Introduction
The splitting probability is a fundamental concept in the theory of stochastic processes and first-passage phenomena [3 ] [1, 2]. It measures the probability that the stochastic trajectory reaches one designated target before any of the other available targets, thus providing a natural way to characterize competition between multiple possible outcomes. A famous example arises in financial risk management, where a gambler tries to reach a target fortune before going bankrupt [4 ] . In this context, the splitting probability gives the likelihood of eventual success versus zero-fortune failure. Also, in transport problems, the transmission and reflection probabilities of particles through a medium can be modeled within the splitting probability framework [5 , 6 ] . Further applications can be found in chemical kinetics [7 , 8 ] , single-molecule experiments [9 , 10 ] and in studies of polymers [11 , 12 ] . These diverse applications have led to systematic investigations of the splitting problem across a broad spectrum of stochastic processes, such as random walk and diffusion [3 , 4 , 13 , 14 , 15 ] [1, 2], random acceleration [16 , 17 ] , active models [18 , 19 , 20 , 21 , 22 , 23 ] , resetting systems [24 , 25 , 26 ] , fractional Brownian motion [27 , 28 ] , and disordered systems [29 , 30 ] . We refer to standard textbooks [1, 2] for more details on the splitting problem. In this paper, we investigate a quantum analogue of this process.
As a concrete example of the classical problem, consider a discrete-space, continuous-time random walker hopping along a one-dimensional lattice chain of size N N N , with lattice sites labeled as x ∈ { 1 , 2 , … , N } x \in\{1, 2, \ldots, N\} x ∈ { 1 , 2 , … , N } . Starting from an initial position x 0 x_0 x 0 (with 1 ≤ x 0 ≤ N 1 \le x_0 \le N 1 ≤ x 0 ≤ N ), the walker can jump to one of its neighboring sites with a rate γ \gamma γ (> 0 > 0 > 0 ). We further have absorbing conditions at the left and the right boundaries, x L = 1 x_L = 1 x L = 1 and x R = N x_R = N x R = N , so that the walker is instantly absorbed and the process terminates upon reaching either boundary. In this set-up, the splitting probability characterizes the likelihood that the walker, starting from x 0 x_0 x 0 , will be absorbed by one boundary before it reaches the other one.
For this discrete-space, continuous-time random walker, the splitting probabilities, P L ( c l ) ( x 0 ) P_L^{(\mathrm{cl})}(x_0) P L ( cl ) ( x 0 ) and P R ( c l ) ( x 0 ) P_R^{(\mathrm{cl})}(x_0) P R ( cl ) ( x 0 ) , are given by the well-known gambler’s ruin formula [2]
P L ( c l ) ( x 0 ) = N − x 0 N − 1 , P R ( c l ) ( x 0 ) = 1 − P L ( c l ) ( x 0 ) . (1) P_L^{(\mathrm{cl})}(x_0)=\frac{N-x_0}{N-1},\qquad P_R^{(\mathrm{cl})}(x_0)=1-P_L^{(\mathrm{cl})}(x_0). \tag*{(1)} P L ( cl ) ( x 0 ) = N − 1 N − x 0 , P R ( cl ) ( x 0 ) = 1 − P L ( cl ) ( x 0 ) . (1)
The superscript ‘cl’ in our notation explicitly indicates the classical nature of the process, which will later be contrasted with the quantum case. These splitting probabilities are linear in x 0 x_0 x 0 , and are independent of the hopping rate γ \gamma γ . The fact that they satisfy P L ( c l ) ( x 0 ) + P R ( c l ) ( x 0 ) = 1 P_L^{(\mathrm{cl})}(x_0)+P_R^{(\mathrm{cl})}(x_0)=1 P L ( cl ) ( x 0 ) + P R ( cl ) ( x 0 ) = 1 implies that the walker will be always absorbed by one of the boundaries. Also, Eq. (1) implies the proximity effect whereby P L ( c l ) ( x 0 ) > P R ( c l ) ( x 0 ) P_L^{(\mathrm{cl})}(x_0)>P_R^{(\mathrm{cl})}(x_0) P L ( cl ) ( x 0 ) > P R ( cl ) ( x 0 ) if x 0 x_0 x 0 is closer to the left boundary and vice versa, so that starting closer to a boundary increases the probability of absorption there. Finally, we emphasize that the same expressions in Eq. (1) hold for the continuum diffusion process, as well as for the discrete-space, discrete-time random walk [5 ] [2].
