Protecting quantum systems from decoherence with unitary operations
E.O. Kiktenko,1, 2, 3 A.S. Mastiukova,1, 2 and A.K. Fedorov1, 2 1Russian Quantum Center, Skolkovo, Moscow 143025, Russia 2Moscow Institute of Physics and Technology, Dolgoprudny, Moscow Region 141700, Russia 3Department of Mathematical Methods for Quantum Technologies, Steklov Mathematical Institute of Russian Academy of Sciences, Moscow 119991, Russia (Dated: February 25, 2020) Decoherence is a fundamental obstacle to the implementation of large-scale and low-noise quantum information processing devices. In this work, we suggest an approach for suppressing errors by employing pre-processing and post-processing unitary operations, which precede and follow the action of a decoherence channel. In contrast to quantum error correction and measurement-based methods, the suggested approach relies on specifically designed unitary operators for a particular state without the need in ancillary qubits or post-selection procedures. We consider the case of decoherence channels acting on a single qubit belonging to a many-qubit state. Pre-processing and post-processing operators can be either individual, that is acting on the qubit effected by the decoherence channel only, or collective, that is acting on the whole multi-qubit state. We give a classification of possible strategies for the protection scheme, analyze them, and derive expressions for the optimal unitary operators providing the maximal value of the fidelity regarding initial and final states. Specifically, we demonstrate the equivalence of the schemes where one of the unitary operations is individual while the other is collective. We then consider the realization of our approach for the basic decoherence models, which include single-qubit depolarizing, dephasing, and amplitude damping channels. We also demonstrate that the decoherence robustness of multi-qubit states for these decoherence models is determined by the entropy of the reduced state of the qubit undergoing the decoherence channel.
I. INTRODUCTION Pre-processing Post-processing Decoherence Input state unitary operation unitary operation Output state U V ⇢ | i A concept of quantum internet relies on the ability of E quantum computers to process and exchange an essen- tially arbitrary number of quantum states of any struc- Figure 1. Error suppression scheme based on pre-processing and post-processing unitary operations, which are designed tures [1]. Although key building blocks both for quantum specifically for the input state and decoherence channel in computing devices and quantum communication systems order to maximize the fidelity of the output state. have been developed [2–6], the current challenge is to scale such devices with respect to the number of qubits inside quantum computers and the distance of quantum tanglement in a distributed quantum state after its degra- state transfer [6]. A major barrier is protecting quantum dation due to losses in communication lines, which can be systems from decoherence, which is the main source of solved efficiently with the use of quantum catalysis [35]. errors in quantum information processing devices [7–12]. Unfortunately, the entanglement level increases at the An ultimate solution for eliminating the decoherence ef- cost of employing post-selection. fect on quantum systems would be the use of quantum Here we propose a technique for protecting quantum error correction [13–22]. However, existing methods are states from decoherence, which is free from using an- hardly implementable since they typically require addi- cillary information carriers or post-selection procedures tional resources, such as ancillary qubits, or they are de- (see Fig.1). First, we use a pre-processing procedure for manding from the viewpoint of code sizes [17–22]. arXiv:1907.01971v2 [quant-ph] 23 Feb 2020 preparing a given known quantum state of the system Here, instead of considering quantum error correction in a specific form. Next, we use a post-processing opera- methods, we stress on possible methods for error sup- tion, which follows the action of a decoherence channel. pression that would reduce, but not eliminate the ef- These operations can be implemented as unitary oper- fect of decoherence in transferring quantum states. This ators, and their particular form can be efficiently con- problem has been studied in recent decades in various structed based on prior knowledge of the state under the aspects. Particular schemes include but not limited to protection and decoherence channel, i.e. the method is the use of non-Gaussian states [23], controlling informa- state-dependent. We study the case of local decoherence tion asymmetry in multi-qubit systems [24–27], as well as channels, which act on a particular qubit of the n-qubit employing quantum Zeno effect [28–30], specific measure- input state. The pre-processing and post-processing uni- ments [31–34] and many other methods control. These tary operators are considered to be either individual, that methods are useful under certain conditions, however, is acting on the same qubit, or collective, that is acting they encounter some difficulties for the experimental re- on the whole n-qubit state. As a model for decoherence alization. A particular task is improving the level of en- processes, we consider single-qubit depolarizing, dephas- 2
