ebook img

The time-averaged limit measure of the Wojcik model PDF

0.25 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The time-averaged limit measure of the Wojcik model

The time-averaged limit measure of the Wojcik model 4 1 TAKAKOENDO 0 Department of Physics, 2 Ochanomizu University,2-1-1 Ohtsuka, Bunkyo, Tokyo, 112-0012, Japan l u NORIOKONNO J Department of Applied Mathematics, Faculty of Engineering, 0 Yokohama National University,Hodogaya, Yokohama, 240-8501, Japan 1 ] h p Abstract - h Weinvestigate“theWojcikmodel”introducedandstudiedbyWojciketal. [1],whichisaone-defectquantum t a walk (QW) having a single phase at the origin. They reported that giving a phase at one point causes an m astonishing effect for localization. There are three types of measures having important roles in the study of QWs: time-averaged limit measure, weak limit measure, and stationary measure. The first two measures imply [ a coexistence of localized behavior and the ballistic spreading in the QW. As Konno et al. [3] suggested, the 2 time-averaged limit and stationary measures are closely related to each other for some models. In this paper,we v focusonarelationbetweenthetwomeasuresfortheWojcikmodel. Thestationarymeasurewasalreadyobtained 0 by our previous work [2]. Here, we get the time-averaged limit measure by several methods. Our results show 7 that the stationary measure is a special case of thetime-averaged limit measure. 0 3 . 1 0 1 Introduction 4 1 Asaquantumcounterpartoftherandomwalk,quantumwalks(QWs)describemanykindsofphenomenainquantum : scale [11, 12]. There are two distinct types of QWs, one is the discrete time walk and the other is the continuous v i one. Discrete time QWs have been intensively studied in [8, 18]. Here, we focus on a two-state discrete time QW X in one dimension. The two-state corresponds to left and right chiralities, respectively [4]. It has been reported that r a one-dimensional discrete time QWs have characteristic properties, that is, localization and the ballistic spreading. TherearetwokindsoflimittheoremstoshowtheasymptoticbehavioroftheQWs: thetime-averagedlimittheorem correspondingto localization,andthe weaklimittheoremcorrespondingto the ballisticspreading. Inthis paper,we say that the walk starting from the origin exhibits localization if and only if its time-averaged limit measure at the origin is strictly positive. As Konno et al. [3] reported, the time-averaged limit and stationary measures are closely related to eachother. Therefore,we clarify the relationbetween the two measuresfor a suitable QWmodel. Wojcik et al. [1] showed that giving a phase at a single point in the QW on the line exhibits an astonishing localization effect. In this paper, we callthe model “the Wojcik model”. Our previous work [2] gavea stationarymeasure of the 1 model, andthis paper isa sequentialworkof[2]. We presentthe time-averagedlimit measure,derivedfromthe pass counting method [5, 7], the