Quantum walks are the analogues of classical random walks with applications in quantum information science and quantum computation [31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 ] . In a recent Letter [43 ] , some of us introduced the splitting problem for monitored quantum walks and showed that it differs qualitatively from its classical counterpart in Eq. (1) . More specifically, the study examined a continuous-time, discrete-space quantum walker evolving on a finite graph under a general time-independent, Hermitian Hamiltonian. In addition, projective measurements are performed at two distinct target states at regular time intervals τ \tau τ to monitor the walker in either target state. In contrast to classical processes, where the splitting problem, as explained above, is formulated in terms of absorbing boundaries, here it is defined through repeated projective measurements at the target states. Also τ \tau τ will be referred to as the sampling time throughout this work. For the one-dimensional lattice chain with N N N sites, with a Hamiltonian possessing reflection symmetry about the centre of the chain and localized measurements performed at the left and the right endpoints of the chain, the dual-target splitting problem can be mapped on to a pair of single-target first-detection problems. This showed that the splitting problem can be viewed as an interference effect between two single-target processes. The single-target detection problem has been quite extensively studied both theoretically [44 , 45 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 , 57 , 58 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67 , 68 ] as well as in experimental platforms involving quantum computers, optical waveguides and trapped ions [69 , 70 , 71 , 72 , 73 , 74 , 75 ] . The mapping therefore allowed us to use established single-target methods to analyze the dual-target splitting problem.
For the nearest-neighbor hopping Hamiltonian with localized initial condition, ∣ ψ 0 ⟩ = ∣ x 0 ⟩ \lvert\psi_0\rangle=\lvert x_0\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ , as shown in Figure 1 (a), it was demonstrated that, in the large-N N N limit, the splitting probabilities undergo a nonanalytic flat-to-fluctuating transition as a function of the sampling time. For 0 ≤ τ < τ c 0\leq\tau<\tau_c 0 ≤ τ < τ c , both splitting probabilities are equal to 1 / 2 1/2 1/2 , independent of x 0 x_0 x 0 and τ \tau τ , while for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c , they differ from 1 / 2 1/2 1/2 and develop a fluctuating pattern of peaks and dips that depend sensitively on both x 0 x_0 x 0 and τ \tau τ . The critical sampling time is τ c = 2 π / Δ E \tau_c=2\pi/\Delta E τ c = 2 π /Δ E in units of ℏ \hbar ℏ , where Δ E \Delta E Δ E is the difference between the maximum and minimum energy eigenvalues. For finite system, one gets resonances in the splitting probabilities at special values of τ \tau τ , related to a form of ergodicity breaking associated with dark states.
