eα Eβ ) acting on d . These superoperators are character- HS 2 | iihh | 2 ⊗ 4 4 d −1 d −1 ized with the d d matrices [ αβ,γδ]α,β,γ,δ=0 and ∗ † 2 × S ≡ S = Eˆ ′ Eˆ ′ (M ) ′ ′ , (29) d −1 X α β α β ,αβ χ [χαβ,γδ] . The bijective Choi operator is de- ′ ′ ⊗ α,β,γ,δ=0 α ,β =0 ≡ 4 ⊗4 fined on the d -dimensional H -space d that is identi- ′ ′ 4 H where M ′ ′ is α β ; αβ-entry of the d4 d4 complex fied with the L-space d as follows: α β ,αβ × L matrix M. It is explicitly given as (13) (24) ˆ ˆ (12) (34) χˆ ˆ ˆ (d2ρˆ ρˆ ), (37) † ∗ I I M ′ ′ = Tr( Eˆ ′ Eˆ ′ ) Eˆ Eˆ ≡ S ⊗ I ⊗ α β ;αβ α β α β ⊗4 2 | iihh | ⊗ where the indices refer to the factors in in which the d −1 Hd γ γ corresponding operations have a nontrivial action [41, 42, = Q ′ R ′ , (30) X α α ββ ˆ ˆ γ=0 43]. Then, it is straightforward to show that and χˆ can be written as follows: S 2 where the coefficients d −1 ˆ ˆ ˆ ˆ ˆ γ ˆ ˆ ˆ = X αβ,γδ Eα Fβ Eγ Fδ Qαβ = Eα (I πˆγ ) Eβ , (31) S S | ii| iihh |hh | hh | ⊗ | ii α,β,γ,δ=0 2 d −1 ˆ ˆ∗ ˆ ˆ∗ γ = X χαβ,γδEα Eγ Fβ Fδ . (38) R = Eˆα (ˆπγ Iˆ) Eˆβ (32) ⊗ ⊗ ⊗ αβ hh | ⊗ | ii α,β,γ,δ=0 are the computable matrix elements of the operators 2 ⊗2 d −1 Iˆ πˆγ andπ ˆγ Iˆ on defined with respect to 2 d ⊗ ⊗ d H−1 χˆ = χαβ,γδ Eˆα Fˆβ Eˆγ Fˆδ the basis set Eˆα . These coefficients form X α=0 2 | ii| iihh |hh | 2 2 {| ii} γ γ d −1 α,β,γ,δ=0 the d d complex matrices Q [Qαβ]α,β=0 and 2 2 d −1 ×γ − ≡ Rγ [R ]d 1 . It is easy to confirm that M is Hermi- ˆ ˆ∗ ˆ ˆ∗ αβ α,β=0 = X αβ,γδEα Eγ Fβ Fδ , (39) ≡ † ∗ S ⊗ ⊗ ⊗ tian as well as unitary, i.e., (M )α′β′,αβ = (Mαβ,α′β′ ) = α,β,γ,δ=0 7

It follows that the two operators are certainly bijective: Theorem 3 The linear operation S is completely posi- ˆ ˆ χˆ = Λ Λ( ˆ) and ˆ =Λ Λ(χˆ). From these bijective tive, iﬀ the χ-matrix is positive. relations,⊗ weS can exploreS ⊗ the bijection between the χ- This is natural on physical grounds, because the Choi and -matrices for a two-qudit operation as S operator χˆ should be an unnormalized density operator 2 d −1 associated with the system which was subjected to the γ λ γ λ χ = X Q S R T , (40) quantum operation as will be discussed in the next sec- ⊗ S ⊗ γ,λ=0 tion. In addition to 3, any physical quantum operation should satisfy the following condition.

