Arxiv:1908.03404V2 [Quant-Ph] 12 May 2020

Probability representation of quantum dynamics using pseudostochastic maps

E.O. Kiktenko,1, 2, 3 A.O. Malyshev,1, 3 A.S. Mastiukova,1, 3 V.I. Man’ko,1, 3, 4 A.K. Fedorov,1, 3 and D. Chru´sciński5 1Russian Quantum Center, Skolkovo, Moscow 143025, Russia 2Department of Mathematical Methods for Quantum Technologies, Steklov Mathematical Institute of Russian Academy of Sciences, Moscow 119991, Russia 3Moscow Institute of Physics and Technology, Dolgoprudny 141700, Russia 4P.N. Lebedev Physical Institute, Russian Academy of Sciences, Moscow 119991, Russia 5Institute of Physics, Faculty of Physics, Astronomy and Informatics, Nicolaus Copernicus University, Toruń87100, Poland (Dated: May 13, 2020) In this work, we consider a probability representation of quantum dynamics for finite-dimensional quantum systems with the use of pseudostochastic maps acting on true probability distributions. These probability distributions are obtained via symmetric informationally complete positive operator-valued measure (SIC-POVM) and can be directly accessible in an experiment. We provide SIC-POVM probability representations both for unitary evolution of the density matrix governed by the von Neumann equation and dissipative evolution governed by Markovian master equation. In particular, we discuss whereas the quantum dynamics can be simulated via classical random processes in terms of the conditions for the master equation generator in the SIC-POVM probability representation. We construct practical measures of nonclassicality non-Markovianity of quantum processes and apply them for studying experimental realization of quantum circuits realized with the IBM cloud quantum processor.

I. INTRODUCTION Classical probability Pseudo-stochastic maps Classical probability distributions distributions

Stochastic Quantum technologies require an efficient toolbox for SIC-POVM maps SIC-POVM probability probability synthesis, control, and characterization of the quantum distributions distributions states and processes [1]. The task of characterizing quantum states has quite a rich history of attempts to describe quantum systems with the use of standard methods of statistical physics, such as phase-space probability distributions [2–5]. A quantum analog of classical phase-space probability distributions, known as the Wigner function, Figure 1. General relation among the set of SIC-POVM cannot be fully interpreted as a probability distribution probability distributions, general classical probability distri- because it takes negative values in some cases [2,6,7]. butions of the same dimension, and maps which keep distri- The negativity of the Wigner quasiprobability distribu- butions in a particular set. The set of SIC-POVM probability distributions, so-called Hilbert qplex [38], is smaller than the tion plays an important role in the modern quantum full probability distribution set, while the set of possible maps theory since this effect is a signature of the highly non- turning SIC-POVM probabilities into SIC-POVM probabili- classical character of a quantum state. In particular, it ties (known to be a pseudostochastic) is wider than the set was demonstrated that negativity and contextuality are of classical stochastic maps. Points and arrows demonstrate equivalent notions of nonclassicality [8]. The negativity actions of pseudostochastic (stochastic) maps on SIC-POVM of the Wigner function has been largely studied for quan- (classical) probability distributions. tum information processing both for systems with contin- uous [6] and discrete variables [9–14]. Recent progress in quantum information science has helped us to understand arXiv:1908.03404v2 [quant-ph] 12 May 2020 ing quantum states contain no redundant information; the role of properties of quasiprobability distribution in i.e., the number of probabilities is the minimum possi- the context of verifying quantum resources that provide ble for reconstructing all density-matrix elements. We quantum speedup [15–21]. note that analytic proofs of SIC existence have only been An approach to the description of quantum phenom- found in a number of cases [29]. The approach of describ- ena using the language of probability distributions [22– ing quantum states with SIC-POVM probability distribu- 25] has been extended by the concept of informationally tions has been widely explored in quantum Bayesianism complete POVMs (IC-POVMs) and SIC-POVMs that (QBism) reformulation of quantum mechanics [30, 31]. are based on measurements completely describing quan- SIC-POVM measurements also has been actively used tum states [26–28]. In this framework, quantum states in various experiments [32–37]. Importantly, it turns out are associated with probabilities related to a specific set that the set of possible probability distributions obtained of vectors in the Hilbert space, which is composed of so- with SIC-POVM measurements is smaller than the full called SIC projectors. It is important to point out that in probability distributions set of the same dimension. This the SIC approach, the probability distributions describ- “quantum” part of a classical probability simplex, which 2 is achievable via SIC-POVM measurements, is referred ply our theoretical results to experimental study of pro- to as a “qplex” [38]. An important result of Ref. [38] is cesses realized on the super-conducting quantum cloud the derivation of the properties of qplexes from the very IBM QX4 quantum processor [55]. fundamental assumptions about quantum theory as well Our work is organized as follows. In Sec.II, we present as the description of a link between the symmetry prop- a general scheme for the SIC-POVM probability repre- erties of qplexes and a condition for the existence of a sentation of states and measurements. In Sec.III, we d-dimensional SIC-POVM. derive an equation for a SIC-POVM probability vector The representation of quantum states with the use of which corresponds to the von Neumann equation and the quasiprobability distributions can be further generalized Markovian master equation. In Sec.IV, we study a re- to the representation of quantum processes (channels) lation between pseudostochasticity and non-classicality, with the use of quasistochastic matrices [39–42]. In con- and construct practical measures of nonclassicality and trast to traditionally used stochastic matrices, which de- non-Markovianity of quantum processes. In Sec.V, we scribe the evolution of classical probability distributions, apply our theoretical result to experimental study of the quasi-stochastic matrices can posses negative elements. IBM Q4 superconducting quantum processor. We sum- In Ref. [40], a functorial embedding of the quantum chan- marize main results and conclude in Sec.VI. nels category into the category of quasi-stochastic matrices has been provided. Thus, the formalism of quasi- stochastic matrices can serve as an alternative formula- II. PROBABILITY REPRESENTATION OF tion of the quantum theory and looks promising in the STATES, MEASUREMENTS AND LINEAR MAPS framework of quantum resources analysis. This appara- tus, however, has not been consistently applied to quan- We start our consideration by introducing SIC-POVM tum information processing tasks yet. effects, which can be used in the construction of probabil- In this work, we focus on the dynamics of probability ity representation of states and measurements for finite- distributions obtained in SIC-POVM measurements. In dimensional systems. Consider a d-dimensional Hilbert line with Refs. [41, 42], we refer to the resulting matrices space with d 2. In what follows, we assume that H ≥ 2 (maps), which define the evolution of SIC-POVM proba- it is possible to find out a set of d normalized states d2 bility distributions as pseudostochastic rather than qua- ψi i=1 belonging to such that {| i} H sistochastic. This is because these matrices correspond to the evolution of true experimentally accessible probabil- 2 dδi,j + 1 ψi ψj = Tr(ΠiΠj) = , (1) ities rather then quasi-probabilities. As was mentioned, |h | i| d + 1 pseudostochastic matrices are an analog of conditional probability matrices without restrictions of positivity for where Πi := ψi ψi and δi,j stands for the Kronecker | ih | d2 symbol. We note that the set Πi forms a basis in matrix elements (see Fig.1). In fact, pseudostochastic { }i=1 matrices appear in the discussion of unitary evolution of the space ( ) of linear operators acting on . L H d2 H quantum states described by qplexes (see Appendix B of The set Πi/d i=1 is called a SIC-POVM. By its def- { }1 P Ref. [38]) and are discussed in detail in Ref. [40]. How- inition, we have d Πi 0 and i Πi = 1d, where 1 de- ever, a connection between pseudo(quasi)stochastic ma- notes identity operator.≥ At this moment, SIC-POVMs trices and common equations used for describing quan- are found for d = 2–151, 168, 172, 195, 199, 228, 259, tum system dynamics, to the best of our knowledge, is 323, and 844. The obtained solutions are available on- not considered in the literature. It is then important line [59] (for a review, see also Ref. [29] and reference to investigate how pseudostochastic matrices appear as therein). We also note that numerical methods are heav- solutions of quantum dynamical equation, and we ad- ily involved for SIC-POVM search [60, 61]. dress this question in this work. More precisely, we derive a dynamics equation for a SIC-POVM probability vector, which corresponds to the von Neumann equation A. Representation of states and dissipative evolution governed by Markovian master equation. Next we consider a general scheme of quan- Let us consider a quantum state given by a unit- tum mechanical experiments and demonstrate that the trace semipositive Hermitian density operator ρ ( ) developed SIC-POVM representation shines additional (ρ 0, Trρ = 1). The probability of obtaining an ∈ith L out-H light on nonclassical features of quantum process. We ≥ come corresponding to the effect Πi/d after SIC-POVM consider necessary and sufficient conditions for the time- measurement is given by independent Markovian generator to produce classical- like quantum evolution. These conditions allow us to 1 p = Tr(ρΠ ). (2) construct practical measure for nonclassicality of quan- i d i tum processes. We show that the considered SIC-POVM probability representation also allows introducing new Let us write these probabilities in the form of vector p := T practical measure of non-Markovianity, which has been p1 . . . pd2 , which we further refer to as a SIC-POVM actively studied in recent decades [43–54]. Finally, we ap- probability vector. 3