Figure 1. Panel (a) shows a chain of size N = 5 N = 5 N = 5 in which a system with ∣ ψ 0 ⟩ = ∣ 4 ⟩ |\psi_0\rangle= |4\rangle ∣ ψ 0 ⟩ = ∣4 ⟩ (see the pink shaded region) evolves with the nearest-neighbor hopping Hamiltonian in Eq. (2) . After every time interval τ \tau τ , projective measurements are performed at ∣ D 1 ⟩ = ∣ 1 ⟩ |D_1\rangle= |1\rangle ∣ D 1 ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ N ⟩ |D_2\rangle= |N\rangle ∣ D 2 ⟩ = ∣ N ⟩ to check the system at these states. Panel (b) illustrates the splitting probability detection protocol. During the first τ \tau τ interval, the system evolves unitarily with the operator U ( τ ) U(\tau) U ( τ ) in Eq. (3) . After this evolution, a projective measurement ∣ D 1 ⟩ ⟨ D 1 ∣ |D_1\rangle\langle D_1| ∣ D 1 ⟩ ⟨ D 1 ∣ is performed at the first target state. If this detection is successful, the process terminates. If not, then the state of the system is projected onto the subspace complementary to ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ , and a second projective measurement ∣ D 2 ⟩ ⟨ D 2 ∣ |D_2\rangle\langle D_2| ∣ D 2 ⟩ ⟨ D 2 ∣ is immediately made at the second target state. If the system is detected, the process again terminates. If this detection also fails, the system’s state is projected onto the subspace complementary to ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ . With this projected state, the system evolves unitarily for another interval τ \tau τ , and the procedure is repeated, in principle, for an infinite number of times or until a successful detection is made at either of the target states. We implement this detection protocol to study the splitting probabilities at the two targets.
In this paper, we extend the analysis on the splitting problem in monitored quantum systems and report several new results. We develop a renewal-theory formulation for the first-detection amplitudes at two arbitrary targets for a general Hamiltonian. This formulation connects the associated two detection amplitudes to quantities involving only the measurement-free unitary dynamics of the system. As a result, by combining renewal theory with the generating-function technique, one can treat the dual-target splitting problem entirely in terms of the measurement-free propagators. Renewal ideas have a long history in classical first-passage problems [76 ] , going back to Schrödinger [77 ] , and have more recently proved useful in quantum first-detection problems [52 , 53 ] . Here, we construct them for the quantum splitting problem.
We further investigate the dependence of the splitting probabilities on the initial position x 0 x_0 x 0 for the one-dimensional chain with nearest-neighbor hopping Hamiltonian. Apart from the flat-to-fluctuating transition at τ c \tau_c τ c , the splitting probabilities exhibit a notable spatial structure: for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c , exhibiting pronounced oscillations as functions of x 0 x_0 x 0 . Question is: what physical mechanism determines the characteristic spatial scales of these oscillations, and how do these scales depend on the sampling time τ \tau τ ? To address this question, we analyze the oscillations in the Fourier space of x 0 x_0 x 0 and show that they are governed by phase-matching events between different momentum modes of the quantum walk. Their wavenumbers depend systematically on the sampling time τ \tau τ , as shown in Eq. (13) below. Beyond localized initial conditions, we also consider ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ involving coherent superpositions over multiple sites. Does the flat-to-fluctuating transition at τ c \tau_c τ c persist for such coherent initial states, and if so, is it modified by the spatial separation and relative phase between different components of the initial state? As we show in the paper, the transition remains present, although some of its features differ from those found for localized initial conditions. Finally, we study how a finite number of measurements, as commonly encountered in experiments [72 , 73 , 74 ] , affect the splitting probabilities.
The rest of our paper is structured as follows: In Sec. II, we introduce our model and the splitting probability detection protocol. We outline the main results of our paper in § 2 . Sec. IV derives the quantum renewal equations. The renewal framework is then combined with the generating function techniques in § 4 to derive the splitting probabilities for arbitrary Hamiltonian, sampling time, initial condition and target states. For one-dimensional chain with Hamiltonians satisfying reflection symmetry and target states coinciding with the endpoints, we show in § 4 .A that these general expressions reduce to the mapping mentioned above. Sec. VI focuses on the nearest-neighbor hopping case. The Zeno limit is analyzed in Sec. VII, followed by a discussion on the oscillations in the splitting probabilities and the flat-to-fluctuating transition in § 7 . § 8 then derives the conditions for observing the resonances at special sampling times. We present detailed results for our case study, N = 5 N = 5 N = 5 , in § 9 . We discuss the role of coherent initial states in § 10 , and the effect of finite number of measurements in § 11 . Finally, we present our conclusions and indicate future directions in § 12 .