CGMV method [15], and the generating function method [3] explained in Sect. 5. Our result implies that the stationary measure is a special case of the time-averagedlimit measure. The rest of this paper is organized as follows. Section 2 gives the definition of the time-averaged limit measure and localization for the discrete time QW starting at the origin. In Sect. 3, we introduce the Wojcik model and presentourmainresults,Theorems1and2. Section4isdevotedtotheresultbasedontheCGMVmethod. Wegive the proofs of Lemma 1 in Sect. 5 and Theorem 2 in Sect. 6, respectively. Appendix A gives the proof of Theorem 1, and Appendix B presents the proof of Lemma 7. 2 The time-averaged limit measure and localization Inthissection,weintroducethetime-averagedlimitmeasureanddefinelocalizationfortheQWstartingattheorigin. First, we givethe notationof the space-inhomogeneousQWs onthe line. The walkerhas a coinstate described by a two-dimensionalvector which is called “the probability amplitude”. We define the coin state at position x and time n by ΨL(x) Ψ (x)= n . n ΨR(x) (cid:20) n (cid:21) The upper and lower elements express left and right chiralities, respectively. Let Ψ =T[...,ΨL( 1),ΨR( 1),ΨL(0),ΨR(0),ΨL(1),ΨR(1),...], n n − n − n n n n where T means the transposedoperation. The time evolutionis defined byits initialcoinstate Ψ and2 2unitary 0 × matrices U (x Z) : x ∈ a b U = x x , x c d x x (cid:20) (cid:21) where subscript x Z denotes the location. Then the evolution is determined by the following recurrence formula: ∈ Ψ (x)=P Ψ (x+1)+Q Ψ (x 1), n+1 x+1 n x−1 n − where a b 0 0 P = x x , Q = . x 0 0 x c d x x (cid:20) (cid:21) (cid:20) (cid:21) Note that P (resp. Q ) expresses that the walker moves to the left (resp. right) at position x in each time step. x x Let R =[0, ). Then for + ∞ Ψn =T ..., ΨΨRLn((−11)) , ΨΨRLn((00)) , ΨΨRLn((11)) ,... ∈(C2)Z, (cid:20) (cid:20) n − (cid:21) (cid:20) n (cid:21) (cid:20) n (cid:21) (cid:21) we define a map µ :Z [0, ] as n → ∞ µ (x)= ΨL(x)2+ ΨR(x)2 (x Z). n | n | | n | ∈ Our interest in this paper is the sequence of measures: µ ,µ ,µ ,... . 0 1 2 { } 2 If µ is a probability measure, let X be a random variable defined by µ , that is, for x Z, n n n ∈ P(X =x)=µ (x). n n Now we introduce the time average of µ (x) and its limit. The time average of µ (x) is defined by n n T−1 1 µ (x)= µ (x), T T n n=0 X and if the limit exists, we define the limit of µ (x) by T T−1 1 µ (x)= lim µ (x)= lim P(X =x). (1) ∞ T→∞ T T→∞T n n=0 X Here we put = µ =µΨ0 ZZ 0 :Ψ CZ , (2) M∞ { ∞ ∞ ∈ +\{ } 0 ∈ } whereµΨ∞0 representsthe dependence onthe initialstateΨ0 and{0}=T[...,0,0,0,...]