2 The density matrix ρ can be reconstructed back from Pd where p, s = i=1 pisi is a standard dot product of two the SIC-POVM probability vector in the following way: SIC-POVMh i probability vectors. Since Tr(ρσ) [0, 1], we also have ∈ d2 X 1 1 2 ρ = (d + 1)pi Πi p, s , (10) − d d(d + 1) ≤ h i ≤ d(d + 1) i=1 d2 where the minimum is achieved for two orthogonal states X X ρ and σ, while the maximum is achieved for s = p and a = [(d + 1)Πi 1] pi = Kipi, (3) − pure state ρ = σ = ψ ψ . i=1 i As was already mentioned| ih | in the introduction, the set where Ki = (d + 1)Πi 1. of possible SIC-POVM probability vectors is smaller than − 2 We mention the following useful property of Ki: the full set of all possible d -dimensional probability vectors. This fact can be easily verified by noticing that Tr(KiΠi) = (d + 1)Tr(ΠiΠj) Tr(Πj) = dδi,j. (4) − the maximum probability in the SIC-POVM probability vector can not exceed the value d−1 due to the struc- It is also useful to introduce a vectorized representation ture of SIC-POVM effects. We refer readers to Ref. [38], of linear operators. Let i be an orthonormal basis where properties of the SIC-POVM probability vectors in and A ( ) is some{| i} linear operator. We can H P ∈ L H set (qplex) are studied in detail. write A = Ai,j i j , where Ai,j = i A j are matrix i,j | ih | h | | i In AppendixA, we provide a relation between the con- elements of A in the i representation. Next, we refer P {| i} sidered SIC-POVM probability representation and an al- to A := Ai,j i j , as the “ket” vector | ii i,j | i ⊗ | i ∈ H ⊗ H ternative probability representation based on mutually representation of A. In a similar way, we can introduce unbiased measurements (MUB) in the case of d = 2. an adjoint vector A := P A∗ i j ∗ ∗. It hh | i,j i,jh | ⊗ h | ∈ H ⊗ H is easy to check that for any two operators A, B ( ), ∈† L H their Hilbert-Schmidt product takes the form Tr(A B) = B. Representation of measurements A B . hh Let|| A,ii B, U, V ( ). It is also easy to check that the ∈† L H Let us now consider a question how the SIC-POVM identity B = UAV corresponds to the following identi- probability vectors determine the probabilities for ar- ties in the vector representation: bitrary measurements. Consider a POVM E = P B = U V ∗ A , B = A U † V T. (5) E1,...,Em with Ei 0 and i Ei = 1. Let us in- { } ≥ T | ii ⊗ | ii hh | hh | ⊗ troduce an m-dimensional vector q := q1 . . . qm with Using the introduced vectorized representation, one elements given by probabilities of obtaining different out- can rewrite Eq. (3) in the following form: comes in measuring E for some state ρ: qi = Tr(Eiρ). The relation between the SIC-POVM probability vec- d2 X tor p (corresponding to ρ) and a new probability vector ρ = Ki pi = Kp, (6) | ii | ii q for an arbitrary measurement can be obtained using i=1 Eq. (3): where q = Mp, (11) 2 where K = K1 ... K | ii | d i   2 TrE1 = (d + 1) Π1 ... Πd 1 ... 1 (7) | ii | ii − | ii | ii  .  1 ... 1 M = (d + 1)m  .  (12) is a d2 d2 matrix. − | {z } TrE 2 Thus,× the linear transformation K defines the map m d elements 2 from the SIC-POVM probability vector p to the vec- is m d matrix with elements given by mi,j = Tr(EiΠj). torized representation of the density matrix ρ. Using The× matrix m is a stochastic rectangular matrix, that is, −1 P Eq. (2), we obtain the inverse matrix K given by it satisfies the following properties: mi,j 0, mi,j = P ≥ i Tr ( i EiΠj) = 1. We note that m appears to be bis- −1 1 T K = Π1 ... Πd2 . (8) tochastic; that is, its rows also sum up to 1, in the case 2 −1 2 d hh | hh | where m = d and TrEi = d for any i 1, . . . , d . ∈ { } It may be the case when Ei is also a SIC-POVM. We also note the following correspondence between { } the Hilbert-Schmidt product of states and dot product It easy to check from Eq. (12) that the matrix M is of SIC-POVM probability vectors. Let ρ and σ be two pseudostochastic, that is, the sum of its elements in each P column equals to unity: Mi,j = (d + 1) Tr1 = 1; arbitrary density matrices, and p and s be their corre- i − sponding SIC-POVM probability vectors. One can check however, some elements Mi,j may be negative. It may be that the case when Ei is proportional to the projector on the state orthogonal to some ψj . Then we obtain Mi,j = | i Tr(ρσ) = ρ σ = d(d + 1) p, s 1, (9) TrEj < 0. hh || ii h i − − 4

C. Representation of linear maps as We can also deﬁne Φ using the Choi state pseudostochastic matrices 1 X ρΦ = i j Φ[ i j ]. (18) d | ih | ⊗ | ih | Here we consider a representation of positive in i,j trace-preserving (PTP) and completely positive trace- preserving (CPTP) maps acting on quantum states in The vectorized version of the Choi state then reads 1 P the SIC-POVM framework. R = i,j Φ[ i j ] i j . In this form, the output din | | ih | iih | ⊗ h | First, consider a PTP linear map Φ : ( ) state can be calculated in the following way: ρout = in | ii ( ), where and are d -L andH d →- dinR ρin . By comparing this result with Eq. (17), we out in out in out | ii LdimensionalH HilbertH spaces, respectively.H obtain Being a PTP map Φ transforms density operators in 1 1 R = A = K SK−1. (19) in into density operators in out, so we can consider an d Φ d out in actionH of Φ on some input stateH ρin resulting in output in in 2 2 out in in din out dout −1 state ρ = Φ[ρ ]. Let Πi i=1 and Πi i=1 be SIC- and S = dinKoutRKin. { } { } in POVM projectors in in and out, respectively. Let p It is important to note that Φ is CPTP map if and only out H H and p be SIC-POVM probability vectors correspond- if ρΦ 0 and TrHin ρΦ = 1/din. These requirements allow ing to ρin and ρout. Simple algebra leads to pout = Spin, one to≥ check whether a given pseudostochastic matrix S where corresponds to any CPTP map.

1 out Si,j= (din + 1)si,j Tr Πi Φ(1in) , (13) − dout III. DYNAMICS OF SIC-POVM PROBABILITY 1 out in VECTORS si,j = Tr Πi Φ(Πj ) , (14) dout Here we study the dynamics of the SIC-POVM proba- and 1in is the identity operator acting in in. One can P H bility vector p. For this purpose, we analyze (i) the rep- see that s is a stochastic matrix: si,j = 1, si,j 0, P i ≥ resentation for dynamical equation for p corresponding while Si,j is pseudostochastic one: i Si,j = 1 with some to the unitary evolution of the density matrix governed elements may be negative. by the von Neumann equation, (ii) the representation for If Φ is also unital, that is, Φ(1in) = 1out, then s is unitary operators as the solution of dynamical equations, bistochastic and Eq. (13) reduces to Si,j = (din +1)si,j and (iii) the representation for dissipative evolution for p −1 − dout . The examples of PTP maps for the qubit case corresponding to Markovian master equation. We place (d = 2) are presented in AppendixB. the summary of results of each following subsection in Let now Φ be a CPTP map that is representing a quan- TableI. tum channel. Any such map gives rise to a Kraus representation A. Stochastic representation of the von Neumann out in X in † X † ρ = Φ[ρ ] = Akρ Ak, AkAk = 1in. (15) equation k k Consider a standard von Neumann evolution equation In this case Eq. (13) may be rewritten as follows: i ρ˙ = [H, ρ], (20) " # ~ 1 X † out 2 Si,j = (din + 1)si,j Tr Πi AkAk (16) where [ , ] stands for commutator, i = 1,ρ ˙ is the time − dout k derivative· · of the state ρ, H = H† is the− system Hamilto- nian, and ~ is the Plank constant. In the general case, the −1 P h in † outi where si,j = dout k Tr AkΠi AkΠj . Again, s is a Hamiltonian H may depend on time t. In what follows, stochastic matrix, while S is pseudostochastic. we use dimensionless units and set ~ := 1. Let us observe that we can deﬁne S in another way. To make a transition from the density matrix ρ to the Let us rewrite Eq. (15) in the vectorized form: corresponding SIC-POVM probability vector p, we mul- tiply both sides of Eq. (20) by Πi and take the trace. X ∗ Thus, we obtain ρout = Ak A ρin = A ρin , (17) | ii ⊗ k| ii | ii k iTr(ρ ˙Πi) = Tr([H, ρ]Πi). (21) P ∗ where A = k Ak Ak. Consequently, we obtain Taking into account Eq. (4), the left-hand side of Eq. (21) out in ⊗ Koutp = AKinp , where Kout and Kin are deﬁned can be rewritten as follows: via Eq. (7), which is related to the corresponding SIC- X POVM elements. Finally, we have the following expres- iTr(ρ ˙Πi) = i Tr(KjΠi)p ˙j = idp˙i. (22) −1 sion for the pseudostochastic matrix S: S = KoutAKin. j 5