(κ) ing, and amplitude damping channels. In our consider- In the general case, the state ρ1,...,n has less than one ation, the main studied characteristic is the fidelity re- fidelity with respect to Ψ 1,...,n. garding input and output states. We present an analysis The main goal of our| worki is to develop a method for of possible strategies for the protection scheme and derive improving the fidelity of the state after decoherence by expressions for the optimal unitary operators providing means of employing unitary operations acting on qubits. the maximal value of the fidelity regarding initial and fi- We study using two types of unitary operations: nal states. Specifically, we show the equivalence of the individual operations acting on κth qubit only; schemes where one of the unitary operations is individual • while the other is collective. collective acting on the whole n-qubit system. We also consider relations between the form of the re- • Specifically, we consider employing two unitary opera- duced state of the qubit undergoing an individual deco- tions just before and after an impact of the decoherence herence process and losses of the fidelity of the whole channel . We refer to these operations as pre-processing n-qubit state. We demonstrate that in the case of per- and post-processingE unitary operators. Since each of the forming pre- and post-processing this kind of the robust- processing operators can be either individual or collec- ness to decoherence is determined by the linear entropy tive, we obtain four possible strategies: the reduced state. This feature appears to be crucial in the case where one can select a part of the whole quan- 1. ‘both individual’ scheme: both operations are indi- tum state, which is affected by the decoherence. It is vidual; important to note that the suggested scheme is able in 2. ‘individual-then-collective’ scheme: pre-processing principle to supplement existing error correction and er- operation is individual, while the post-processing ror suppression techniques [13–22, 36, 37]. operation is collective; Our paper is organized as follows. In Sec.II, we present a general framework for a method of protecting quan- 3. ‘collective-then-individual’ scheme: pre-processing tum states from decoherence with unitary operations. In operation is collective, while the post-processing Sec.III, we apply the presented method for the case of operation is individual; acting basic single-qubit decoherence channels, which in- 4. ‘both collective’ scheme: scheme with two collective clude depolarizing, dephasing, and amplitude damping operations. channels. In Sec.IV, we discuss the obtained results and conclude. We denote κth qubit as Q, while the subsystem of all the rest qubits 1, . . . , n / qi as R. Using the Schmidt decomposition,{ the initial} state{ } (1) can be written in the II. PROTECTING STATES FROM following form: DECOHERENCE WITH UNITARY OPERATIONS Ψ = λ ψ ζ , (4) | iQR i | iiQ ⊗ | iiR i=0,1 Consider a n-qubit quantum system, which consists of X p n subsystems with indices 1, . . . , n. Let the whole system where a pair of states ψ0 , ψ1 forms the orthonormal be initialized in a pure state written in the following form: basis in the Hilbert space{| i of| Qi}, ζ , ζ are two or- {| 0i | 1i} thogonal normalized vectors in the space of R, λ0 and λ1 Ψ = cx x , (1) are non-negative real numbers such that λ0 + λ1 = 1 and | i1,...,n | i1,...,n x 0,1 n λ0 λ1, which always can be achieved by an appropriate ∈{X} choice≥ of ψ and ζ . To characterize the informa- {| ii} {| ii} where complex coefficients cx obey a standard normal- tion properties of the considered quantum system, we { }2 ization condition x 0,1 n cx = 1. Then let a particu- also employ linear entropy, ∈{ } | | lar κth qubit (κ 1, . . . , n ) undergo a decoherence pro- 2 P Slin(ρ) := 1 tr(ρ ), (5) cess described by∈ { a completely} positive trace-preserving − (CPTP) map , which we refer as a decoherence chan- von Neumann entropy, nel. The initialE state (1) then turns into a new (generally, SvN(ρ) := tr(ρ log ρ), (6) mixed) state as follows: − and min-entropy, (κ) ρ1,...,n = Id 1,...,n / κ κ[ Ψ 1,...,n Ψ ], (2) S (ρ) := log Λ (ρ), (7) { } { } ⊗ E | i h | min − max where Id 1,...,n / i denotes the identical channel acting where log stands for base-2 logarithm, ρ is a density ma- { } { } on all qubits except κth and κ is a decoherence channel trix of arbitrary dimension, and Λmax(ρ) is operator giv- acting on κth qubit. As the mainE target characteristic, we ing the largest eigenvalue of ρ. We note that von Neu- consider the fidelity regarding input and output states, mann entropy and min-entropy are special cases of the which is given by the following expression: R´enyi entropy:
1 α (κ) Sα(ρ) := log tr(ρ ) (8) F := Ψ 1,...,n ρ1,...,n Ψ 1,...,n . (3) 1 α h | | i − 3 for α 1 and α + , correspondingly. Using the fact that ζ and ζ are orthogonal and → → ∞ | 0i | 1i The initial reduced state of qubit Q is as follows: normalized, we obtain the following expression for the resulting fidelity:
ρQ := trR Ψ Ψ = λi ψi ψi . (9) | iQR h | | iQ h | dec i=0,1 Find ind = Ψ (Vind I)ρ (V † I) Ψ (17) X h |QR ⊗ QR ind ⊗ | iQR = λiλj χi Q [ φi Q φj ] χj , (18) The considered entropies for the state ρQ then take the h | E | i h | | iQ i,j following forms: X n 1 Q where I stands for 2 dimensional identity operator S := S (ρ ) = λ log λ (1 λ ) log(1 λ (10)), − vN vN Q − 0 0 − − 0 − 0 acting in the space of the subsystem R. SQ := S (ρ ) = 1 λ2 = 2λ (1 λ ), (11) lin lin Q − i=0,1 i 0 − 0 The maximal value of the fidelity can be obtained Q by optimization over both unitary operators and can be S := Smin(PρQ) = log λ0. (12) min − written in the following form: All these quantities can be expressed via single param- opt eter λ only, and they monotonously decrease with λ F := max Find ind 0 0 ind ind U ,V as it ranges from 1/2 to 1. Thus, there are one-to-one ind ind correspondences between linear entropy, von Neumann (19) entropy, and min-entropy for the qubit case. We also = max λiλj χi Q [ φi Q φj ] χj Q . φi , χi h | E | i h | | i note that for the pure state Ψ the von Neumann en- {| i} {| i} i,j | iQR X tropy SQ equals to the entropy of entanglement and the vN We note that F opt depends on λ defining the initial quantum discord [27, 38–40] with respect to the R Q ind ind { i} partitioning. − reduced state of the qubit Q. Below all four strategies for our protection scheme that depicted in Fig.2 are considered, and corresponding val- ues of maximal achievable fidelity are derived. B. ‘Individual-then-collective’ scheme
The second scheme, which is depicted in Fig.2(b), be- A. ‘Both individual’ scheme gins with applying the individual unitary operator Uind and decoherence channel in the same way as in the E Here we consider the scheme depicted in Fig.2(a) step ‘both individual’ scheme. We then obtain the same state by step. First, we apply, as a pre-processing procedure, of the system given by Eq. (15). We can write its spectral the following individual unitary operation Uind: decomposition given by the following expression:
2n 1 Uind = φi ψi , (13) dec − | i h | ρ = pi ξi ξi , (20) i=0,1 QR | iQR h | X i=0 X where φi is a new orthonormal basis in the qubit 2n 1 n {| i} where ξi − forms an orthonormal basis in whole 2 - space. After acting operation Uind, the state of the whole {| i}i=0 system is as follows: dimensional Hilbert space, and we assume that pi pj for i > j. Since we are able to apply an arbitrary unitary≤ transformation V to the whole state of the system, we Φ = λ φ ζ . (14) col | iQR i | iiQ ⊗ | iiR can achieve the maximal possible fidelity (for a fixed form i=0,1 dec X p of Uind) equal to the largest eigenvalue of ρQR. It can be dec Then, after the action of the decoherence channel , written using min-entropy of ρQR in the following form: state (14) transforms as follows: E dec prelim dec Smin(ρQR) Find col = p0 = Λmax(ρQR) = 2− . (21) ρdec = [ φ φ ] ζ ζ . (15) QR E | iiQ h j| ⊗ | iiR h j| i,j This value of the fidelity can be achieved with Vcol trans- X forming the pure state with the highest eigenvalue ξ | 0iQR As a post-processing procedure, we apply the second into the initial state Ψ . Thus, we arrive to the opti- | iQR individual unitary operator Vind of the following form: mal form of Vcol given by:
opt V = ψ χ , (16) V = Ψ ξ0 + [...], (22) ind | ii h i| col | i h | i=0,1 X where [...] stands for any appropriate remaining part of where χ is another orthonormal basis. the unitary operator. {| ii} 4
Post-processing operations
Individual Collective R (a) R (c)
Vcol | i Q | i Q Uind Vind Uind E E Individual
R (b) R (d) U U V | i col | i col col
Pre-processingoperations Q Q Collective Vind E E
Figure 2. Schemes for protecting a pure state |ΨiQR from the decoherence channel E, which acts on a particular qubit Q by employing unitary operations. In (a) ‘both individual’ scheme with both unitary operators are individual and they act on Q is presented. In (b) ‘collective-then-individual’ scheme with the first operator is individual, while the second is non-individual, is shown. In (c) ‘individual-then-collective’ scheme with the first operator is collective, while the second is individual, is presented. In (d) the ‘both collective’ scheme with the both operators are non-individual is shown.
The maximal value of the fidelity for the whole where is a conjugate map to of the following form: ‘individual-then-collective’ scheme can be obtained by re- E E maining optimization over U : [ρ] := A† ρAk. (27) ind e E k dec k opt prelim Smin(ρQR) X Find col := max Find col = max 2− e Uind φi The maximal fidelity in the ‘collective-then-individual’ {| i} h i (23) scheme is given by the similar expression as in Eq. (23): dec = exp ln(2) min Smin(ρQR) . − φi {| i} opt dec Fcol ind := exp ln(2) min Smin(ρQR) , (28) − φi {| i} C. ‘Collective-then-individual’ scheme where e