. We callthe elementofM∞ the time-averagedlimit measure of the QW. Then, localization for discrete time QW is defined as follows. Definition 1 We say that localization for the QW starting at the origin happens if µ (0)>0. ∞ 3 Model and main results 3.1 Model In this paper, we treat a space-inhomogeneous QW, “the Wojcik model”, introduced by Wojcik et al. [1], whose time evolution is defined by the unitary matrices U (x Z) as follows. x ∈ H (x Z 0 ), Ux = ωH (x∈=0)\, { } (3) (cid:26) where ω =e2πiφ with φ (0,1). The model has a weight e2πiφ at the origin. Here, H is “the Hadamard matrix”: ∈ 1 1 1 H = . (4) √2 1 1 (cid:20) − (cid:21) In particular, if φ 0, then the Wojcik model becomes space-homogeneous and is equivalent to the well-known → Hadamardwalk whichis one ofthe most intensively studied QWs. We shouldnote thatKonno et al. [3] treatedthe QW in which det(U ) does not depend on the position x Z. However,the Wojcik model has x ∈ det(U )= 1, det(U )= ω2 (= 1if x Z 0 ). 0 x − − 6 − ∈ \{ } In this paper, we assume that the walk starts at the origin with the initial coin state ϕ = T[α,β], where α,β C ∈ and α2+ β 2 =1. | | | | 3 3.2 Main result 1: Time-averaged limit measure at the origin Inthissubsection,wegivethetime-averagedlimitmeasureattheorigin. Weshouldremarkthatwetreatthemeasure for x 1 case in subsection 3.3. Let us consider the initial coin states ϕ = T[1/√2,i/√2], or ϕ = T[1/√2, i/√2] | | ≥ − for a while. At first, we focus on the Hadamard walk (φ 0 case). For the initial coin state, the probability → distributionofthewalkissymmetricfortheoriginatanytime. Ifµ isaprobabilitymeasure,letX betherandom n n variable of the walk for the position x at time n. We compute the return probability at time n, which we denote it as r(H)(0)=P(X =0). We should note that r(H) (0)=0(n 0). By a brief calculation, we have n n 2n+1 ≥ r(H)(0)=0.5, r(H)(0)=0.125, r(H)(0)=0.125, r(H)(0)=0.07031, 2 4 6 8 r(H)(0)=0.07031, r(H)(0)=0.04882, r(H)(0)=0.04882,.... 10 12 14 In fact, we see lim r(H)(0)=0, (5) 2n n→∞ for example, see [5]. Equation (5) suggests that the Hadamard walk (φ 0 case) does not show localization → . From now on, we consider a space-inhomogeneous case, that is, φ (0,1) case. By a simple calculation, we ∈ have the same probability measure as that of the Hadamard walk at time n = 1,2,3 for the initial coin states ϕ = T[1/√2,ηi/√2] (η = 1, 1). However, we see that the probability measure at time n = 4 depends on the − parameter φ. Actually, we have 1 2(2+E) P(X = 4)=P(X =4) = , P(X = 2)=P(X =2)= , 4 4 4 4 − 16 − 16 2(3 2E) P(X =0) = − , 4 16 where E = C +ηS (η = 1, 1), C = cos(2πφ), and S = sin(2πφ). Hereafter, we present the time-averaged limit − measure at the origin for the Wojcik model. Let ΨL (0) Ψ (0)= 2n 2n ΨR (0) (cid:20) 2n (cid:21) be the probability amplitude at time 2n at the origin. Then, we obtain an explicit expression for Ψ (0) as follows. 