2 d −1 Condition 4 (Trace non-increasing) The physical quan- γ λ γ λ = X Q S χR T , (41) S S ⊗ ⊗ tum operation should be trace non-increasing; i.e., the γ,λ=0 mapped operator should have trace less than one. where d2 d2 matrices Qγ and Rγ are given by Eqs. × This condition is simply expressed as the restriction on (31) and (32), and the matrix entries of Sλ and T λ are (1) the χ-matrix: Tr1χ I for a single-qudit operation ˆ (2) ≤ (n) given by the matrix elements of the operators I πˆγ and Tr13χ I for a two-qudit operation, where I ˆ ⊗2 ⊗ ≤ n andπ ˆγ I on d defined with respect to the basis set is an identity matrix with size d . ⊗2 H ˆ d −1 It is needless to say that the χ- and -matrices can Fα α=0 : {| ii} be defined with respect to arbitrary operatorS basis sets. γ ˆ ˆ ˆ Sαβ = Fα (I πˆγ ) Fβ , (42) Once these matrices are given with respect to a particu- hh | ⊗ | ii lar operator basis set, they can be converted into those defined with respect to the other basis set. It is obvious γ from Eqs. (19), (20), (39), and (38) that the two matri- T = Fˆα (ˆπγ Iˆ) Fˆβ . (43) αβ hh | ⊗ | ii ces defined with respect to different bases are unitarily We can further extend the above formulation to describe equivalent. To be specific, let χE and E be the χ- and n-qudit operations. In this case, the bijective relation -matrices for a single-qudit operationS defined with re- S 2 ˆ ⊗n ˆ ˆ ⊗n ˆ d −1 F between ˆ and χˆ reads χˆ =Λ ( ˆ) and ˆ =Λ (χˆ). By spect to the operator basis set Eα α=0 , and χ and S S S F be those defined with respect{ to} the operator basis using these relations, we can straightforwardly extend 2 S ˆ d −1 Eqs. (40) and (41) to the bijection between the χ- and set Fα α=0 , where the two bases are unitarily related -matrices for an n-qudit operation. as shown{ } in Eqs. (15) and (16). Then, these matrices S It is obvious that not all the space of χ- and -matrices should be written as corresponds to physically realizable operations.S For ex- F = † E , (44) ample, we can describe an anti-unitary operation by us- S U S U ing the χ- and -matrices, which is evidently an unphysi- cal operation. TheS requirement for the χ- and -matrices S to represent physical quantum operations has been ex- χF = †χE , (45) tensively studied by many researchers. In the following, U U 2 the requirements common for the single- and two-qudit d −1 2 2 where [ αβ ]α=0 is a d d unitary matrix. For the operations are summarized [21]. case ofU two-qudit ≡ U operation,× we need to extend the set Condition 1 (Hermiticity) The physical quantum oper- of the operator basis to cover all the possible basis sets ation S should preserve Hermiticity; i.e., S maps any defined for the two-qudit operator space, that is, one that Hermite operator into an Hermite operator. is a factorable set as well as not a factrorable set with respect to the first and second systems. The general set 4 Condition 2 (Positivity) The physical quantum opera- ˆ d −1 ⊗2 of operator basis Φγ γ=0 on d should be unitarily tion S should be positive; i.e., S maps any positive op- { } H related to the factorable operator bases Eˆα Fˆβ. If we erator into an positive operator. ⊗ 4 4 4 d −1 introduce a d d unitary matrix [ αβ]α=0 that Condition 3 (Complete positivity) The physical quan- × U ≡ U relates Φˆ γ and Eˆα Fˆβ: tum operation S should be completely positive; i.e., posi- ⊗ 2 tivity is preserved if we extend the L-space and HS-space d −1 ˆ ˆ ˆ ˆ ˆ by adding more qudits. That is, the superoperator ˆ ˆ Φγ = X Eα Fβ [α,β]γ , (46) ⊗ U on the extended spaces should be positive. S⊗ I α,β=0 It is known that 3 is sufficient for 1 and 2, and 2 is suffi- where [α, β] := d2α + β, it follows that Eqs. (44) and cient for 1. Therefore, we require complete positivity for (45) also hold for two-qudit operations. a physical quantum operation. Complete positivity can The χ- and -matrices can be diagonalized by choosing be expressed as a particularly simple condition for the the appropriateS operator basis sets, but are not neces- χ-matrix. sarily diagonalized simultaneously by a unique set. The 8 operator basis set that diagonalizes the χ-matrix, each a sequence of local measurements assisted by classical element of which is multiplied by the square root of the communication [7, 9, 11, 31, 42]. associated