Model and preliminaries
Consider a discrete-state, continuous-time quantum system characterized by a time-independent, Hermitian Hamiltonian H H H , defined on a Hilbert space H \mathcal{H} H with a finite N N N states. The system is governed by the Schrödinger equation i ℏ ∣ ψ ˙ ⟩ = H ∣ ψ ⟩ i\hbar|\dot{\psi}\rangle= H|\psi\rangle i ℏ∣ ψ ˙ ⟩ = H ∣ ψ ⟩ with the initial condition ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ and overhead dot denoting the time derivative. For example, H H H could represent the nearest-neighbor hopping Hamiltonian on a finite one-dimensional chain of size N N N
H = − γ ∑ x = 1 N − 1 [ ∣ x ⟩ ⟨ x + 1 ∣ + ∣ x + 1 ⟩ ⟨ x ∣ ] , (2) H = -\gamma\sum_{x=1}^{N-1} \left[ |x\rangle\langle x+1| + |x+1\rangle\langle x| \right], \tag*{(2)} H = − γ x = 1 ∑ N − 1 [ ∣ x ⟩ ⟨ x + 1∣ + ∣ x + 1 ⟩ ⟨ x ∣ ] , (2)
where γ \gamma γ is the hopping amplitude; see Figure 1 a. Then H H H takes the form of a discrete Laplacian, which can be conveniently represented as the adjacency matrix of a graph [39 ] . We later return to this model. For now, we keep our analysis general.
In the absence of any measurement, the system evolves unitarily with the operator
U ( t ) = exp ( − i H t ) , (3) U(t) = \exp(-iHt), \tag*{(3)} U ( t ) = exp ( − i H t ) , (3)
where we set ℏ = 1 \hbar= 1 ℏ = 1 . When measurements are performed, this purely unitary evolution is however interrupted by wave function collapses at the measurement times. We now explain our detection protocol at two distinct target states.
Splitting probability detection protocol
Projective measurements are made on two distinct target states, ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , satisfying the orthogonality condition, ⟨ D 1 ∣ D 2 ⟩ = 0 \langle D_1 \vert D_2\rangle= 0 ⟨ D 1 ∣ D 2 ⟩ = 0 . The measurements are performed stroboscopically at times t = τ t = \tau t = τ , 2 τ 2\tau 2 τ , 3 τ 3\tau 3 τ and so on, such that the time interval τ \tau τ between successive measurements is kept fixed. As mentioned before, τ \tau τ is called the sampling time. Our detection protocol is the following:
Starting from ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ , we evolve the system for a time duration τ \tau τ as ∣ ψ ( τ ) ⟩ = U ( τ ) ∣ ψ 0 ⟩ \lvert\psi(\tau)\rangle= U(\tau)\lvert\psi_0\rangle ∣ ψ ( τ )⟩ = U ( τ ) ∣ ψ 0 ⟩ , where U ( τ ) U(\tau) U ( τ ) is the unitary operator in Eq. (3) .
At the end of this evolution, we perform a projective measurement ∣ D 1 ⟩ ⟨ D 1 ∣ \lvert D_1\rangle\langle D_1\rvert ∣ D 1 ⟩ ⟨ D 1 ∣ to check the system at the first target state. If detected, the process terminates with detection at this target. If no detection occurs, then the state ∣ ψ ( τ ) ⟩ \lvert\psi(\tau)\rangle ∣ ψ ( τ )⟩ is projected onto the subspace complementary to ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ as ∣ ψ ( τ + ) ⟩ = N ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) ∣ ψ ( τ ) ⟩ \lvert\psi(\tau^+)\rangle= \mathcal{N}\left(1-\lvert D_1\rangle\langle D_1\rvert\right)\lvert\psi(\tau)\rangle ∣ ψ ( τ + )⟩ = N ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) ∣ ψ ( τ )⟩ . Here 1 \mathbb{1} 1 is the identity matrix, τ + \tau^+ τ + is the instant immediately after the null measurement and N \mathcal{N} N is the normalization factor chosen such that ⟨ ψ ( τ + ) ∣ ψ ( τ + ) ⟩ = 1 \langle\psi(\tau^+)\vert\psi(\tau^+)\rangle= 1 ⟨ ψ ( τ + ) ∣ ψ ( τ + )⟩ = 1 .