2n Lemma 1 Let ϕ=ϕ(η)=T[1/√2,ηi/√2](η =1, 1) be the initial coin state. Then, we have − n k k 1 ω( 1+ηi) 1 Ψ (0)= r∗ − , 2n √2  2aj−1 2 ηi kX=1(a1,...,aXk)∈(Z>)k: jY=1 (cid:18) (cid:19) (cid:20) (cid:21) a1+···+ak=n   for n 1, where Z = 1,2, , and > ≥ { ···} ∞ 1 z2+√1+z4 r∗zn = − − . n z n=1 X We prove Lemma 1 in Sect. 5. Noting that the return probability is defined as P(X = 0) = Ψ (0) 2 = 2n 2n k k ΨL (0)2+ ΨR (0)2, we have | 2n | | 2n | 4 Lemma 2 Let ϕ = ϕ(η) = T[1/√2,ηi/√2] (η = 1, 1) be the initial coin state. Then the limit of the return − probability at time 2n for the parameter φ is given as follows. c(φ) = lim r (0) 2n n→∞ 2 2 1 √2C 1 √2C − + = 4 − I (φ)I (η)+4 − I (φ)I (η), 3 2√2C−! (1/4,1) {1} 3 2√2C+! (0,3/4) {−1} − − where I (x)=1(x A), I (x)=0(x A), and A A ∈ 6∈ π √2 C =cos 2πφ = cos(2πφ)+sin(2πφ) , − − 4 2 { } (cid:16) π(cid:17) √2 C =cos 2πφ+ = cos(2πφ) sin(2πφ) . + 4 2 { − } (cid:16) (cid:17) Interestingly, when φ (0,1), we see that the inequality ∈ c(φ)>0 holds except for η = 1 with φ (0,1/4] or η = 1 with φ [3/4,1). On the other hand, when φ 0, we have ∈ − ∈ → c(φ)=0 which implies that the Hadamard walk does not exhibit localization. As for the generalcase,that is,for the initial coinstate ϕ=T[α,β](α,β C, α2+ β 2 =1), we canobtainthe ∈ | | | | probability amplitudes and the time-averagedlimit measure by Lemma 1. Now, we should note that for the general initial coin state ϕ, we have n k 1 Ψ (0)= Ξ∗ ϕ 2n √2  2aj Xk=1(a1,...,aXk)∈(Z>)k: jY=1 a1+···+ak=n   1 n k ω k 1 1 k α = r∗ − √2  2aj−1 2 1 1 β Xk=1(a1,...,aXk)∈(Z>)k: jY=1 (cid:16) (cid:17) (cid:20) − − (cid:21) (cid:20) (cid:21) a1+···+ak=n   for n 1, where Z = 1,2,... and > ≥ { } ∞ 1 z2+√1+z4 r∗zn = − − . n z n=1 X Here we should note α iβ α+iβ k ( 1+i)k − +( 1 i)k 1 1 α − 2 − − 2 − = (cid:18) (cid:19) (cid:18) (cid:19) . 1 1 β α iβ α+iβ (cid:20) − − (cid:21) (cid:20) (cid:21) ( 1+i)ki − +( 1 i)k( i)  − 2 − − − 2   (cid:18) (cid:19) (cid:18) (cid:19)    Therefore, we obtain the concrete expression of Ψ (0) for the general initial coin state ϕ as follows. 2n 5 Lemma 3 For the initial coin state ϕ=T[α,β](α,β C, α2+ β 2 =1), we have ∈ | | | | n k k α iβ ω( 1+i) 1 Ψ (0)= − r∗ − 2n 2  2aj−1 2 i (cid:18) (cid:19)Xk=1(a1,...,aXk)∈(Z>)k: jY=1 (cid:18) (cid:19) (cid:20) (cid:21) a1+···+ak=n   n k k α+iβ ω( 1 i) 1 + r∗ − − , 2  2aj−1 2 i (cid:18) (cid:19)kX=1(a1,...,aXk)∈(Z>)k: jY=1 (cid:18) (cid:19) (cid:20) − (cid:21) a1+···+ak=n   for n 1. ≥ Thus,multiplyingtheresultsby(α ηiβ)/√2(η =1, 1)foreachinitialcoinstateϕ=ϕ(η)=T[1/√2,ηi/√2](η = − − 1, 1) and then summing the results each other gives the time-averagedlimit measure. By Lemma 3, we obtain the − following one of our main results. Theorem 1 1. For the initial coin state ϕ=T[α,β](α,β C, α2+ β 2 =1), we have ∈ | | | | 1 E 1 E Ψ(L,ℜ)(0) (α iβ) − + cos(nθ )I (φ)+(α+iβ) − − cos(nθ )I (φ), 2n ∼ − 3 2E 0 (1/4,1) 3 2E 0 (0,3/4) + − − − 1 E S C 1 E S+C Ψ(L,ℑ)(0) (α iβ) − + − sin(nθ )I (φ)+(α+iβ) − − sin(nθ )I (φ), 2n ∼ − 3 2E S C 0 (1/4,1) 3 2E S+C 0 (0,3/4) + − − | − | − | | 1 E S C 1 E S+C Ψ(R,ℜ)(0) (α iβ) − + − sin(nθ )I (φ).+(α+iβ) − − sin(nθ )I (φ), 2n ∼− − 3 2E S C 0 (1/4,1) 3 2E S+C 0 (0,3/4) + − − | − | − | | 1 E 1 E Ψ(R,ℑ)(0) (α iβ) − + cos(nθ )I (φ) (α+iβ) − − cos(nθ )I (φ), 2n ∼ − 3 2E 0 (1/4,1) − 3 2E 0 (0,3/4) + − − − for n 1, where ≥ (j,ℜ) (j) (j,ℑ) (j) Ψ (0) = Ψ (0) , Ψ (0)= Ψ (0) (j =L,R), 2n ℜ 2n 2n ℑ 2n (cid:16) (cid:17) (cid:16) (cid:17) E = C S =cos(2πφ) sin(2πφ), ± ± ± 2(1 E)2 (2 E)S C cosθ − , sinθ = − | − |. 