eigenvalue, forms a particular set of Kraus op- Consider first the identification of a single-qudit opera- erators in the Kraus form of the quantum operation. Any tion. Equations (19) and (20) show that all the elements set of Kraus operators can be obtained by noting the uni- of the χ- and -matrices are given by the expansion co- S tary freedom in the Kraus form [2]. It follows from Eqs. ˆ ˆ (19), (20), (39), and (38) that the same operator basis efficients of the operators χˆ or with respect to two different types of operator basisS on ⊗2. Of these two set with the associated set of eigenvalues also gives an Hd operator-Schmidt decomposition for the L-space superoperators, the Choi operator χˆ is particularly useful since ˆ it is a positive operator associated with the physical state operator ˆ [20]. Therefore, the Kraus rank for the HS- S ˆ of the bipartite object. To be specific, it can be inter- space superoperator ˆ( ) and Schmidt number of the preted as the unnormalized output state from the system S ˆ⊙ L-space superoperator ˆ must be equal. in question where qudit 1 of the two qudits prepared in S an isotropic state is input into the system and undergoes the quantum operation S while qudit 2 is left untouched. IV. APPLICATIONS OF χ- AND S-MATRICES Therefore, we can prepare the output state corresponding ˆ ˆ to the normalized Choi operator χ/dˆ ˆ ˆ(ˆρ ) with the ≡ S⊗I I This section presents the several applications of the χ- use of several copies of the isotropic-state input for the and -matrices. We discuss how these matrices and the two qudits. Thus, the identification of a single-qudit op- presentS formulation are useful for analysis and design of eration reduces to the identification of a two-qudit state. quantum operations. It follows from Eq. (22) that every element of the - matrix can be directly obtained by determining the ex-S Experimental identification of quantum opera- pectation value of the corresponding product operator basis Eˆ Eˆ∗ for the output states after the quantum tions h α ⊗ βi ˆ In the first example, we explain how useful the present operation has taken place. If the basis Eα is chosen to formulation is for experimental identification of quantum be the Hermitian operator basis λˆα, it suffices to make operations [6, 7, 8, 9, 10, 11, 12]. This task is impor- a set of d4-independent local measurements assisted by tant because the development of any quantum device classical communication to determine the whole set of the or circuit for quantum computation and communication, real expectation values λˆα λˆβ . Accordingly, we can which can be considered as an input-output system that obtain the -matrix definedh ⊗ withi respect to the Hermi- 2 S ˆ d −1 performs an intended quantum operation on its input tian operator basis set λα α=0 . Once the -matrix is state and transforms it into its output state, necessarily obtained, it is easy to convert{ } it to the χ-matrixS defined requires experimental benchmarking of its performance. with respect to the same basis by using Eqs. (31)-(33) The identification of a two-qudit device is particularly and also into the χ- and -matrices defined with respect interesting from a practical viewpoint as well as a scien- to the arbitrary chosen basisS by using Eqs. (44) and (45). tific one because it may involve a nonseparable operation The identification of a two-qudit operation can be which has a purely quantum mechanical nature, i.e., it carried out in the same way as the identification of a cannot be simulated by using any classical method. single-qudit operation. In this case, we prepare the state Identification of an input-output system amounts to corresponding to the Choi operator in Eq. (37) with the identifying its χ- or -matrix, since these matrices char- S use of several copies of the product of isotropic states acterize the system in question completely as far as input prepared in the four qudits. To prepare the output and output data are concerned. The evaluated matrices state, we initially prepare the product of isotropic states should reproduce the behavior of the system well enough ρˆ(12) ρˆ(34) in two pairs of two qudits (qudits 1-2 and when the system is stimulated by any class of inputs of I I qudits⊗ 3-4). Then qudit 1 and qudit 3 are input into the interest, and