d2 Standard representation Probability representation for SIC-POVM {Πi/d}i=1 T d-dimensional quantum state ρ – d × d Hermitian unit-trace semi- p = p1 . . . pd2 is the real probability vector with non- positive complex matrix negative elements Pd2 Trρ = 1, ρ ≥ 0. i=1 pi = 1, pi ≥ 0. Measurement results proba- qi = Tr(ρEi) q = Mp, where M = (d + 1)m − E, Ei,j = TrEi, bilities for a POVM {Ei} mi,j = Tr(EiΠj )(M is pseudostochastic matrix). −1 Evolution equation deﬁned iρ ˙ = [H, ρ] p˙ = Hp, where Hi,j = (d + 1)d Tr(H[Πj , Πi]) by a Hamiltonian H (H is Hermitian) (H is real antisymmetric). Solution of the evolution ρ = U(t)ρinU †(t), where p = U(t)pin, where equation n R t 0 0o n R t 0 0o U(t) = T exp −i t0=0 H(t )dt , U(t) = T exp t0=0 H(t )dt (U(t) is unitary). (U(t) is unitary and pseudobistochastic). Markovian master equation ρ˙ = −i[H, ρ]+ p˙ = Lp, where −1 ∗ P ∗ governed by the GKSL P † 1 † † L = K ΛK, Λ = −i(C ⊗ 1 − 1 ⊗ C ) + Vk ⊗ V , + VkρV − [V Vkρ + ρV Vk] k k generator k k 2 k k i P † C = H − 2 k Vk Vk. out P in † out in −1 h out P † i Quantum channel deﬁned by ρ = i Aiρ Ai p = Sp , where S = (din + 1)s − doutTr Πi k AkAk , Kraus operators {Ai} −1 P in † out si,j = dout k Tr AkΠi AkΠj (S is pseudostochastic).

Table I. Correspondence between standard and SIC-POVM probability representations.

The right-hand side of Eq. (21) can be written in the 3. Each row and column of H summing to 0: P P following way: i Hi,j = 0, j Hi,j = 0 (here we employed the P fact that ψi ψi = 1d). X i | ih | Tr ([H, ρ]Πi) = Tr ([H, ((d + 1)Πj 1)pj]Πi) − j The number of independent parameters defining the X d2 d2 matrix with such properties is equal to N = = p (d + 1)Tr([H, Π ]Π ) H j j i (d2× 1)(d2 2)/2. Meanwhile, the physical properties of j − − 2 the Hamiltonian is defined with NH = d 1 parameters X − = pj(d + 1)Tr(H[Πj, Πi]). (23) (the term 1 comes from the fact that the Hamiltonian is − j defined up to a term proportional to the identity matrix). One can see that NH > NH for d > 2, so there should be By combining Eqs. (22) and (23), we obtain some additional constraints on H on the top of the listed p˙ = Hp, (24) properties. In order to study these constraints, we introduce a set 2 2 2 where H is a d d matrix with elements given by Hi,j = σ(j) d −1 of orthogonal (with respect to the Hilbert- d+1 × { }j=1 id Tr(H[Πj, Πi]). One can see that the elements of H Schmidt distance) traceless Hermitian matrices in ( ) can be also rewritten as (e.g., Pauli matrices for d = 2) and a renormalizedL H 2 2(d + 1) identity matrix σ(d ) := p2/d1 satisfying the relation Hi,j = Im [ ψj ψi ψi H ψj ] , (25) (i) (j) 2 d h | ih | | i Tr(σ σ ) = 2δi,j for any i, j 1, . . . , d . ∈ { } where Im [ ] stands for the imaginary part. Then the Hamiltonian can be written in the form Let us also· consider an alternative approach to ob- d2 taining the form of H. We can write the von Neu- X H = λ σ(i), (27) mann equation (20) in the vectorized form: i ρ˙ = i (H 1 1 H∗) ρ . Then by taking into account Eq.| ii (6), i=1 ⊗ − ⊗ | ii ∗ we have iKp˙ = (H 1 1 H )Kp. Comparing this re- 2 (i) d −1 ⊗ − ⊗ where λi = Tr Hσ /2. We note that all λi can sult with Eq. (24), we obtain the following representation { }i=0 of the Hamiltonian: take arbitrary real values. Substituting representation (27) into Eq. (26), we ob- H = iK−1(H 1 1 H∗)K. (26) tain − ⊗ − ⊗ The matrix H in the form given by Eq. (25) has a d2−1 X −1 (i) (i)∗ number of important properties. H = λi iK σ 1 1 σ K − ⊗ − ⊗ i=0 1. H is real and antisymmetric: Hi,j = Hji. − d2−1 2. Diagonal elements and trace of H are zero: H = 0, X (i) ii = λiH , (28) TrH = 0. i=1 6

(i) −1 (i) (i)∗ 2 2 where H = iK σ 1 1 σ K. We see where eI is the d d matrix with all elements equal to − ⊗ − ⊗ × that the parameter λ0 does not participate in deﬁning unity (i.e., Ii,j = 1). H. We can highlight the following basic properties of the (i) d2 It turns out that the matrices from the set H i=1 operator U(t). are orthogonal to each other: { } 1. U(t) is pseudobistochastic: Tr H(i)H(j)T = Tr H(i)H(j) X X − Ui,j(t) = Ui,j(t) = 1, (37) (i) (j) (i)∗ (j)∗ i j = Tr σ σ 1 + 1 σ σ = 4dδi,j. (29) ⊗ ⊗ and some its elements of may be negative. Thus, the set H(i) forms a basis of the (d2 1)- dimensional linear{ subspace} of real antisymmetric− ma- 2. U(t) is orthogonal: trices in the space of all antisymmetric matrices corre- T T sponding to physical processes. U(t)U (t) = U (t)U(t) = I, (38) We can also introduce a projector Ham on this sub- 2 2 P where I is a d d identity matrix. space, which acts on arbitrary real d2 d2 matrix He as × follows: × We note that according to these two properties an action

d−1 of U(t) preserves both l1 and l2 norms. 1 X (i)T (i) Finally, the generalization on time-dependent Hamil- unit(He ) = Tr(H He )H . (30) P 4d tonians can be provided in a straightforward way. In this i=1 case of t > 0, the evolution operator is given by Therefore, a matrix H corresponding to a physical Hamil- tonian satisﬁes the relation Z t U(t) = T exp H(t0)dt0 , (39) t0=0 unit(H) = H. (31) P where T is the standard time-ordering operator. It is easy This is an important relation for the representation of to see that the mentioned properties for U(t) remain true the SIC-POVM probability vector dynamics. in this case. We present an example for the construction of H and U(t) in AppendixC. B. Representation of unitary operators