0 0 − 3 2E 3 2E − − Here, we should note that (z) is the real part and (z) is the imaginary part of z (z C). Moreover, we have ℜ ℑ ∈ 6 2. r (0) 2n µ (0) = lim ∞ n→∞ 2 2 2 1 E 1 E = − + α iβ 2I (φ)+ − − α+iβ 2I (φ). 3 2E | − | (1/4,1) 3 2E | | (0,3/4) (cid:18) − +(cid:19) (cid:18) − −(cid:19) 2 2 1 √2C 1 √2C = − − α iβ 2I (φ)+ − + α+iβ 2I (φ), 3 2√2C−! | − | (1/4,1) 3 2√2C+! | | (0,3/4) − − where π 1 C =cos 2πφ = cos(2πφ) sin(2πφ) . ± ± 4 √2{ ∓ } (cid:16) (cid:17) The proof of Theorem 1 appears in Appendix A. 3.3 Main result 2: Time-averaged limit measure for general x Z ∈ Let us consider the Wojcik model starting at the origin with the initial coin state ϕ = T[α,β], where α,β C with ∈ α2+ β 2 =1. In this subsection, we present the time-averagedlimit measure µ (x)(x Z). | | | | ∞ ∈ Theorem 2 1. x=0. µ (0)=µ(1)(0)+µ(2)(0). ∞ 2. x=0. 6 1 |x| 1 |x| µ (x)=(2 √2C ) µ(1)(0)+(2 √2C ) µ(2)(0), ∞ − + 3 2√2C − − 3 2√2C (cid:18) − +(cid:19) (cid:18) − −(cid:19) where (1 √2C )2 µ(1)(0)= − + α+iβ 2I (φ), (3 2√2C )2| | (0,3/4) + − and (1 √2C )2 µ(2)(0)= − − α iβ 2I (φ), (3 2√2C )2| − | (1/4,1) − − with π 1 C =cos 2πφ+ = cos(2πφ) sin(2πφ) , + 4 √2{ − }  (cid:16) π(cid:17) 1  C− =cos 2πφ− 4 = √2{cos(2πφ)+sin(2πφ)}. (cid:16) (cid:17) We emphasize that the time-averagedlimit measure is symmetric for the originand localizationheavily depends on the choice of the initial coin state ϕ and parameter φ. For instance, when α = iβ and φ (3/4,1), we see ∈ µ (x) = 0(x Z) holds. When α= iβ and φ (0,1/4), we also have µ (x) =0 (x Z). Moreover, our results ∞ ∈ − ∈ ∞ ∈ imply that the stationary measure given by [2] stated below is a special case for the time-averaged limit measure. The proof of Theorem 2 is given in Sect. 6. Hereweconsidertherelationbetweenthetime-averagedandstationarymeasures. First,wepresentthestatioary measure for the Wojcik model in Theorem 2 of Ref. [2] as follows: 7 Theorem 3 Γ(φ) (x=0), µ(x)=kΨ(x)k2 =2|α|2|θs|2|x|× 1 (x=6 0), (cid:26) where 2 cos(2πφ) sin(2πφ) (β =iα), Γ(φ)= − − 2 cos(2πφ)+sin(2πφ) (β = iα), (cid:26) − − and 1 (β =iα), 3 2cos(2πφ) 2sin(2πφ) θ 2 = − − (6) | s|  1 (β = iα). 3 2cos(2πφ)+2sin(2πφ) −  − We should note that 1 1 = 3 2√2C 3 2cos(2πφ) sin(2πφ) − + − − and 1 1 = 3 2√2C 3 2cos(2πφ)+2sin(2πφ) − − − in Theorem 2 agree with θ 2: s | | 1 (β =iα), 3 2cos(2πφ) 2sin(2πφ) θ 2 = − − | s|  1 (β = iα). 3 2cos(2πφ)+2sin(2πφ) −  − From now on, we consider the two cases. When α = 1/√2 and β = i/√2 for instance, we have α+iβ = 0 and α iβ =√2. Then we have the time-averagedlimit measure as follows. − 2(1 √2C )2 − − I (φ) (x=0), (3 2√2C )2 (1/4,1) µ∞(x)= (2−−√2C−−) 3 21√2C |x|µ∞(0) (x6=0). (7) (cid:18) − −(cid:19) On the other hand, Theorem 2 in Ref. [2] gives the stationary measure as c2 (x=0), | | µ (x)= 1 |x| (8) ∞  (2 √2C )c2 (x=0). −  − | | 3 2√2C 6 (cid:18) − −(cid:19)  Equations (7) and (8) suggest that when c2 =2(1 √2C )2/(3 2√2C )2, then the time-averagedlimit measure − − | | − − coincides with the stationary measure. 