they should be useful for engineering the system in question, undergo the quantum operation S system of interest, e.g., to permit control of the system, jointly while the other qudits are left untouched. This to allow transmission of information through the system, 2 to yield predictions of future behavior, etc. Identification setup leads to an output state χ/dˆ in four-qudits. Thus, problems are commonly regarded as inversion problems, the identification of a two-qudit operation reduces to the where the χ- or -matrix is to be statistically estimated identification of a four-qudit state. It follows from Eq. from incompleteS prior knowledge of the system, using (38) that every element of the -matrix for the two-qudit S prior knowledge of corresponding inputs and the collec- operation can be directly obtained by determining the tion of data obtained by measurement of outputs that real expectation value of the corresponding product usually contain noise. In what follows, it will be shown operator basis λˆα λˆγ λˆβ λˆδ for the output states h ⊗ ⊗ ⊗ i that this common belief is not the case for identification after the quantum operation has taken place, if all the of quantum operations. To be specific, we can estimate relevant basis sets are chosen to be the Hermitian opera- 8 both the χ- and -matrices without any inversion pro- tor basis λˆα. It suffices to make a set of d -independent cedure if we can makeS use of an entangled resource and local measurements assisted by classical communication 9 to determine the whole set of real expectation values of the system in question [10, 18, 44, 45]. ˆ ˆ ˆ ˆ λα λγ λβ λδ . Accordingly, we can obtain the The next example concerns the analysis of a quantum h -matrix⊗ ⊗ defined⊗ withi respect to the Hermitian operator 2 operation acting on the composite system. As mentioned S d −1 basis λˆα λˆβ . The -matrix can be converted before, the χ-matrix of a two-qudit operation is inter- { ⊗ }α,β=0 S into the χ-matrix by using Eqs. (40)-(43). The - esting from a practical as well as a scientific viewpoint. and χ-matrices defined with respect to arbitrary chosenS The Jamiolkowski isomorphism for a two-qudit opera- bases can be obtained by applying the appropriate tion (Eqs. (37) and (38)) implies that the notion of matrix unitary transformation. entanglement can be extended from quantum states to quantum operations. Analogously to what happens for Matrix analysis of quantum operations states, quantum operations on a composite system can This section discusses in what way the χ- and - be entangled [46, 47]. A quantum operation acting on matrices contribute to developing quantum devices andS two subsystems is said to be separable if its action can circuits for quantum computation and communication. be expressed in the Kraus form We consider two classes of applications in which these matrices offer useful mathematical models for quantum ˆ ˆ ˆ ˆ ˆ † ( )= X (Ai Bi) (Ai Bi) , (47) operations. The first one concerns physical and infor- S ⊙ ⊗ ⊙ ⊗ mation theoretic analysis of quantum operations and the i other concerns a logical calculus of quantum circuits or algorithms comprised of a sequence of quantum opera- where Aî and Bî are operators acting on each subsystem tions. [48, 49, 50]. Otherwise, we say that it is nonseparable Let us first consider the physical and information the- (or entangled). Quantum operations that can be per- oretic analysis of quantum operations. For this purpose, formed by local operations and classical communications it is preferable to use the χ-matrix. This stems partly (the class of LOCC operations) are described by separa- from the fact that the χ-matrix is positive and isomor- ble quantum operations, yet there are separable quantum phic to the density matrix in the doubled Hilbert space. operations that cannot be implemented with LOCC op- Physically, the diagonal elements of the process matrix erations with probability one [41, 43, 51]. Anyway, these show the populations of, and its off-diagonal elements are useless for creating entanglement in an initially un- show the coherences between, the basis operators mak- entangled system. It has been pointed out by several ing up the quantum operation, analogous to the inter- authors that the separability and entangling properties