With ∣ ψ ( τ + ) ⟩ \lvert\psi(\tau^+)\rangle ∣ ψ ( τ + )⟩ , a second projective measurement ∣ D 2 ⟩ ⟨ D 2 ∣ \lvert D_2\rangle\langle D_2\rvert ∣ D 2 ⟩ ⟨ D 2 ∣ is immediately performed at the second target state. If the system is detected there, the process terminates with detection at this target. If this detection also fails, then ∣ ψ ( τ + ) ⟩ \lvert\psi(\tau^+)\rangle ∣ ψ ( τ + )⟩ is projected onto the complementary subspace using ( 1 − ∣ D 2 ⟩ ⟨ D 2 ∣ ) \left(1-\lvert D_2\rangle\langle D_2\rvert\right) ( 1 − ∣ D 2 ⟩ ⟨ D 2 ∣ ) .
The new projected state is then evolved for another duration τ \tau τ , followed by the repetition of steps 2-4, in principle, for an infinite number of times or until a successful detection is made.
With this protocol, we investigate the quantum splitting probabilities corresponding to detection at the target states ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ . Figure 1 shows a schematic illustration of this protocol for the one-dimensional chain with the nearest neighbor hopping Hamiltonian in Eq. (2) . Figure 1 (a) shows the chain of size N = 5 N = 5 N = 5 , with initial condition ∣ ψ 0 ⟩ = ∣ 4 ⟩ \lvert\psi_0\rangle= \lvert4\rangle ∣ ψ 0 ⟩ = ∣ 4 ⟩ , and measurements at the endpoints of the chain, ∣ D 1 ⟩ = ∣ 1 ⟩ \lvert D_1\rangle= \lvert1\rangle ∣ D 1 ⟩ = ∣ 1 ⟩ and ∣ D 2 ⟩ = ∣ 5 ⟩ \lvert D_2\rangle= \lvert5\rangle ∣ D 2 ⟩ = ∣ 5 ⟩ . Figure 1 (b) illustrates the detection protocol: after each interval of unitary evolution U ( τ ) U(\tau) U ( τ ) , the measurement is first performed at the ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ state and, if unsuccessful, immediately at ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , before the next evolution cycle begins. Observe that in our setup, a projective measurement is first performed at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ , and subsequently at ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ . One may equally consider the opposite ordering, where measurement is first made at the ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ state, and then at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ . We will later show that the two orderings are completely equivalent due to the orthogonality of ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , and the fact that the two measurements are performed immediately one after the other.
More generally, let F n ( D 1 ) ( ψ 0 ) F_n^{(D_1)}(\psi_0) F n ( D 1 ) ( ψ 0 ) and F n ( D 2 ) ( ψ 0 ) F_n^{(D_2)}(\psi_0) F n ( D 2 ) ( ψ 0 ) be the first-detection probabilities corresponding to the ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ target states, respectively, on the n n n -th measurement attempt. The splitting probabilities are then given by
P D 1 ( ψ 0 ) = ∑ n = 1 ∞ F n ( D 1 ) ( ψ 0 ) , P D 2 ( ψ 0 ) = ∑ n = 1 ∞ F n ( D 2 ) ( ψ 0 ) . (4) P_{D_1}(\psi_0) = \sum_{n=1}^{\infty} F_n^{(D_1)}(\psi_0), \qquad P_{D_2}(\psi_0) = \sum_{n=1}^{\infty} F_n^{(D_2)}(\psi_0). \tag*{(4)} P D 1 ( ψ 0 ) = n = 1 ∑ ∞ F n ( D 1 ) ( ψ 0 ) , P D 2 ( ψ 0 ) = n = 1 ∑ ∞ F n ( D 2 ) ( ψ 0 ) . (4)
The sums run over all possible first-detection events, namely detection at the first measurement attempt, the second, the third, and so forth. We next summarize the main results of our paper.