dec The third scheme is shown in Fig.2(c) can be consid- ρQR := [ φi φj ] ζi ζj . (29) E | iQ h | ⊗ | iR h | i,j ered as the second scheme in the reverse order. To study X this scheme it is useful to write the action of the map e We thene show that the maximal fidelities is given by via Kraus operators as follows: E Eq. (28) and Eq. (23) are the same. Let ξopt and opt opt | i φ are such that F achieves its maximal value [ρ] = AkρAk† , (24) i ind col E of{| thei} following form: Xk where Ak obeys the standard CPTP condition F opt = λ λ { } ind col i j× A† Ak = 1, where 1 stands for identity operator in k k k,i,jX (30) the qubit space. opt opt opt opt P ξ A φ φ A† ζ ζ ξ . The resulting fidelity for some Ucol and Vind takes the × h |QR k | i i h j | k ⊗ | iiR h j| | iQR following form: One can see that ξopt is an eigenvector corresponding | iQR F = Ψ (V I)A U Ψ Ψ to the largest eigenvalue of ρdec obtained after optimal col ind h |QR ind ⊗ k col | iQR h | × QR k opt X individual operation defined by φi . We introduce matrix elements{| of Krausi} operators in Ucol† Ak† (Vind† I) Ψ QR = Ψ QR Ucol† Ak† (Vind† I) ⊗ | i h | ⊗ × the basis of φopt : Xk {| i i} Ψ QR Ψ (Vind I)AkUcol Ψ QR . (25) k opt opt | i h | ⊗ | i a := φ Ak φ , (31) ij h i | | j i One can see that the resulting expression is the same as opt the fidelity in the ‘individual-then-collective’ scheme up and also introduce the overlap between ξ and the opt | i to the following change (see Fig.4): vector set φ ζj ij: {| i i ⊗ | i} opt opt V U † ,U V † , , (26) C := φ ζ ξ . (32) col ↔ col ind ↔ ind E ↔ E ij h i | ⊗ h j| | i e 5
Then expression (30) takes the following form: D. ‘Both collective’ scheme
opt k k Finally, we consider a scheme, which is presented in Find col = λiλjCmi∗ amianj∗Cnj k,i,j,m,n Fig.2(d). Although this scheme may seem far from prac- X tical application, it is important from a theoretical point k k of view since it provides an upper bound on the fidelity = λ C∗ a λ C a ∗ i mi mi j nj nj (33) level achieved in the considered class of schemes. k i,m j,n X X X We can use any operator Ucol in order to ‘re-prepare’ 2 any desired state before acting of the decoherence chan- k = λ C∗ a . nel. In this way, the optimal strategy is to choose Ucol i mi mi k i,m such that the resulting pure state is affected by as less X X E as possible. Since acts individually on Q, the best fi- delity can be achievedE with U transforming Ψ in Let us then return to the ‘collective-then-individual’ col QR a separable state of the following form: | i scheme. As we demonstrated above, it is equivalent to ‘individual-then-collective’ with the only difference that Ucol Ψ = Υ Ξ , (39) the map is changed by the dual map . One can write | iQR | iQ ⊗ | iR the fidelityE of the ‘collective-then-individual’E scheme in with some normalized pure states Υ Q and Ξ R in 2- the form similar to Eq. (30): e n 1 | i | i dimensional and 2 − -dimensional Hilbert spaces corre- spondingly. Fcol ind = Since the second unitary Vcol can turn any pure state (34) into Ψ , the best final fidelity will be achieved with Υ λiλj ξ A† φi φj Ak ζi ζj ξ , | i | i b h |QR k | i h | ⊗ | iR h | | iQR minimizing a min-entropy of [ Υ Υ ]. Denote k,i,jX E | i h | e e e e opt Υ := arg min Smin( [ Υ Υ ]). (40) where ξ is some 2n-dimensional normalized vector and | i Υ E | i h | | i | i φ is an orthonormal basis in qubit space. We can set {| ii} We note that the minimum for min-entropy is the same e a minimum for linear and von Neumann entropies since opt e φi := φ , (35) is a qubit channel: | i | i i E opt e Υ = arg min Slin( [ Υ Υ ]) = arg min SvN( [ Υ Υ ]). and choose ξ such that | i Υ E | i h | Υ E | i h | | i | i | i (41) e opt φi ζj ξ = Cji∗ . (36) We then obtain the following expression for the maximal h | ⊗ h | | i fidelity as follows: Then we obtain the following expressione for the fidelity: opt opt opt F = Λmax( [ Υ Υ ]) col col E | i h | (42) k k = exp ln(2)S ( [ Υopt Υopt ]) . Fcol ind = λiλjCimaim∗ ajnCjn∗ min E | i h | k,i,j,m,n X This value is achieved under the action of unitary oper- b (37) ators of the following form: k opt = λjajnCjn∗ = Find col. opt k j,n Ucol = Υ Ξ Ψ + [...], (43) X X | i ⊗ | i h |
Vcol = Ψ Θ Ξ + [...], (44) On the one hand, we have the following expression: | i h | ⊗ h | where Θ stands for an eigenstate of [ Υopt Υopt ] cor- | i E | iopt h |opt opt opt responding to largest eigenvalue Λmax( [ Υ Υ ]) F Fcol ind = F . (38) E | i h | col ind ≥ ind col and [...] stands for the arbitrary appropriate remaining part of the unitary operators. On the other hand, by employingb a similar line reasoning opt opt In conclusion of the section, we note the following re- we can obtain Find col Fcol ind. Finally, we arrive the opt≥ opt lation between all the considered approaches: following relation: Fcol ind = Find col, i.e. there is a full equivalence of the performance of both schemes. The ob- opt opt opt opt Find ind Fcol ind = Find col Fcol col. (45) served identity reveals a time-symmetry aspect of trans- ≤ ≤ formation (26). That is why are all our subsequent con- It holds true since the scheme with two individual oper- siderations we will take into account the ‘individual-then- ations is a particular case for individual-then-collective collective’ scheme only since all the results for ‘collective- (or collective-then-individual) scheme, while the latter then-individual’ scheme can be obtained from it straight- schemes are particular cases for the scheme with two col- forwardly. lective operations. 6 24
(a1) (b1) (c1) opt opt opt opt opt opt opt opt opt Find ind Find col Fcol col Find ind Find col Fcol col Find ind Find col Fcol col
1 1 1 0.75 0.75 0.75 0.5 0.5 0.5 0.25 0.25 0.25 0 0 0 0.25 0.25 0.25 0 0.1 0.5 0 0.1 0.5 0 0.1 0.5 0.2 0.3 0.75 0.2 0.3 0.75 0.2 0.3 0.75 0.4 0.51 p 0.4 0.51 p 0.4 0.51 p Q Q Q Slin Slin Slin