pretation of density matrix elements as populations of, of quantum operations acting on two systems can be dis- and coherences between, basis states. Owing to Jami- cussed in terms of the Choi operator for two-qudit op- olkowski isomorphism, the dynamic problems concerning erations (Eq. (38)) [41, 42, 43]. In the present context, quantum operations can be turned into kinematic prob- this reduces to discussing the separability properties of lems concerning quantum states in a higher dimensional the χ-matrices. For example, there is a condition for space, and one can make use of a well-understood state- the χ-matrix equivalent to Eq. (47): a quantum opera- based technique for analyzing the quantum operation. In tion acting on two subsystems is separable if its χ-matrix what follows, we show several illustrative examples and (A) (B) (A) can be written as χ = Pi χi χi , where χi and interesting problems from the physical and information (B) ⊗ theoretic viewpoints. χi are the χ-matrices for the quantum operation act- The first example concerns the fidelity or distance mea- ing on each subsystem. Thus, the separability of general sure between two quantum operations. Several measures quantum operations acting on the composite system is that make use of the above isomorphism have been pro- reduced to the separability of its χ-matrix. Since the separability criterion and measure for the general d4 d4 posed to quantify how close the quantum operation in × question is to the ideal operation (usually a unitary op- positive matrix is not fully understood, it remains as an eration) we are trying to implement. For example, the important problem for quantum information science to state fidelity defined between two states is extended to find such a criterion and measure for general two-qudit quantum operations. compare the two operations. The process fidelity Fp is defined by using the χ-matrixχ ˜ of the system in ques- Let us turn our attention to a logical calculus of quan- tion and the rank one χ-matrix χideal of the ideal system tum circuits or algorithms. For this purpose, the - 1 S in the state-fidelity formula, that is, Fp = d2n Tr˜χχideal, matrix is practically useful. This follows from the fact where n=1 for single-qudit operation and n=2 for a two- that L-space superoperator algebra works just like Dirac qudit operation. The average gate fidelity F¯ defined as operator algebra. For example, consider the scenario in the state fidelity between the output state after the quan- which two-quantum operations S1 and S2 act sequen- tum operation and the ideal output can be calculated tially on a quantum system. Assume that the associated from the process fidelity. The purity defined for the den- -matrices are given with respect to the same operator S ˆ N−1 2 sity matrix can be extended to characterize how much basis set Eα α=0 in N , where N = d for single- of a mixture the quantum operation introduces, which is qudit operation{| ii} and N =L d4 for two-qudit operation. also represented by the simple function of the χ-matrix Then the composite operation S = S S is described 1 ◦ 2 10 by the multiplication of L-space superoperators The present formulation will offer us the way of building

N−1 bridges across the two applicational classes. For exam- ˆ ˆ ˆ ˆ = ˆ ˆ = Eˆ Eˆ , (48) ple, the entangling properties of the quantum circuits S S1S2 X Sαβ| αiihh β | comprised of a sequence of single- and two-qudit quan- α,β=0 tum operations acting on several qudits can be discussed. where The present formulation will also help us to analyze and N−1 benchmark the quantum operation realized in the actual = ( ) ( ) . (49) device. Sαβ X S1 αγ S2 γβ γ=0 The extension to the case in which a sequence of a finite number of quantum operations is applied to the same V. CONCLUSIONS quantum system is trivial. Equation (49) implies that the -matrix of the composite operation reduces to the We have considered two matrix representations of multiplicationS of the -matrices of the individual opera- single- and two-qudit quantum operations defined with tions. This makes itS practically advantageous using the respect to an arbitrary operator basis, i.e., the χ- -matrix to make a logical calculus of quantum circuits or and -matrices. We have provided various change-of- S