Summary of the results
(1) Quantum renewal equation: We show that the probability F n ( D α ) ( ψ 0 ) F_n^{(D_\alpha)}(\psi_0) F n ( D α ) ( ψ 0 ) with α ∈ { 1 , 2 } \alpha\in\{1,2\} α ∈ { 1 , 2 } is given by
F n ( D α ) ( ψ 0 ) = ∣ ϕ n ( D α ) ( ψ 0 ) ∣ 2 , with ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ , (5) F_n^{(D_\alpha)}(\psi_0) = \left|\phi_n^{(D_\alpha)}(\psi_0)\right|^2, \quad\text{with}\ \phi_n^{(D_\alpha)}(\psi_0) = \langle D_\alpha\vert U(\tau)\left[(1-\hat{D})U(\tau)\right]^{n-1}\lvert\psi_0\rangle, \tag*{(5)} F n ( D α ) ( ψ 0 ) = ϕ n ( D α ) ( ψ 0 ) 2 , with ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ , (5)
where D ^ = [ ∣ D 1 ⟩ ⟨ D 1 ∣ + ∣ D 2 ⟩ ⟨ D 2 ∣ ] \hat{D} = [|D_{1}\rangle\langle D_{1}| + |D_{2}\rangle\langle D_{2}|] D ^ = [ ∣ D 1 ⟩ ⟨ D 1 ∣ + ∣ D 2 ⟩ ⟨ D 2 ∣ ] is the projection operator in the detector subspace and U ( τ ) U(\tau) U ( τ ) recall is the unitary operator in Eq. (3) . The quantity ϕ n ( D α ) ( ψ 0 ) \phi_{n}^{(D_{\alpha})}(\psi_{0}) ϕ n ( D α ) ( ψ 0 ) is the first-detection amplitude at the ∣ D α ⟩ |D_{\alpha}\rangle ∣ D α ⟩ target state at the n n n -th stroboscopic attempt. We prove that ϕ n ( D α ) ( ψ 0 ) \phi_{n}^{(D_{\alpha})}(\psi_{0}) ϕ n ( D α ) ( ψ 0 ) satisfies the relation
ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( n τ ) ∣ ψ 0 ⟩ − ∑ β = 1 2 ∑ l = 1 n − 1 ϕ l ( D β ) ( ψ 0 ) ⟨ D α ∣ U ( ( n − l ) τ ) ∣ D β ⟩ . (6) \phi_{n}^{(D_{\alpha})}(\psi_{0}) = \langle D_{\alpha}|U(n\tau)|\psi_{0}\rangle- \sum_{\beta=1}^{2}\sum_{l=1}^{n-1}\phi_{l}^{(D_{\beta})}(\psi_{0})\langle D_{\alpha}|U((n-l)\tau)|D_{\beta}\rangle. \tag*{(6)} ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( n τ ) ∣ ψ 0 ⟩ − β = 1 ∑ 2 l = 1 ∑ n − 1 ϕ l ( D β ) ( ψ 0 ) ⟨ D α ∣ U (( n − l ) τ ) ∣ D β ⟩ . (6)
For example, when n = 1 n = 1 n = 1 and n = 2 n = 2 n = 2 , Eq. (6) reduces to
n = 1 : ϕ 1 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) ∣ ψ 0 ⟩ , n = 2 : ϕ 2 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( 2 τ ) ∣ ψ 0 ⟩ − ϕ 1 ( D 1 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 1 ⟩ − ϕ 1 ( D 2 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 2 ⟩ . (7) \begin{aligned}
n = 1:\quad&\phi_{1}^{(D_{\alpha})}(\psi_{0}) = \langle D_{\alpha}|U(\tau)|\psi_{0}\rangle,\\
n = 2:\quad&\phi_{2}^{(D_{\alpha})}(\psi_{0}) = \langle D_{\alpha}|U(2\tau)|\psi_{0}\rangle- \phi_{1}^{(D_{1})}(\psi_{0})\langle D_{\alpha}|U(\tau)|D_{1}\rangle- \phi_{1}^{(D_{2})}(\psi_{0})\langle D_{\alpha}|U(\tau)|D_{2}\rangle. \tag*{(7)}