If the matrix H does not depend on t (the Hamiltonian C. Stochastic representation of dissipative H does not depend on time t), then the solution of the evolution basic dynamics equation for the SIC-POVM probability vector Eq. (24) is as follows: Consider now the Markovian master equation in p(t) = U(t)p , U(t) = exp (Ht) , (32) ρ˙ = L(ρ), (40) where pin is a SIC-POVM probability vector at t = 0. which is governed by the Gorini-Kossakowski-Sudarshan- At the same time, we know that the solution for ρ via Lindblad (GKSL) generator [56, 57] of the following form: standard evolution operator is given by: in † X † 1 † † ρ(t) = U(t)ρ U (t),U(t) = exp ( iHt) , (33) L(ρ) = i[H, ρ] + VkρV [V Vkρ + ρV Vk] , − − k − 2 k k in k where ρin is initial density matrix (corresponded to p ). (41) Then we can obtain where H is Hamiltonian and Vk are so-called noise opera- 1 tors. By fixing an orthonormal basis i in , one defines X in | i H pi(t) = (d + 1)ui,j(t) pj , (34) the following matrix − d j Ki,j := Tr [PiL(Pj)] , (42) 1 † where ui,j(t) = d Tr[U(t)ΠjU (t)Πi]. It easy to see that 2 2 u is a d d bistochastic matrix: where Pi = i i . The matrix Ki,j defines a Kol- × | ih | X X mogorov generator [63] as follows: Ki,j 0, i = j, ui,j(t) = ui,j(t) = 1, ui,j(t) 0. (35) and P K = 0, j = 1, . . . , d. Let us then≥ define6 the ≥ i i,j i j stochastic representation of Eq. (40) using SIC-POMV. Equation (40) gives rise to the following equation for the Finally, we arrive at the following expression: probability vector p: 1 U(t) = (d + 1)u(t) eI, (36) − d p˙ = Lp, (43) 7 where the d2 d2 matrix L is defined as follows: reduced down to d2 rank(v), we prefer to treat V as × d2 d2 matrix with possibly× zero columns. 1 1 × Li,j := (d + 1) Tr [ΠiL(Πj)] Tr [ΠiL(1)] . (44) Using the matrix V as a parametrization of the Marko- d − d vian noise generator D, we define a function d2 P X † The matrix Li,j satisfies the condition i=1 Li,j = 0, D(V) := [VV ] Ω . (51) 2 i,j i,j j = 1, . . . , d ; however, condition Li,j 0 for i = j needs ≥ 6 i,j not be satisfied. We may call Li,j a pseudo-Kolmogorov generator. It allows us to introduce a function (projector) which Finally, defining the vectorization of L outputs a matrix of physical time-independent Marko- vian generator which is closest, in terms of the Frobenius ∗ X ∗ 2 2 Λ := i(C 1 1 C ) + Vk Vk , (45) norm, to given d d matrix De : − ⊗ − ⊗ ⊗ × k ! 2 i P † (D) := D arg min Tr D(V) D , (52) with C = H V Vk, one finds the following rep- Mark e e 2 k k P V∈M − resentation for− the pseudo-Kolmogorov generator: L = d2 −1 K ΛK. 2 2 where 2 is set of complex d d matrices. We note, The solution of the master equation (43) takes the form Md × that it is practical to consider V in the form V = Vre + p(t) = S(t)pin, (46) iVim, where Vre and Vim are real matrices, and perform an optimization over two spaces of real matrices. where the pseudostochastic matrix S is given by S(t) = So the necessary and sufficient condition for a gen- exp(Lt) for time-independent generator L and erator L to correspond to time-independent Markovian evolution is to satisfy a relation Z t 0 0 S(t) = T exp L(t )dt , (47) unit(L) + Mark(L unit(L)) = L (53) t0=0 P P − P which is a generalization of the condition (31). if there is time dependence in the generator. Finally, we introduce a projector on the set of real matrices corresponded to time-independent Markovian noise IV. RELATION BETWEEN generators in Eq. (43). For this purpose, we represent the PSEUDOSTOCHASTICITY AND generator L in the form L = H + D, where the first term NONCLASSICALITY H is a Hamiltonian part, defined by Eq. (26), and D is a remaining part corresponding to the Markovian noise. Here we apply the developed SIC-POVM probability Next, we refer to H and D as unitary evolution generator representation to study nonclassical features of quantum and Markovian noise generators correspondingly. system dynamics. We consider a general scheme of a Consider a decomposition of the noise operators in the quantum-mechanical experiment (see Fig.2), which has Pd2 (i) (i) form Vk = i=1 vk , where vk are arbitrary complex three parts: (i) preparation of a quantum state controlled numbers. Then the operator D takes the form by a random classical input X taking values from some finite set ; (ii) evaluation of this stated under the se- X X (i) (j)∗ X D = Ωi,j vk vk , (48) quence of quantum channels; and (iii) measurement of i,j k the resulting state, which provides a (generally random) outcome Y from some finite set . Let us consider these where steps in more detail and in the frameworkY of the pseudostochastic representation. We assume that it is possible −1 (i) (j)∗ 1 (j) (i) Ωi,j = K σ σ σ σ 1 to construct SIC-POVMs for all involved dimensions of ⊗ − 2 ⊗ quantum systems. 1 Let the input X have a probability distribution p (x) 1 σ(i)∗σ(j)∗ K. (49) X 2 (x ). We can also introduce a corresponding prob- − ⊗ ∈ X ability vector pX with components given by values (i) (j)∗ |X | Noticing that P v v can be considered as (i, j)th pX (x). Let for every x the prepared state in the k k k ∈ X case X = x is described by a din din density matrix ρx. element of some semipositive matrix v, we can introduce × a complex matrix V such that Then the preparation of the quantum state can be de- 2 scribed with din stochastic matrix G, each of whose † X (i) (j)∗ columns is a SIC-POVM×|X | vector pin,x corresponded to the [VV ]i,j = vi,j = vk vk (50) k state ρx. By multiplying G on the vector pX we obtain in P in,x a SIC-POVM vector p = pX (x)p , which cor- Px∈X (hereafter, [a]i,j stands for a (i, j)th matrix element of responds to a prepared state x∈X pX (x)ρx. So one can a matrix a). Although the dimensionality of V can be see that every quantum state preparation process can be 8 considered as a classical probabilistic process. However, dent and satisfies relation (53). We raise the question the opposite statement is not true. Each of columns of about whether the resulting evolution demonstrates non- 2 G strictly belongs to din-dimensional qplex, so the set classical features related to negative elements in the pseu- of possible stochastic matrices corresponded to quantum dostochastic matrix S(t) = eLt or it is classical-like. The state preparation is strictly inside the set of all possible results are as follows. 2 din stochastic matrices. We× |X then | consider an evolution of the prepared state Theorem 1. Consider a time-independent Markovian during actions of quantum channels. As discussed above, master equation governed by the generator L given in the action of each channel can be described by a multi- the SIC-POVM probability representation. The resulting Lt plication on a pseudostochastic matrix on an input SIC- map S(t) = e is stochastic for any t > 0 if and only if POVM probability vector. Therefore, the total action of all nondiagonal elements of L are non-negative: Li,j 0 for any i = j. ≥ the channels on the quantum states is given by a prod- 6 uct of pseudostochastic matrices, which is also pseudos- Proof. First, we prove that if all nondiagonal elements of tochastic. Let the output state be described with the L are non-negative then S(t) is stochastic. In order to dout dout density matrix. Then the action of quantum × 2 2 do so, we represent S(t) for some fixed t in the form channels is given by the corresponding dout din pseudostochastic matrix S and the resulting state× is given by t n a SIC-POVM probability vector pout = Spin. eLt = lim I + L . (54) Finally, we consider a measurement process whose out- n→+∞ n put is described by a discrete random variable Y taking Let el := maxi Li,i . Since Li,j 0 for i = j, all elements values from a finite set . Let M be a pseudostochastic | | ≥ 6 matrix of the measurementY under consideration. Then of the matrix I + L t are non-negative for n elt . So n ≥ d e -dimensional probability vector pY of the random vari- the expansion (54) is product of matrices, whose elements Y in able Y can be obtained as pY = Mp = QpX , where are non-negative, and therefore the resulting matrix also Q = MSG is a stochastic matrix. It is easy consists of non-negative elements only. to see that by varying|Y| × |XM | , S, and G it is possible to Then, we prove that because S(t) is stochastic for all construct any desired stochastic matrix Q. t > 0, then all non-diagonal elements of L are non- From this analysis, one can note that the non- negative. We prove this statement by contraposition. Let ∗ ∗ classicality of the considered process is related to negative Li∗,j∗ < 0 for some i , j . Consider small values of t, and conditional probabilities of pseudostochastic maps, which an expansion of S(t) in the form correspond to quantum channels and quantum measure- 2 ment. Otherwise, if all matrices, which are involved S(t) = 1 + Lt + ∆L(t), ∆L(t)i,j O(t ). (55) ∈ in our consideration, are stochastic, then we can conclude that the whole process appears to be classical-like Since ∆L(t)i∗j∗ is of the second order of smallness, it is stochastic process. In this case, obtaining of measure- possible to find small enough t > 0 such that Li∗j∗ t + ment results in the real quantum experiment can be sim- ∆L(t)i∗j∗ < 0 and thus Si∗j∗ (t) < 0. ulated by a classical sampling technique: One can first sample a classical random variable x from the distribu- The above theorem has an important corollary regard- tion p , then sample a random variable from the dis- ing to noiseless unitary processes with L = unit(L) = H. X P tribution pin,x, next sample a random variable from the certain column of the stochastic matrix corresponding to Corollary 1. Any non-zero unitary evolution generator H = 0, given in the SIC-POVM probability representa- the first channel, and so on, until the final measurement. 6 Ht We note that this technique turns out to be invalid in the tion, spawns a pseudostochastic matrix S(t) = e with case of pseudostochastic matrices with negative elements at least one negative element for some t > 0. due to the necessity of sampling from distributions with Proof. The statement directly follows from facts that negative elements. Below we consider formulate neces- Hi,j = Hji and H is non-zero matrix. sary and sufficient for obtaining non-classical behavior of − the resulting process. We would like to note an alter- Thus, as one may expect, the necessary condition for native consideration of non-classical features of quantum quantum evolution to become classical-like is the pres- system dynamics in the context of SIC-POVM is pre- ence of decoherence processes, which appear, e.g., due to sented in Ref. [64]. Markovian noise. We also note that for some values of t > 0 the resulting evolution matrix S(t) = eHt can be stochastic; e.g., in the case of two-level system of Rabi A. Characterizing nonclassicality of oscillations S(t) is stochastic and equals to the identity time-independent Markovian dynamics matrix for t = nT , where n is a positive integer and T is a period of oscillations. However, between these dis- Here we consider a time-independent Markovian dy- crete moments S(t) appears to be pseudostochastic with namics governed by Eq. (43), where L is time indepen- negative elements. 9