(a2) (b2) (c2)
opt opt opt opt opt opt R Find col Find ind Find col Find ind F F ind col ind ind
0.1 Ucol 0.15 0.25 | i Q 0.2 0.1 V 0.15 0.05 ind E 0.05 0.1 0.05 0 0 0.25 0 0.25 0.25 0 0.5 0.1 0.2 0.5 0.1 0.2 0.5 0.1 0.2 0.3 0.4 0.75 0.3 0.4 0.75 0.3 0.4 0.75 0.51 = p 0.51 p 0.51 p Q SQ SQ S R lin lin lin Figure 3. The resultsFigure of 4. employing The results of the employing proposed the proposed protecting protecting schemes. schemes. In (a) In results (a)results for depolarizing for depolarizing channels, in (b) results channels, in (b) results for dephasing channels, are in (c) results for amplitude damping channels are presented. In (a1), (b1), and (c1) the maximal for dephasing channels, are in (c) results for amplitudeU † damping channels are presented. In (a1), (b1), and (c1) the maximal fidelities for all different| i strategies are shown. In (a2),col (b2), and (c2) the advantage of using ‘individual-then-collective’ over ‘both individual’ isfidelities demonstrated.Q for all di↵erent We strategies note are that shown. ‘individual-then-collective’ In (a2), (b2), and (c2) the advantage and of using ‘collective-then-individual’ ‘individual-then-collective’ over provides the same † results. Vind
R same basis has the following form:
√λ0 Ucol 0 Ψ = , (54) | i Q | iQR 0 Vind √λ1, E one can employ the operator as follows:
= Vcol = W (2γ), (55) R where U † cos(θ/2) 0 0 sin(θ/2) col 0 1 0 0 | i Q W (θ) := . (56) † 0 0 1 0 Vind
The state afterW the(✓) depolarizing channel in the X := 1 0 + 0 1 , X := ξ1 ξ0 + ξ0 ξ1 . (58) = | i h | | i h | | i h | | i h | ‘individual-then-collective’Q scheme with Uind := 1 is Q The rotation operation around y-axis of the Bloch sphere, given by Eq. (49). The optimal fidelityX in the consid-Ry(✓) X ered scheme equals to its largest eigenvalue and it is as cos(θ/2) sin(θ/2) R (θ) = , (59) follows: y sin(θ/2)− cos(θ/2) 1 1 1 in the central controlled-unitary gate is assumed to be F opt = p + (p 2)2 2p(4 3p)SQ . (51) ind col 2 − 4 4 − − − lin performed if the subsystem R is in the state ζ1 . q | i The behaviour the fidelity F opt is presented in ind col 3. ‘Both collective’ scheme Fig.3(a1), while the difference F opt F opt is given ind col − ind ind in Fig.3(a2). We note that both fidelities F opt and ind ind Since in the case of depolarizing channel the action of F opt decrease with increasing of SQ . The efficiency of ind col lin individual unitary operation cannot affect the resulting employing a single collective operator over two individual entropy, as it follows from the property Eq. (48), we can operators growth with an increase of p and it is maximal set the pre-processing unitary as follows: for p = 1 and SQ 0.375. lin ≈ The subtle question is how to obtain the optimal form Ucol := W (2ς), ς := arctan λ1/λ0, (60) of the post-processing operator V . This operator has col which turn the state Ψ into thep state 0 ξ . to turn the eigenstate ξ , which corresponds to the QR Q 0 R 0 QR Then after acting the depolarizing| i channel, we| obtain:i | i dec| i largest eigenvalue of ρQR to the initial state Ψ QR. For | i dec p p the state given by Eq. (49), we have the following rep- ρQR = (1 ) 0 Q 0 Q + 1 Q 1 Q ξ0 R ξ0 , resentation of ξ in the basis 0 ζ , 0 − 2 | i h | 2 | i h | ⊗ | i h | | 0iQR {| iQ ⊗ | 0iR | iQ ⊗ (61) ζ1 R , 1 Q ζ0 R , 1 Q ζ1 R : with the largest eigenvalue equal to 1 p/2 and corre- | i | i ⊗ | i | i ⊗ | i } − sponding eigenvector 0 Q ξ0 R. Thus, finally, we have cos υ the following expression:| i | i
0 opt p ξ0 QR = (52) F = 1 , (62) | i 0 col col − 2 sin υ It is achieved if the operators are as follows: with V := U † = W ( 2ς), (63) col col − 2Find col (2 p)λ0 where ς is given by Eq. (60). υ = arctan − − . (53) The unavoidable decrease of the fidelity can be ex- 2(1 p) SQ /2 − lin plained by an unavoidable entropy production effect in- q herent to the depolarizing channel. The behaviour of In order to transform ξ into Ψ , which in the F opt is also presented in Fig.3(a1). | iQR | iQR col col 8
B. Dephasing channel We demonstrate the behaviour of the fidelity Find col in Fig.3(b1) together with the difference F F ind col − ind ind The action of the dephasing channel on a qubit state in Fig.3(b2). One can see that the maximal difference ρ has the following form: in the case of the dephasing is in two times higher than the one for depolarizing channel and is achieved at the Q [ρ] = (1 p)ρ + p (ρ00 0 0 + ρ11 1 1 ) , (64) same point p = 1, S 0.375. E − | i h | | i h | lin ≈ where ρii := i ρ i are diagonal elements of ρ and p [0, 1] is againh | a| i strength of decoherence. One can 3. ‘Both collective’ scheme note∈ that the dephasing channel destroys non-diagonal elements of the density matrix ρ with probability p and One can see that the state 0 ξ0 is not affected by does not change the state with probability 1 p. In | iQ | iQ − dephasing channels, that is why we can employ the pre- contrast to the depolarizing channel, we see that the de- processing and post-processing collective operators in the phasing channel does not affect the diagonal states. This forms given by Eq. (60) and Eq. (63), correspondingly. feature can be efficiently employed in protection schemes. opt They allow achieving the maximal fidelity Fcol col = 1.