algorithmsS comprised of a sequence of elementary single- representation formulas for these matrices including bi- and two-qudit quantum operations. jections between the χ- and -matrices. These matrices S Consider next the quantum circuit or algorithm com- are defined with the expansion coefficients of two oper- prised of a sequence of quantum operations each of which ators on a doubled Hilbert space, that is, the L-space acts not necessarily on the same quantum system. This superoperator and the Choi operator. These operators offers a general model for the quantum circuit acting on are mutually convertible through a particular bijection the large numbers of qudits [3]. To analyze and design by which the Kronecker products of the relevant opera- such a quantum circuit, we need to consider quantum tor basis and the dyadic products of the associated state operations acting on the whole set of qudits and associ- basis are mutually converted. From this fact, the mutual ated extended -matrix. Such an extended -matrix is conversion formulas between two matrices are established non-trivial, butS its bijective χ-matrix is triviallyS obtained as computable matrix multiplication formulas. Extention by taking the tensor product with identity matrix, that to multi-qudit quantum operation is also trivial. These is, the χ-matrix for the identity operation on the irrele- matrices are useful for their own particular classes of ap- vant system. If the χ-matrix for the quantum operation plications, which might be interesting from a practical as S on the relevant space is given by χ, the χ-matrix for well as a scientific point of view. the extended quantum operation is given by χ I. On We have presented possible applications of the present the other hand, we can trivially extend the conversion⊗ formulation. By using the present formulation, an exper- formulas (40) and (41) to those formulas for the quan- imental identification of a quantum operation can be re- tum operation acting on more qudits. Therefore, we can duced to determining the expectation values of a Hermi- calculate the extended -matrix for each quantum op- tian operator basis set on a doubled Hilbert space. This eration from the associatedS -matrix for the quantum can be done if we prepare several copies of the isotropic- operation acting on the relevantS space. The -matrix for state input or the product of isotropic states input. By a sequence of operations acting on the spaceS of the whole using the χ-matrix, we can make a physical as well as quantum systems can be calculated by multiplication of a quantum information theoretic characterization of the the extended -matrices for the individual quantum op- quantum operation. In particular, the χ-matrix is useful erations. S to discuss the entangling properties of the quantum oper- The -matrix analysis of the quantum operation has ation acting on the composite system, since the problem the followingS potential advantage. It can deal with a of the separability of the quantum operation is reduced non-unitary operation in which mixed state evolution to the problem of the separability of the χ-matrix. On occurs. Noisy quantum operation, probabilistic subrou- the other hand, the –matrix is useful when we discuss tines, measurements, and even trace-decreasing quantum the typical quantumS circuit comprised of a sequence of filters can be treated. This is in contrast to the usual single- and two-qudit quantum operations each of which analysis based on unitary matrix which can deal only acts on different quantum qubits. It is possible by consid- with unitary gate in which only the pure state evolution ering the extended –matrix of each quantum operation is allowed. It thus offers us an mathematical model to acting on the wholeS state space of the relevant qudits. analyze and design the logical operation of wider range Such extended –matrices can be calculated from the as- of the complex quantum circuits and algorithms [3]. sociated, bijectiveS χ-matrices by taking the tensor prod- In the above discussion, we considered two applica- uct with the appropriate identity matrix. Accordingly, tional classes, i.e., physical and information theoretic we can calculate the –matrix for a quantum circuit by analysis and logical calculus of quantum operations, in multiplying the extendedS –matrices of each operation. which the χ- and -matrices matrices offer useful mathe- This should be very usefulS to analyze and design a wide matical models. TheyS have their own useful applications. range of the quantum circuits and algorithms involving 11 non-unitary operation. the CREST program of the Japan Science and Technol- We thank Satoshi Ishizaka and Akihisa Tomita for ogy Agency. their helpful discussions. This work was supported by