stoch. matrix where L := log S. We can define the non-Markovianity stoch. matrix ps.-stoch. matrix ps.-stoch. matrix ps.-stoch. matrix measure δnMark(S) as Quantum state Quantum ... Quantum Quantum preparation channel 1 channel N measurement 1 p δnMark(S) := (S, SMark), (60) d2 D classical random 2 variables where (S1, S2) = Tr (S1 S2) is the Hilbert- D − Figure 2. Scheme of a quantum-mechanical experiment. Schmidt distance between matrices S1 and S2. The values of δnMark(S) have a clear physical meaning: They show an average difference between elements of S and The result of Theorem1 allows us to construct a mea- corresponding Markovian projection SMark. We note sure of nonclassicality of time-independent Markovian that all elements Si,j can be measured experimentally dynamics determined by a generator L. Let with some absolute accuracy δ. So in order to conclude with confidence that S exhibit non-Markovianity, one (Le) = max max(0, Lei,j) (56) should have δnMark(S) larger than δ. N i6=j − be a magnitude of the smallest negative non-diagonal V. EXPERIMENTAL STUDY OF element of some matrix L. In line with the results of e SUPERCONDUCTING CIRCUITS Theorem1, (L) seems to be suitable for the characterization ofN the nonclassicality of L. However, it is not invariant under change of the basis of the underlying d- In this section, we apply the SIC-POVM probability dimensional Hilbert space determined by applying sone representation to experimental study of quantum pro- unitary operator U to computational basis vectors. It is cesses realizing during implementation of quantum gates of the IBM QX4 cloud-based superconducting quantum easy to see that a transformation ψi U ψi of ba- 2 | i → | i processor [55]. We consider a qubit dynamics deter- sic SIC-POVM vectors ψ d yields a transformation i i=1 mined by implementation of single-qubit gate, followed L ULUT, where U {|is ai} pseudostochastic representa- by a SIC-POVM measurement. The SIC-POVM mea- tion→ of U. Consequently, we can introduce the measure surement is realized using an additional qubit in the pure of nonclassicality δ (L) for given generator L in the quant state following form: T 1 1 + 1/√3 1/√3 i/√3 δquant(L) := max (ULU ), (57) φ φ = − , (61) U∈U N | ih | 2 1/√3 + i/√3 1 1/√3 − where the set of all possible pseudostochastic matrices U applying controlled-NOT (CNOT) gate followed by the corresponding to unitary operators is given by Hadamard gate, and performing standard projective measurements in the computational basis on both qubits d2 X (i) [see Fig.3(a)]. We note that alternative ways of perform- = exp H : H = λiH , λi R . (58) U { ∈ } ing SIC-POVM measurements can also be used [65]. Let i=1 ab be a classical two-bit string produced at each launch of the considered measurement scheme. It is easy to check that there is one-to-one correspondence between possi- B. Measuring non-Markovianity of the dynamics ble outcomes ab = 00, 01, 10, 11 and SIC-POVM effects Π1/2, Π2/2, Π3/2, and Π4/2 defined by projectors In the above consideration, we stressed on a partic- 1 (4) 1 (1) (2) (3) ular case of time-independent Markovian generator and Π1 = ψ1 ψ1 = σ + (σ σ + σ ) , discussed its nonclassicality. Here we introduce an addi- | ih | 2 √3 − tional quantity, which indicates how accurately the con- 1 (4) 1 (1) (2) (3) Π2 = ψ2 ψ2 = σ + (σ + σ σ ) , sidered Markovian type of dynamics can describe the ob- | ih | 2 √3 − served dynamics, which is given by a pseudostochastic 1 (4) 1 (1) (2) (3) matrix S. In other words, we are considering a measure Π3 = ψ3 ψ3 = σ + ( σ + σ + σ ) , 2 √ non-Markovianity of quantum processes. We note that | ih | 3 − our approach in general is similar to the one proposed in 1 (4) 1 (1) (2) (3) Π4 = ψ4 ψ4 = σ + ( σ σ σ ) Ref. [43]. | ih | 2 √3 − − − Consider a projection of S on a set of pseudostochas- (62) tic matrices corresponding to Markovian evolution with (here σ(1), σ(2) and σ(3) stand for standard x-, y- and z- time-independent generators: Pauli matrices respectively and σ(4) is a 2 2 identity matrix). Bloch vectors of SIC-POVM projectors× and the SMark := exp [ unit(L) + Mark(L unit(L))] , (59) state φ φ are shown in Fig.4. P P − P | ih | 10

Quantum state Quantum SIC-POVM measurement alization, it would be given by preparation channel a  1 1 1 1  ϕ H − ideal 1 1 1 1 1 (a) SU =  −  . (65) in,i b 2  1 1 1 1 ψ U 1 1 1− 1 − Because of experimental imperfections and decoherence, i pin,i pout,i a,b S has some forms different from Sideal, and, actually, (b) S U U G U Sdec the purpose of the experiment is to find out. We note that ideal SU possesses negative elements of maximal possible for this dimensionality magnitude equal to 1/d = 1/2 [see Figure 3. The experimental setup. For the tomography of the Eq. (36)]. SIC-POVM measurement operation an applying of U (S ) is U Then the multiplication by a pseudostochastic ma- skipped. In panel (a), standard representation as a circuit is presented (standard notations for CNOT and Hadamard trix, which corresponds to the SIC-POVM measurement gates are used). In panel (b), SIC-POVM probability repre- comes. In the ideal case, it would be described by the sentation is shown. 4 4 identity matrix. However, the realistic experimental× conditions it can also suffer from different kinds z of imperfections, especially because of employing two- qubit CNOT gate. In order to take imperfection into ψ account, we model the real SIC-POVM measurement | 3i with a sequence of a decoherence channel described by ψ | 1i φ some pseudostochastic matrix Sdec and ideal SIC-POVM | i measurement described with the identity matrix. The y measurement outcome ab is obtained by sampling a ran- ψ | 4i dom variable from the resulting probability distribution out,i in,i x p = Sdec SU p . ψ For each input state, which is given by i 1, 2, 3, 4 , | 2i we ran the experiment N = 1024 times and∈ { calculate} the output four-dimensional probability vectors pout,i as Figure 4. The Bloch vectors corresponded to SIC-POVM pro- frequencies of obtaining corresponding outcomes in SIC- 4 jectors {|ψiih|ψii|}i=1 and the state |φihφ|. POVM measurement. Then, by employing known ele- in,i ments of input probability vectors pj , we obtain a complete system of 12 linear equations on the elements of Our experiment is designed as the tomographic recon- a pseudostochastic matrix SdecSU from general matrix struction of the quantum channel related to the imple- equations mentation of a single-qubit gate of the following form: pout,i = (S S ) pin,i, i = 1, 2, 3, 4. (66) 1 0 dec U U = , (63) 0 i We note that the d2 d2 (pseudo)stochastic matrix is characterized by d2(d2× 1) independent elements due to also known as the S gate. In order to perform the the normalization requirement− for each column. tomography protocol, we apply U to a set of four in- raw 4 Let Sdec,U be a solution of the the system of linear put states ψi i=1–the same which were used for SIC- POVM construction–and{| i} then measure resulting states equations. The reconstruction error of each element of with the designed SIC-POVM measurement. pseudostochastic matrices can be estimated at a level of δ 0.031 as shown in AppendixD. However, the result- The corresponding scheme of the experiment in the ≈ SIC-POVM probability representation is presented in ing matrix is not guaranteed to be physical. It means Fig.3(b). The quantum state preparation is presented that the corresponding Choi matrix may have negative by stochastic matrix G = pin,1 pin,2 pin,3 pin,4, where eigenvalues. In order to obtain a physical result with cer- tainty we introduce an operation of projection on a set each pin,i is a SIC-POVM probability vector, correspond- of pseudostochastic maps corresponding to CPTP maps. ing to the state ψi ψi , with elements | ih | We derive it in a way similar to the derivation of Mark( ). We consider the Kraus representation of anP arbitrary· pin,i = Tr(Π Π )/2 = (2δ + 1)/6. (64) j i j i,j CPTP map presented in Eq. (15). Taking into account that each Kraus operator can be decomposed in the form We note that for a particular known value of i the proba- (i) (i) in,i P (i) bility distribution is given by p . Then this probability Ak = i ak σ with complex coefficient ak , we obtain distribution is multiplied by a pseudostochastic matrix a corresponding pseudostochastic matrix in the form S = P P (i) (j)∗ −1 (i) (j) S corresponded to the gate U. In the case of ideal re- a a Θi,j, where Θi,j = K σ σ Kin. U i,j k k k out ⊗ 11