8 Next, when α=1/√2 and β = i/√2, we have α+iβ =√2 and α iβ =0. Then we obtain the time-averaged − − limit measure as follows. 2(1 √2C )2 + − I (φ) (x=0), (3 2√2C )2 (0,3/4) µ∞(x)= (2−−√2C++) 3 21√2C |x|µ∞(0) (x6=0). (9) (cid:18) − +(cid:19) On the other hand, Theorem 2 in Ref. [2] gives the stationary measure. c2 (x=0), | | µ (x)= 1 |x| (10) ∞  (2 √2C )c2 (x=0). +  − | | (cid:18)3−2√2C+(cid:19) 6 Equations(9)and(10)suggestthatwhen c2 =2(1 √2C )2/(3 2√2C )2,then the time-averagedlimit measure + + | | − − also coincides with the stationary measure. 4 Result via the CGMV method We can derive the time-averaged limit measure at the origin µ (0) also from the CGMV method [15]. From now ∞ on, we use the same expressions as in Ref. [15]. Applying the CGMV method to the Wojcik model, we have i i 1 i a= e−2πφi, b= , w =e2πφi, ζ (b)= + . ± √2 √2 ±√2 √2 As condition , we see that the following inequality holds. + M 1 <φ<1. 4 On the other hand, as condition , we obtain − M 3 0<φ< . 4 Moreover,we have σ π σ =0, σ =π, σ =σ +σ =π, θ = = , 1 2 1 2 2 2 τ =2πφ, τ =2πφ+π, τ =τ +τ =4πφ+π, 1 2 1 2 and 1 1 C = (C+S), C = (C S), + − √2 √2 − with C =cos(2πφ), S =sin(2πφ). Conditions and imply 1/4<φ<1 and 0<φ<3/4, respectively. According to the CGMV method, we get + M M 1 ρ2 2 (αˆ 2 βˆ2) b+2ρ (ωαˆβˆ) lim P(0)(2n)= 1 a 1 | | −| | ℜ bℜ , (11) n→∞ α,β 2(cid:18) − |ζ±(b)−a|2(cid:19) ( ∓ √1−ℑ2b ) 9 whereP(0)(2n)istheprobabilitythatthewalkerreturntotheoriginattime2nwiththeinitialcoinstateϕ=T[α,β] α,β where α,β C and α2+ β 2 =1. Here, ∈ | | | | 1 1 1 ρ = , ζ (b) a2 = (3 2(C S))= (3 2√2C ), a ± ± √2 | − | 2 − ± 2 − 1 1 b= , b=0, ρ = , αˆ =λˆ(1)α=α, ℑ √2 ℜ b √2 0 βˆ=λˆ(2)β =ei((σ2−σ1)/2+τ2−σ2)β =eπi/2ω =iω. 1 Therefore, Eq. (11) becomes 2(1 √2C )2 − ± α iβ 2 (φ (1/4,1)), (3 2√2C )2| − | ∈ lim P(0)(2n)= − ± n→∞ α,β  2(1−√2C±)2 α+iβ 2 (φ (0,3/4)). (3 2√2C )2| | ∈ Thus, we obtain the time-averagedlimit measur−e µ (0)±as follows. ∞ 1 µ (0)= lim P(0)(2n) ∞ 2n→∞ α,β 2 2 1 √2C 1 √2C = − + α iβ 2I (φ)+ − − α+iβ 2I (φ), 3 2√2C+! | − | (1/4,1) 3 2√2C−! | | (0,3/4) − − which agrees with our result. 5 Proof of Lemma 1 At first, we consider the case of the Hadamard walk on Z = 0,1,2, starting at m( 1). ≥ { ···} ≥ 1 1 1 U =H = (x 0), (12) x √2 1 1 ≥ (cid:20) − (cid:21) which can be devided into P and Q as x x U =P +Q (x 1), (13) x x x ≥ where 1 1 1 1 0 0 P =P = , Q =Q= . x √2 0 0 x √2 1 1 (cid:20) (cid:21) (cid:20) − (cid:21) Next let Ξ(∞,m) be the sum of all the passages starting at m ( 1) and arrive at the origin at time n for the first n ≥ time. For instance, we have Ξ(∞,1) =P2QPQ+P3Q2. 5 Here we introduce R and S as 1 1 1 1 0 0 R= − , S = . √2 0 0 √2 1 1 (cid:20) (cid:21) (cid:20) (cid:21) 10

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.