[1] K. Kraus, States Effects and Operations: Fundamental [28] A. Ben-Reuven, Adv. Chem. Phys. 33, 235 (1975). Notions of Quantum Theory, Lecture Notes in Physics [29] A. Ben-Reuven, Phys. Rev. A 22, 2572 (1980); 22, 2585 Vol. 190 (Springer-Verlag, Berlin, 1983). (1980). [2] M.A. Nielsen, I.L. Chuang, Quantum Computation and [30] Y. Prior and A. Ben–Reuven, Phys. Rev. A 33, 2362 Quantum Information (Cambridge University Press, Ox- (1986). ford, 2000). [31] G. M. D’Ariano, and P. Lo Presti, Phys. Rev. Lett. 86, [3] D. Aharonov, A. Kitaev, and N. Nisan, e-print 4195 (2001); 91, 47902 (2003). quant-ph/9806029. [32] M. Horodecki and P. Horodecki, Phys. Rev. A 59, 4206 [4] P. Stelmachovi˘cand˘ V. Bu˘zek, Phys. Rev. A 64, 062106 (1999). (2001); 67, 029902(E) (2003). [33] B. M. Terhal and K. G. H. Vollbrecht, Phys. Rev. Lett. [5] H. Hayashi, G. Kimura, and Y. Ota, Phys. Rev. A 67, 85, 2625-2628 (2000). 062109 (2003). [34] G. Mahler and V. A. Weberruß, Quantum Networks: Dy- [6] Y. Nambu et al., Proc. SPIE Int. Soc. Opt. Eng. 4917, namics of Open Nanostructures, 2nd ed. (Springer, New 13 (2002). York, 1998). [7] J. B. Altepeter et al., Phys. Rev. Lett. 90, 193601 (2003). [35] C. Miquel, J. P. Paz, and M. Saraceno, Phys. Rev. A 65, [8] M. W. Mitchell, C. W. Ellenor, S. Schneider, and A. M. 062309 (2002). Steinberg , Phys. Rev. Lett. 91, 120402 (2003). [36] M. L. Aolita, I. Garc´ıa-Mata, and M. Saraceno, Phys. [9] F. De Martini et al., Phys. Rev. A 67, 062307 (2003). Rev. A 70, 062301 (2004). [10] J. L. O’Brien et al., Phys. Rev. Lett. 93, 080502 (2004). [37] M.-D. Choi, Linear Algebr. Appl. 10, 285 (1975). [11] V. Secondi, F. Sciarrino, and F. De Martini, Phys. Rev. [38] A. Jamiolkowski, Linear transformations which preserve A 70, 040301(R) (2004). trace and positive semidefiniteness of operators, Rep. [12] Y. Nambu and K. Nakamura, Phys. Rev. Lett. 94, 010404 Math. Phys. 3, 275 (1972). (2005). [39] K. Chen, and L-A.Wu, Phys. Lett. 306, 14 (2002); Quan- [13] C. M. Caves, J. Superconductivity 12, 707 (1999). tum Info. Comp., 3 193 (2003). [14] V.E. Tarasov, J. Phys. A. 35, 5207 (2002). [40] T. F. Havel, J. Math. Phys. 44, 534 (2003). [15] L. P. Hughston, R. Jozsa, and W. K. Wootters, Phys. [41] J. I. Cirac, W. Dür, B. Kraus, and M. Lewenstein, Phys. Lett. 183, 14 (1993). Rev. Lett. 86, 544 (2001). [16] G. M. D’Ariano, P. Lo Presti, and M. F. Sacchi , Phys. [42] W. Dür and J. I. Cirac, Phys. Rev. A 64, 12317 (2001). Lett. A 272, 32 (2000). [43] A. W. Harrow and M. A. Nielsen, Phys. Rev. A 68, 12308 [17] I. L. Chuang and M. A. Nielsen, J. Mod. Opt. 44, 2455 (2003). (1997). [44] M. Raginsky, Phys. Lett. A 290, 11 (2001). [18] A. G. White et.al., e-print quant-ph/0308115. [45] A. Gilchrist, N. K. Langford, and M. A. Nielsen, e-print [19] E. C. G. Sudarshan, P. M. Mathews, and J. Rau., Phys. quant-ph/0408063. Rev.,121, 920 (1961). [46] P. Zanardi, Phys. Rev. A 63, 040304(R) (2001). [20] K. Zyczkowski˙ and I. Bengtsson, Open Syst. Inf. Dyn. [47] X. Wang and P. Zanardi, Phys. Rev. A 66, 044303 11, 3-42 (2004). (2002). [21] P. Arrighi and C. Patricot, Annals Phys. 311, 26 (2004). [48] V. Vedral, M.B. Plenio, M.A. Rippin, and P.L. Knight, [22] K. Blum, Density Matrix Theory and Applications Phys. Rev. Lett. 78, 2275 (1997). (Plenum Press, New York, 1981). [49] H. Barnum, M.A. Nielsen, and B.W. Schumacher, Phys. [23] A. Royer, Phys. Rev. A 43, 44 (1991); 45, 793 (1992). Rev. A 57, 4153 (1998). [24] S. W. Hawking, Phys. Rev. D 14, 2460 (1976). [50] E. M. Rains, Phys. Rev. A 60, 173 (1999). [25] R. M. Wald, Phys. Rev. D 21, 2742 (1980). [51] C.H. Bennett, D.P. DiVincenzo, C.A. Fuchs, T. Mor, [26] A. Fujiwara and P. Algoet, Phys. Rev. A 59, 3290 (1999). E. Rains, P.W. Shor, J.A. Smolin, and W.K. Wootters, [27] A. Ben-Reuven, Phys. Rev. 141, 34 (1966) . Phys. Rev. A 59, 1070 (1999).