1. ‘Both individual’ scheme C. Amplitude damping channel
It is easy to check that the optimal fidelity for the ‘both We complete our consideration of the basic decoher- individual’ scheme is achieved for Uind = Vind = 1, since ence models with an amplitude damping channel. We these operations left the reduced state of Q in the ‘safe’ can write its action in terms of Kraus operators: diagonal form. For the identity pre-processing operator, we obtain the state after dephasing as follows: 1 0 0 √p A = ,A = . (69) 1 0 √1 p 2 0 0 ρdec = (1 p) Ψ Ψ +p λ i i ξ ξ . (65) − QR − | iQR h | i | iQ h |⊗| iR h | i Here p [0, 1] is again a decoherence strength. The am- X plitude∈ damping process corresponds to the relaxation of After applying the identity post-processing the value of the ‘excited’ state 1 to the ‘ground’ state 0 . We note fidelity takes the form: that there is an apparent| i asymmetry between| i the be- haviour of 0 and 1 : 0 remains untouched by the chan- F opt = (1 p) + p λ2 = 1 pSQ . (66) | i | i | i ind ind − i − lin nel, while 1 undergoes a transformation to 0 . Here we i | i | i X would like to remember that according to our convention we have λ λ in the initial state (46). It corresponds One can see that in the case of the depolarizing channel 0 ≥ 1 (50) the fidelity in dephasing channel decreases linearly to the fact that we assume that Q has the ‘ground’ level Q to be populated more than the ‘excited’ one. with a growth of p and Slin (see also Fig.3(b1)).
1. ‘Both individual’ scheme 2. ‘Individual-then-collective’ scheme
The optimal fidelity in the ‘both individual’ scheme is In the ‘individual-then-collective’ scheme the optimal achieved for the U and V being equal to identity op- individual operator is again the identity operator U := ind ind ind erators: U = V = 1. The state after the decoherence 1, which provides the state after decoherence given by ind ind takes the following form: Eq. (65). Its maximal eigenvalue determines the maximal
fidelity as follows: λ0 0 0 √λ0√λ1√p dec 0 pλ1 0 0 1 1 Q ρ = , (70) F = + 1 2p(2 p)S (67) QR 0 0 0 0 ind col 2 2 − − lin √λ0√λ1√p 0 0 pλ1 q and the corresponding eigenstate has the same form as in Eq. (52) with the only difference that where p := 1 p and the matrix is written in the basis 0 ζ ,−0 ζ , 1 ζ , 1 ζ . {| iQ ⊗ | 0iR | iQ ⊗ | 1iR | iQ ⊗ | 0iR | iQ ⊗ | 1iR} F opt λ The resulting fidelity takes the following form: υ := arctan ind col − 0 . (68) Q opt 2 2 2 (1 p) S /2 Find ind = λ0 + λ1 λ1p + 2λ0 1 pλ1. (71) − lin − − q p Thus, the optimal collective post-processing operator is By using the substitution, given by Eq. (55) with γ calculated using Eq. (57) with 1 1 λ = + 1 2SQ , (72) the updated value of υ from Eq. (68). 0 2 2 − lin q 9
R ‘Both individual’ ‘individual-then-collective’ & ‘collective-then- ‘Both
individual’ Ucol collective’ | i Q q Q 2 Q Depolarizing 1 − p(1 + S )/2 1/2 − p/4 + (p − 2) − 2p(4Vind− 3p)S /4 1 − p/2 lin E lin Q q Q Dephasing 1 − pSlin 1/2 + 1 − 2p(2 − p)Slin/2 1 √ Q Q q Q = Amplitude damping 1 − pSlin + (1 − p/2)(1 − Slin) + 1 − p 1 −R 1 − 2Slin /2 1 q Q p 1 − 2Slin/2 Ucol† | i Q † V
2. ‘Individual-then-collective’ scheme IV. DISCUSSION AND CONCLUSION
In the ‘individual-then-collective’ scheme the best fi- We summarize the main results of the present work. delity is also achieved for Uind = 1. The maximal fidelity We have suggested the method for improving fidelity is given by the largest eigenvalue of state (70): based on the class of state-dependent operations pro- 1 1 tecting the system from decoherence. Specifically, we F opt = 1 p(1 λ ) = 1 p 1 2SQ . ind col − − 0 − 2 − 2 − lin have considered the case where the decoherence affects q (74) a single κth qubit from an n-qubit pure state, while The corresponding eigenstate has the same form as in the operators can be employed either on the same κth Eq. (52) with qubit or on the whole state. We have shown that two schemes of this class, namely ‘individual-then-collective’ 2(1 p) and ‘collective-then-individual’ provide the same maxi- υ := arctan λ1 −Q . (75) mal achievable levels of the fidelity. s S ! lin We have considered these schemes for three basic de- The optimal collective post-processing operator is then coherence models given by depolarizing, dephasing, and given by the same expression (55) with γ calculated by amplitude damping channels. The main results on the Eq. (57) with updated value of υ given by Eq. (75). comparison between all the strategies of error suppres- We show the behaviour of Find col in Fig.3(c1). The sion for various decoherence models are summarized in corresponding difference