The obtained results are presented in TableII. First of all, we note that the level of non-Markovianity for both processes is quite small compared to measurement error δ. Thus, the observed processes can be eﬃciently described with time-independent Markovian equations. Second, we see that the imperfect SIC-POVM measurement can be described by purely classical stochastic process. In the chosen basis, it has small nondiagonal negative elements; however, they can be completely removed by shifting to a basis: This fact corresponds to δquant(Hdec + Ddec) = 0. In contrast, the process of the Figure 5. Experimentally reconstructed pseudostochastic ma- gate implementation shows a clear nonclassical behavior: trices of the single-qubit gate U (a) and the imperfect SIC- There are negative elements in the pseudostochastic ma- POVM measurement (b). trix which can not be removed by any basis change. So we see that SIC-POVM probability representation opens interesting possibilities for deep studying of quantum dy- P (i) (j)∗ namics and revealing its classical and nonclassical fea- Then we can treat k ak ak as elements of some semipositive matrix VV†. This trick allows us to introduce a tures. function

X † S(V) = [VV ]i,jΘi,j (67) VI. CONCLUSION i,j In this work, we have considered a probability repre- which parametrizes a CPTP map with complex d2 d2 × sentation of quantum mechanics based on employing SIC- matrix V. Finally, we introduce a projection operator POVM. In this representation, d-dimensional quantum systems are described with true d2-dimensional proba- ! 2 bility distributions (probability vectors), while their dy- CPTP(Se) = S arg min Tr S(V) Se , (68) P V∈Md2 − namics and calculating measurement outcomes are governed by pseudostochastic matrices–matrices of condi- which gives a physical pseudostochastic matrix closes to tional probabilities without restriction on elements posi- some matrix Se. tivity. We employ the constructed projector for obtaining We have derived an equation of a SIC-POVM probabil- raw Sdec,U := CPTP(S ). Then in order to get Sdec and ity vector evolution corresponding to the von Neumann P dec,U SU separately, we performed the same experiment with- equation and the GKSL master equation. In the case out implementation of the gate U. It allows reconstruct- of unitary evolution, we have shown that the evolution ing Sdec and obtaining SU with the use of the inverse generator in the SIC-POVM probability representation is −1 2 2 matrix Sdec and known matrix Sdec,U . We note that the given by d d real antisymmetric matrix from a certain similar trick is used for improving results of quantum (d2 1) linear× subspace of all d2 d2 real antisymmetric state and process tomography [58, 62]. Because of the matrices.− This fact has allowed us× to construct a projec- same number of circuit runs N = 1024 the experimental tor on the space of physical unitary evolution generators. error of the reconstruction δ remained the same. Then we have also shown that a corresponding evolution The results of Sdec and SU reconstruction are pre- operator is the SIC-POVM probability representation is sented in the first row of TableII and Fig.5. The values given by orthogonal pseudobistochastic matrix preserv- of fidelities regard to the identity matrix and U [see (63)] ing both l1 and l2 norms. In the case of the dissipative equal 0.89 and 0.94, respectively. evolution, we have also shown that is possible to con- We then study a possibility to describe the observed struct a projector on a set of time-independent Marko- processes with a time-independent Markovian master vian dissipators in the SIC-POVM probability represen- equation. For this purpose, we compute ln(Sdec) and tation. ln(SU ) and then extract unitary evolution and Markovian We have applied our results to studying non-classical evolution parts by employing corresponding projectors. features of quantum system dynamics. We have proven We note that the unitary term is extracted first (e.g., the theorem about necessary and sufficient conditions on HU = unit[ln(Sdec)]), and then Mark is applied to the the evolution generator which make the resulting evolu- P P orgthogonal part (e.g. DU = Mark[ln(Sdec) HU ]). We tion described with stochastic (but not pseudostochastic) also computed measures of non-MarkovianityP − given by matrices. Then we have constructed a practical measures δnMark(Sdec) and δnMark(SU ) [see Eq. (60)] and measures of nonclassicality and non-Markovianity for the observed of non-classicality for Markovian approximations, which quantum processes described with pseudostochastic ma- are given by δquant(HU + DU ) and δquant(Hdec + Ddec) trices. [see Eq. (57)]. Finally, we have applied our approaches to the experi- 12

Gate implementation SIC-POVM measurement  0.517 −0.399 0.478 0.489  0.893 0.002 0.100 0.010  0.449 0.504 −0.418 0.473  0.018 0.877 0.012 0.071 Pseudostochastic matrix SU =   Sdec =    0.502 0.403 0.466 −0.448 0.102 0.018 0.924 0.058 −0.467 0.493 0.475 0.487 0.014 0.111 −0.002 0.874  0 −0.768 −0.033 0.801   0 −0.010 0.012 −0.002  0.768 0 −0.761 −0.007  0.010 0 0.011 −0.021 Unitary evolution generator HU =   Hdec =    0.033 0.761 0 −0.794 −0.012 −0.011 0 0.023  −0.801 0.007 0.794 0 0.002 0.021 −0.023 0 −0.044 0.046 0.070 0.001  −0.131 0.008 0.089 0.007   0.013 −0.105 0.005 0.030   0.007 −0.136 −0.001 0.098  Markovian evolution generator DU =   Ddec =    0.033 0.001 −0.110 0.003   0.113 0.025 −0.102 0.035  −0.004 0.058 0.035 −0.034 0.011 0.103 0.014 −0.139 Non-Markovianity δnMark(SU ) = 0.003 δnMark(Sdec) = 0.007 Non-classicality δquant(HU + DU ) = 0.781 δquant(Hdec + Ddec) = 0

Table II. Experimental results. mental study of superconducting quantum circuits run on with a three-dimensional vector the IBM quantum processor. We have demonstrated that  p  a noise appearing in the imperfect SIC-POVM measure- e1 p := p , (A1) ment demonstrates classical behavior, while quantum dy- e  e2  p , namics during application of the single-qubit gate has e3 clear nonclassical features. Meanwhile, we have shown where that both processes are well described by the Markovian approximation that is revealed by our non-Markovianity p1 := + ρ + , p2 := R ρ R , p3 := 0 ρ 0 e h | | i e h | | i e h | | i measure. 1 1 (A2) + := ( 0 + 1 ), R := ( 0 + i 1 ). | i √2 | i | i | i √2 | i | i 3 One can see that the probabilities pi define projec- {e }i=1 ACKNOWLEDGMENTS tions on x, y, and z axes of the Bloch sphere. We note that all three components of pe are independent. In order to obtain the relation between the vector pe We acknowledge use of the IBM Q Experience for this and the SIC-POVM vector p we can employ Eqs. (11) work. The views expressed are those of the authors and and (12). Finally, we obtain do not reflect the official policy or position of IBM or the IBM Q Experience team. The authors thank A. Ulanov, pe = Fp, (A3) M. Gavreev, E. Tiunov, and D. Kurlov for fruitful dis- cussions. Results of Secs.II,IIIA,IIIB, andIV were where obtained by E.O.K. and V.I.M with the support from  1 + √3 1 + √3 1 √3 1 √3  1 the Russian Science Foundation Grant No. 19-71-10091. F = 1 √3 1 + √3 1 +− √3 1 − √3 , (A4) 2   Results of Sec.IIIC have been obtained by D.C. sup- 1 +− √3 1 √3 1 + √3 1 − √3 ported by Narodowe Centrum Nauki under the Grant − − No. 2018/30/A/ST2/00837. Results of Sec.V were ob- where we use the qubit SIC-POVM based on the projec- tained by A.K.F., A.O.M., and A.S.M. with the support tors set (62). from the Grant of the President of the Russian Federa- In order to obtain the opposite relation, we note that tion (Project No. MK923.2019.2). p4 = 1 p1 p2 p3. Then one can verify the following equality:− − − p   2√3 2√3 0 1 √3  p  e1 − 1 p2 1  0 2√3 2√3 1 √3  p2 Appendix A: Relation between SIC-POVM e  =  −    . (A5) probabilities and MUB probabilities for d = 2 pe3 2  2√3 0 2√3 1 √3  p3 1 0 0 0− 2 1 Here we provide a correspondence between the consid- Using (A5), we obtain the following relation: ered SIC-POVM probability representation and the alter-   native probability representation based projective MUB p1 2 2 2 √3 1 p1 − − e measurements in the case of two-level systems, which p2 √3  2 2 2 √3 1  p2   =  − −  e  . (A6) were extensively studied in Refs. [22–24]. In this repre- p3 12  2 2 2 √3 1  p3 − − e sentation, an arbitrary quantum state ρ is characterized p4 2 2 2 √3(1 + √3) 1 − − − 13

It can be rewritten in a more compact form: K−1 =  1 + √3 1 √3 1 + √3 1 √3  p = Tp + c, (A7) − − e 1 √3 + i√3 √3 i√3 √3 i√3 √3 + i√3  − − − −  . where 12 √3 i√3 √3 + i√3 √3 + i√3 √3 i√3 1 −√3 1 + √3− 1 √3− 1 +−√3 − −  1 1 1   3 √3  (C2) − − √3 1 1 1 1  3 √3  T =  −  , c =  −  . (A8) 6  1 1 1  12  3 √3  The antisymmetric matrices forming a basis for phys- − 1 1 1 3 +− 3√3 ical process take the forms − − −  0 0 1 1  0 0 1− 1 Appendix B: Examples of pseudostochastic matrices H(1) =   ,  1 1− 0 0  of PTP Ds for d = 2 −1 1 0 0 −  0 1 1 0  Here we consider two well-known maps in entangle- − T 1 0 0 1 ment theory: transposition map T (X) = X , and so- H(2) =  −  , (C3) called reduction map R : ( ) ( ),  1 0 0 1  L H → L H −0 1 1 0 − 1  0 1 0 1  R(X) = (1TrX X), (B1) d 1 1− 0 1 0 − H(3) =   . −  0 1− 0 1  which is unital PTP, but not CPTP. In the case of din = 1 0 1− 0 dout = d = 2 for SIC-POVM effects given by (62), one − finds the pseudobistochastic matrices corresponding to Consider a Hamiltonian H = σ3/2. In the probabil- transposition ity representations the dynamics is defined by the matrix H = 1 H(3). Then the resulting pseudobistochastic oper-   2 1 1 1 1 ator of the evolution reads 1 1 1 1− 1 ST =  −  , (B2) 2  1 1 1 1  U(t) = exp (Ht) 1− 1 1 1 −  1 + cos(t) sin(t) 1 cos(t) sin(t)  1 sin(t) 1− + cos(t) −sin(t) 1 cos(t) and to reduction map =  − −  , 2  1 cos(t) sin(t) 1 + cos(t) sin(t)  −sin(t) 1 cos(t) sin(t) 1− + cos(t)  1 1 1 1  − − 1 −1 1 1 1 (C4) SR =  −  . (B3) 2  1 1 1 1  which corresponds to the standard evolution operator 1 1− 1 1 − exp(it/2) 0 U(t) = exp ( iHt) = . (C5) − 0 exp( it/2) Appendix C: Construction of pseudostochastic − matrix of a unitary evolution for d = 2

Appendix D: Estimating statistical errors in the Here we present explicit forms of some basic matri- experimental reconstruction of a pseudostochastic ces used for SIC-POVM probability representation in the matrix case (d = 2)-dimensional systems (qubits) and provide an example of a unitary pseudobistochastic evolution oper- In order to reconstruct an unknown pseudostochastic ator construction. We consider the SIC-POVM based matrix S in our quantum process tomography protocol, on the projectors given in Eq. (62). The matrix K and we considered a set of equations its inverse, which deﬁnes transitions between SIC-POVM probability representation and the standard representa- 4 tion, take the following forms: out,i X in,i pj = Sjkpk , i, j = 1, 2, 3, 4 (D1) k=1  1 + √3 1 √3 1 + √3 1 √3  − − 1 √3 + i√3 √3 i√3 √3 i√3 √3 + i√3 where the values of pin,i are given by Eq. (64), and the K =  − − − −  , k 2 √3 i√3 √3 + i√3 √3 + i√3 √3 √3  values of pout,i are obtained from the experiment as fre- − − − − j 1 √3 1 + √3 1 √3 1 + √3 quencies of getting a jth SIC-POVM measurement out- − − (C1) come for ith input state. 14

In order to write down a set of linear equations on the where components of S in a standard form, we can represent S as follows:  5 1 1 1  T S = s(1) s(2) s(3) s(4) , (D2) 1 1− 5 −1 −1 P−1 =  − − −  . (D7) 2  1 1 5 1  − − − where s(i) are four-dimensional column vectors. We note 1 1 1 5 − − − that due to normalization condition on the columns of (4) (1) (2) (3) S we have sj = 1 sj sj sj , so we need to (1) (2) − (3)− − reconstruct s , s and s only. In order to estimate statistical errors of the obtained Then by introducing result, we ﬁrst estimate statistical errors of reconstructed

T output probabilities. The mean squared error for each of (j) h out,1 out,2 out,2 out,4i (i) out,i q := pj pj pj pj (D3) qj = pj can be obtained in the following way: and 2 out,j out,j  in,1 T   (i) (1 pi )pi 1 p 3 1 1 1 δqj − , (D8) in,2 T N 4N p  1 1 3 1 1 ≈ ≤ P :=   =   , (D4) pin,3 T 6  1 1 3 1  in,4 T p 1 1 1 3 where N = 1024 is number measurements performed for each input state. Then the resulting statistical error for we can write the basic system of linear equations on the components of S can be estimated as components of S in a compact form

(j) (j) q = Ps , j = 1, 2, 3. (D5) v u 4 2 (j) uX (j) 1 1 Its solution is then given by δs = t P−1 δq = 0.031. i | ik | k ≤ √ 32 ≈ k=1 N s(j) = P−1q(j), j = 1, 2, 3, (D6) (D9)

[1] M.A. Nielsen and I.L. Chuang, Quantum Computation [12] C. Ferrie and J. Emerson, Framed Hilbert space: Hang- and Quantum Information (Cambridge University Press, ing the quasi-probability pictures of quantum theory, Cambridge, UK, 2000). New J. Phys. 11, 063040 (2009). [2] E.P. Wigner, On the quantum correction for thermody- [13] C. Ferrie, Quasi-probability representations of quantum namic equilibrium, Phys. Rev. 40, 749 (1932). theory with applications to quantum information science, [3] E. Husimi, Some formal properties of the density matrix, Rep. Prog. Phys. 74, 116001 (2011). Proc. Phys. Math. Soc. Jpn. 23, 264 (1940). [14] H. Zhu, Permutation symmetry determines the discrete [4] R.J. Glauber, Photon correlations, Phys. Rev. Lett. 10, Wigner function, Phys. Rev. Lett. 116, 040501 (2016). 84 (1963). [15] E.F. Galvão,Discrete Wigner functions and quantum [5] E.C.G. Sudarshan, Equivalence of semiclassical and computational speedup, Phys. Rev. A 71, 042302 (2005). quantum mechanical descriptions of statistical light [16] C. Cormick, E. F. Galvão,D. Gottesman, J. P. Paz, and beams, Phys. Rev. Lett. 10, 277 (1963). A.O. Pittenger, Classicality in discrete Wigner functions, [6] M. Hillery, R. OConnell, M. Scully, and E. Wigner, Dis- Phys. Rev. A 73, 012301 (2006). tribution functions in physics: Fundamentals, Phys. Rep. [17] V. Veitch, C. Ferrie, D. Gross, and J. Emerson, Negative 106, 121 (1984). quasi-probability as a resource for quantum computation, [7] J. Weinbub and D.K. Ferry, Recent advances in Wigner New J. Phys. 14, 113011 (2012). function approaches, Appl. Phys. Rev. 5, 041104 (2018). [18] V. Veitch, S. A. H. Mousavian, D. Gottesman, and J. [8] R.W. Spekkens, Negativity and contextuality are equiv- Emerson, The resource theory of stabilizer quantum com- alent notions of nonclassicality, Phys. Rev. Lett. 101, putation, New J. Phys. 16, 013009 (2014). 020401 (2008). [19] M. Howard, J. Wallman, V. Veitch, and J. Emerson, Con- [9] W.K. Wootters, A Wigner-function formulation of finite- textuality supplies the magic for quantum computation, state quantum mechanics, Ann. Phys. (N.Y.) 176, 1 Nature (London) 510, 351 (2014). (1987). [20] N. Delfosse, P. Allard Guerin, J. Bian, and R. [10] K.S. Gibbons, M.J. Hoffman, and W.K. Wootters, Dis- Raussendorf, Wigner function negativity and contextu- crete phase space based on finite fields, Phys. Rev. A 70, ality in quantum computation on rebits, Phys. Rev. X 5, 062101 (2004). 021003 (2015). [11] D. Gross, Hudsons theorem for finite-dimensional quan- [21] H. Pashayan, J.J. Wallman, and S.D. Bartlett, Esti- tum systems, J. Math. Phys 47, 122107 (2006). mating outcome probabilities of quantum circuits using 15