F F is presented TableI. For all the channels we have seen that for given ind col − ind ind in Fig.3(c2). The maximal advantage of ‘individual- decoherence strength the maximal fidelities in all the then-collective’ (’collective-then-individual’) scheme over schemes except ‘both collective’ is expressed with the ‘both individual’ schemes in the case of the amplitude Q linear entropy of the qubit under decoherence Slin. In damping channel is two times higher than the one for de- particular, we have obtained that the larger is the linear phasing channel and is achieved for p = 1 and maximally entropy the lower is the fidelity. Q mixed reduced state of Q with Slin = 0.5. This feature provides an answer to the question which qubit from the whole n-qubit system is the most vulner- able and which qubit is the most robust in the sense of 3. ‘Both collective’ schemes preserving the whole n-qubit state after qubit decoher- ence. It turns out that the best choice for κ is defined As we have already emphasized above, the state 0 by the qubit whose reduced state has the lowest linear is not affected by the amplitude damping channel, that| i entropy. For the qubit case, the reduced state with the is why we can employ the pre-processing and post- lowest linear entropy is automatically the state with the processing collective operators in the form given by lowest min-entropy, von Neumann entropy, entropy of en- Eq. (60) and Eq. (63), correspondingly. As in the case tanglement and quantum discord. A promising direction of dephasing channel, they allow achieving the maximal for future research may be also to consider other various opt unit fidelity Fcol col = 1. measures of decoherence [42–44]. 10
The obtained results surely have a clear intuitive expla- schemes seem to be effective in the framework of quan- nation, since the more the entropy of the reduced state tum communication protocols, where several parties em- of the qubit is, the more it is entangled to remaining ploy an entangled state distributed between them. If the qubits of the system, and the more it is ‘informationally communication channels have a different degree of deco- important’ to the whole state. However, we would like herence, then the whole protocol can be adjusted in such to mention that one should be very careful with such a way that the most robust parts of the entangled state kind of intuitive conclusions. E.g. if we turn to speak go through the noisiest communication paths. about quantum correlations (say, entanglement), as it was shown in Refs. [24–26], in the case of depolarizing channel and mixed two-qubit state, the state with the ACKNOWLEDGMENTS largest entropy may be more robust in the sense of pre- serving initial correlations. The work was supported by the grant of the President Finally, we would like to mention that the considered of the Russian Federation (project MK-923.2019.2).
[1] H.J. Kimble, Quantum internet, Nature 453, 1023 [16] E. Knill, R. Laflamme, and L. Viola, Theory of quantum (2008). error correction for general noise, Phys. Rev. Lett. 84, [2] T.D. Ladd, F. Jelezko, R. Laflamme, Y. Nakamura, C. 2525 (2000). Monroe, and J.L. O’Brien, Quantum computers, Nature [17] D. Gottesman, An introduction to quantum error correc- (London) 464, 45 (2010). tion and fault-tolerant quantum computation, in Proceed- [3] C.R. Monroe, R.J. Schoelkopf, and M.D. Lukin, Quan- ings of Symposia in Applied Mathematics 68, pp. 13-58 tum connections, Sci. Am. 314, 50 (2016). (American Mathematical Society, Providence, RI, 2010). [4] S. Muralidharan, L. Li, J. Kim, N. L¨utkenhaus, M.D. [18] L. Jiang, J.M. Taylor, K. Nemoto, W.J. Munro, R. Van Lukin, and L. Jiang, Optimal architectures for long Meter, and M.D. Lukin, Quantum Repeater with encod- distance quantum communication, Sci. Rep. 6, 20463 ing, Phys. Rev. A 79, 032325 (2009). (2016). [19] W.J. Munro, K.A. Harrison, A.M. Stephens, S.J. Devitt, [5] M. Mohseni, P. Read, H. Neven, S. Boixo, V. Denchev, and K. Nemoto, From quantum multiplexing to high- R. Babbush, A. Fowler, V. Smelyanskiy, and J. Marti- performance quantum networking. Nat. Photonics 4, 792 nis, Commercialize quantum technologies in five years, (2010). Nature (London) 543, 171 (2017). [20] A.G. Fowler, D.S. Wang, C.D. Hill, T.D. Ladd, R. Van [6] S. Wehner, D. Elkouss, and R. Hanson, Quantum inter- Meter, and L.C.L. Hollenberg, Surface code quantum net: A vision for the road ahead, Science 362, eaam9288 communication, Phys. Rev. Lett. 104, 180503 (2010). (2018). [21] S. Muralidharan, J. Kim, N. L¨utkenhaus, M.D. Lukin, [7] W.H. Zurek, Decoherence, einselection, and the quantum and L. Jiang, Ultrafast and fault-tolerant quantum com- origins of the classical, Rev. Mod. Phys. 75, 715 (2003). munication across long distances, Phys. Rev. Lett. 112, [8] G.M. Palma, K.-A. Suominen, and A.K. Ekert, Quantum 250501 (2014). computers and dissipation, Proc. Roy. Soc. London Ser. [22] S. Muralidharan, C.L. Zou, L. Li, J. Wen, and L. Jiang, A 452, 567 (1996). Overcoming erasure errors with multilevel systems, New [9] L.-M. Duan and G.-C. Guo, Reducing decoherence in J. Phys. 19, 013026 (2017). quantum-computer memory with all quantum bits cou- [23] S.N. Filippov and M.Ziman, Entanglement sensitivity to pling to the same environment, Phys. Rev. A 57, 737 signal attenuation and amplification, Phys. Rev. A 90, (1998). 010301(R) (2014). [10] D.A. Lidar, I.L. Chuang, and K.B. Whaley, Decoherence- [24] K. Zyczkowski, P. Horodecki, M. Horodecki, and R. free subspaces for quantum computation, Phys. Rev. Horodecki, Dynamics of quantum entanglement, Phys. Lett. 81, 2594 (1998). Rev. A 65, 012101 (2001). [11] A. Buchleitner, V. Viviescas, and M. Tiersch, Entangle- [25] E.O. Kiktenko and S.M. Korotaev, Causal analysis of ment and decoherence: Foundations and modern trends asymmetric entangled states under decoherence, Phys. (Springer-Verlag Berlin Heidelberg, 2009). Lett. A 376, 820 (2012) [12] H.P. Breuer and F. Petruccione, Theory of open quantum [26] S.M. Korotaev and E.O. Kiktenko, Causality and deco- systems (Oxford University Press, Oxford, 2002). herence in the asymmetric states, Phys. Scr. 85, 055006 [13] A.Y. Kitaev, Quantum error correction with imperfect (2012). gates, in Proceedings of the 3rd International Conference [27] E.O. Kiktenko and A.K. Fedorov, Tomographic causal of Quantum Communication and Measurement (New analysis of two-qubit states and tomographic discord, York, Plenum, 1997). Phys. Lett. A 378, 1704 (2014). [14] C.H. Bennett, G. Brassard, S. Popescu, B. Schumacher, [28] S.A. Gurvitz, L. Fedichkin, D. Mozyrsky, and G.P. J.A. Smolin, and W.K. Wootters, Purification of noisy Berman, Relaxation and the Zeno effect in qubit mea- entanglement and faithful teleportation via noisy chan- surements, Phys. Rev. Lett. 91, 066801 (2003). nels, Phys. Rev. Lett. 76, 722 (1996). [29] P. Facchi, D.A. Lidar, and S. Pascazio, Unification of dy- [15] A.M. Steane, Error correcting codes in quantum theory, namical decoupling and the quantum Zeno effect, Phys. Phys. Rev. Lett. 77, 793 (1996). Rev. A 69, 032314 (2004). 11
[30] S. Maniscalco, F. Francica, R.L. Zaffino, N. Lo Gullo, and tum Computers and Computing 1, 84 (2000). F. Plastina, Protecting entanglement via the quantum [37] Y. Ozhigov and L. Fedichkin, A quantum computer with Zeno effect, Phys. Rev. Lett. 100, 090503 (2008). fixed interaction is universal, JETP Lett. 77, 328 (2003). [31] A.N. Korotkov and K. Keane, Decoherence suppression [38] H. Ollivier and W. Zurek, Quantum discord: A measure by quantum measurement reversal, Phys. Rev. A 81, of the quantumness of correlations, Phys. Rev. Lett. 88, 040103(R) (2010). 017901 (2001). [32] Y.-S. Kim, J.-C. Lee, O. Kwon, and Y.-H. Kim, Protect- [39] L. Henderson and V. Vedral, Classical, quantum and to- ing entanglement from decoherence using weak measure- tal correlations, J. Phys. A 34, 6899 (2001). ment and quantum measurement reversal, Nat. Phys. 8, [40] A.K. Fedorov, E.O. Kiktenko, O.V. Manko, and V.I. 117 (2012). Manko. Tomographic causal analysis of two-qubit states [33] K. Roszak, R. Filip, and T. Novotn`y,Decoherence con- and tomographic discord, Phys. Scr. 90, 055101 (2015). trol by quantum decoherence itself, Sci. Rep. 5, 9796 [41] M.A. Nielsen and I.L. Chuang, Quantum computation (2015). and quantum information (Cambridge University Press, [34] A. Basit, F. Badshah, H. Ali, and G.-Q. Ge, Protecting Cambridge, 2000). quantum coherence and discord from decoherence of de- [42] L. Fedichkin, A. Fedorov, and V. Privman, Measures of polarizing noise via weak measurement and measurement decoherence, Proc. SPIE 5105, 243 (2003). reversal, Europhys. Lett. 118, 30002 (2017). [43] L. Fedichkin, A. Fedorov, and V. Privman, Additivity of [35] A.E. Ulanov, I.A. Fedorov, A.A. Pushkina, Y.V. decoherence measures for multiqubit quantum systems, Kurochkin, T.C. Ralph, and A.I. Lvovsky, Undoing the Phys. Lett. A 328, 87 (2004). effect of loss on quantum entanglement, Nat. Photonics [44] L. Fedichkin and V. Privman, Quantitative treatment of 9, 764 (2015). decoherence, Topics Appl. Physics 115, 141 (2009). [36] L. Fedichkin, Polynomial procedure of avoiding multi- qubit errors arising due to qubit-qubit Interaction, Quan-