quasiprobabilities, Phys. Rev. Lett. 115, 070501 (2015). triglia, Stochastic evolution of finite level systems: Clas- [22] V.N. Chernega, O.V. Manko, and V.I. Manko, Probabil- sical versus quantum Phys. Scr. 87, 045015 (2013). ity representation of quantum observables and quantum [42] D. Chru´sciński,V.I. Man’ko, G. Marmo, and F. Ven- states, J. Russ. Laser Res. 38, 324 (2017). triglia, On pseudo-stochastic matrices and pseudo- [23] V.N. Chernega, O.V. Man’ko, and V.I. Man’ko, God positive maps, Phys. Scr. 90, 115202 (2015). plays coins or superposition principle for classical proba- [43] M.M. Wolf, J. Eisert, T.S. Cubitt, and J.I. Cirac, Assess- bilities in quantum suprematism representation of qubit ing non-Markovian quantum dynamics, Phys. Rev. Lett. states, J. Russ. Laser Res. 39, 128 (2018). 101, 150402 (2008). [24] A. Avanesov and Man’ko, Statistical properties of qutrit [44] H.-P. Breuer, E.-M. Laine, and J. Piilo, Measure for the in probability representation of quantum mechanics, degree of non-Markovian behavior of quantum processes Phys. A 533, 121898 (2019). in open systems, Phys. Rev. Lett. 103, 210401 (2009). [25] A. Avanesov and V.I. Man’ko, Probability representation [45] A.´ Rivas, S.F. Huelga, and M. B. Plenio, Entanglement of quantum channels, Lobachevskii J. Math. 40, 1444- and non-Markovianity of quantum evolutions, Phys. Rev. 1449 (2019). Lett. 105, 050403 (2010). [26] C.M. Caves, C.A. Fuchs, and R. Schack, Quantum prob- [46] S. Luo, S. Fu, and H. Song, Quantifying non- abilities as Bayesian probabilities, Phys. Rev. A 65, Markovianity via correlations, Phys. Rev. A 86, 044101 022305 (2002). (2012). [27] J.M. Renes, R. Blume-Kohout, A. J. Scott, and C.M. [47] H.-P. Breuer, Foundations and measures of quantum non- Caves, Quantum probabilities as Bayesian probabilities, Markovianity, J. Phys. B: At. Mol. Opt. Phys. 45, 154001 J. Math. Phys. 45, 2171 (2004). (2012). [28] S.N. Filippov and V.I. Man’ko, Symmetric information- [48] M.J.W. Hall, J.D. Cresser, L. Li, and E. Andersson, ally complete positive operator valued measure and prob- Canonical form of master equations and characterization ability representation of quantum mechanics, J. Russ. of non-Markovianity, Phys. Rev. A 89, 042120 (2014). Laser Res. 31, 211 (2010). [49] D. Chru´sciński and S. Maniscalco, Degree of non- [29] C.A. Fuchs, M. C. Hoang, and B.C. Stacey, The SIC Markovianity of quantum evolution, Phys. Rev. Lett. question: History and state of play, Axioms 6, 21 (2017). 112, 120404 (2014). [30] D.M. Appleby, A. Ericsson, and C.A. Fuchs, Properties [50] B. Bylicka, D. Chru´sciński, and S. Maniscalco, Non- of QBist state spaces, Found. Phys. 41, 564 (2011). Markovianity and reservoir memory of quantum chan- [31] C.A. Fuchs and R. Schack, Quantum-Bayesian coherence, nels: A quantum information theory perspective, Sci. Rev. Mod. Phys. 85, 1693 (2013). Rep. 4 5720 (2014). [32] J. ehek, B.-G. Englert, and D. Kaszlikowski, Minimal [51] D. Chru´scińskiand F. A. Wudarski, Non-Markovianity qubit tomography, Phys. Rev. A 70, 052321 (2004). degree for random unitary evolution, Phys. Rev. A 91, [33] T. Durt, C. Kurtsiefer, A. Lamas-Linares, and A. Ling, 012104 (2015). Wigner tomography of two-qubit states and quantum [52] G. Torre, W. Roga, and F. Illuminati, Non-Markovianity cryptography, Phys. Rev. A 78, 042338 (2008). of Gaussian channels, Phys. Rev. Lett. 115, 070401 [34] Z.E.D. Medendorp, F.A. Torres-Ruiz, L.K. Shalm, (2015). G.N.M. Tabia, C.A. Fuchs, and A.M. Steinberg, Experi- [53] C. Pineda, T. Gorin, D. Davalos, D.A. Wisniacki, and mental characterization of qutrits using symmetric, infor- I. Garc´ıa-Mata,Measuring and using non-Markovianity, mationally complete positive operator-valued measures, Phys. Rev. A 93, 022117 (2016). Phys. Rev. A 83, 051801 (2011). [54] C.-F. Li, G.-C. Guo, and J. Piilo, Non-Markovian quan- [35] N. Bent, H. Qassim, A.A. Tahir, D. Sych, G. Leuchs, tum dynamics: What does it mean? Europhys. Lett. 27, L.L. Sánchez-Soto, E. Karimi, and R.W. Boyd, Ex- 50001 (2019). perimental realization of quantum tomography of pho- [55] IBM Quantum Experience [research.ibm.com/ibm-q]. tonic qudits via symmetric informationally complete pos- [56] V. Gorini, A. Kossakowski, and E.C.G. Sudarshan, Com- itive operator-valued measures, Phys. Rev. X 5, 041006 pletely positive dynamical semigroups of N-level systems, (2015). J. Math. Phys. 17, 821 (1976). [36] Y.-y. Zhao, N.-k. Yu, P. Kurzyński,G.-Y. Xiang, C.-F. [57] G. Lindblad, On the generators of quantum dynamical Li, and G.-C. Guo, Experimental realization of general- semigroups, Commun. Math. Phys. 48, 119 (1976). ized qubit measurements based on quantum walks, Phys. [58] F.B. Maciejewski, Z. Zimborás, and M. Oszmaniec, Miti- Rev. A 91, 042101 (2015). gation of readout noise in near-term quantum devices by [37] Z. Hou, J. Tang, J. Shang, H. Zhu, J. Li, Y. Yuan, K.-D. classical post-processing based on detector tomography, Wu, G.-Y. Xiang, C.-F. Li, and G.-C. Guo, Determin- arXiv:1907.08518. istic realization of collective measurements via photonic [59] J.B. DeBrota, C.A. Fuchs, M.C. Hoang, quantum walks, Nat. Commun. 9, 1414 (2018). B.C. Stacey, QBism Research Group [38] M. Appleby, C.A. Fuchs, B.C. Stacey, and H. Zhu, In- [http://www.physics.umb.edu/Research/QBism]. troducing the qplex: a novel arena for quantum theory, [60] A.J. Scott and M. Grassl, Symmetric informationally Eur. Phys. J. D 71 197 (2017). complete positive-operator-valued measures: A new com- [39] H. Zhu, Quasiprobability representations of quantum me- puter study, J. Math. Phys. 51, 042203 (2010). chanics with minimal negativity, Phys. Rev. Lett. 117, [61] A.J. Scott, SICs: Extending the list of solutions, 120404 (2016). arXiv:1703.03993. [40] J. van de Wetering, Quantum theory is a quasi-stochastic [62] B.I. Bantysh, D.V. Fastovets, and Y.I. Bogdanov, High- process theory, Electron. Notes Theor. Comput. Sci. 266, fidelity quantum tomography with imperfect measure- 179 (2018). ments, Proc. SPIE 11022, 110222N (2018). [41] D. Chru´sciński,V.I. Man’ko, G. Marmo, and F. Ven- 16

[63] N.G. van Kampen, Stochastic Processes in Physics and [65] M. Oszmaniec, F.B. Maciejewski, and Z. Puchala, Sim- Chemistry, 3rd ed, (North Holland, 2007). ulating all quantum measurements using only projec- [64] J.B. DeBrota, C.A. Fuchs, and B.C. Stacey, Symmetric tive measurements and postselection, Phys. Rev. A 100, informationally complete measurements Identify the es- 012351 (2019). sential diﬀerence between classical and quantum, Phys. Rev. Res. 2, 013074 (2020).