\end{aligned} n = 1 : n = 2 : ϕ 1 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) ∣ ψ 0 ⟩ , ϕ 2 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( 2 τ ) ∣ ψ 0 ⟩ − ϕ 1 ( D 1 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 1 ⟩ − ϕ 1 ( D 2 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 2 ⟩ . (7)
One can clearly observe the emerging renewal structure. For n = 2 n = 2 n = 2 , we see that the first-detection amplitude consists of three parts: (i) the direct propagation amplitude ⟨ D α ∣ U ( 2 τ ) ∣ ψ 0 ⟩ \langle D_{\alpha}|U(2\tau)|\psi_{0}\rangle ⟨ D α ∣ U ( 2 τ ) ∣ ψ 0 ⟩ from the initial state ∣ ψ 0 ⟩ |\psi_{0}\rangle ∣ ψ 0 ⟩ to ∣ D α ⟩ |D_{\alpha}\rangle ∣ D α ⟩ in time t = 2 τ t = 2\tau t = 2 τ , (ii) a correction term subtracting histories in which the system is detected in the target state ∣ D 1 ⟩ |D_{1}\rangle ∣ D 1 ⟩ at the earlier time duration τ \tau τ and subsequently evolves from ∣ D 1 ⟩ |D_{1}\rangle ∣ D 1 ⟩ to ∣ D α ⟩ |D_{\alpha}\rangle ∣ D α ⟩ during the remaining time duration τ \tau τ , and (iii) an analogous correction term associated with detection in the second target state ∣ D 2 ⟩ |D_{2}\rangle ∣ D 2 ⟩ at time τ \tau τ , followed by evolution from ∣ D 2 ⟩ |D_{2}\rangle ∣ D 2 ⟩ to ∣ D α ⟩ |D_{\alpha}\rangle ∣ D α ⟩ .
Eq. (6) is the quantum renewal equation mentioned in the Introduction. As discussed there, the renewal formalism has proved useful both in classical first-passage problems [76 , 77 ] and in quantum single-target detection problems [52 , 53 ] . Here, we use this approach and combine it with the generating function techniques to find the detection amplitudes ϕ n ( D α ) ( ψ 0 ) \phi_{n}^{(D_{\alpha})}(\psi_{0}) ϕ n ( D α ) ( ψ 0 ) . We then use Eq. (4) to compute the quantum splitting probabilities.
(2) Quantum mapping: For a one-dimensional lattice chain of size N N N , with a general Hamiltonian symmetric under the reflection transformation x → ( N + 1 − x ) x \to(N + 1-x) x → ( N + 1 − x ) , and with target states localized at the left and the right endpoints, ∣ D 1 ⟩ = ∣ x L ⟩ = ∣ 1 ⟩ |D_{1}\rangle= |x_{L}\rangle= |1\rangle ∣ D 1 ⟩ = ∣ x L ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ |D_{2}\rangle= |x_{R}\rangle= |N\rangle ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ , the first-detection amplitudes in Eq. (5) can be expressed as a linear combination of two terms χ n ( ± ) ( ψ 0 ) \chi_{n}^{(\pm)}(\psi_{0}) χ n ( ± ) ( ψ 0 )
ϕ n ( x L ) ( ψ 0 ) = 1 2 [ χ n ( + ) ( ψ 0 ) + χ n ( − ) ( ψ 0 ) ] , ϕ n ( x R ) ( ψ 0 ) = 1 2 [ χ n ( + ) ( ψ 0 ) − χ n ( − ) ( ψ 0 ) ] . (8) \begin{aligned}
\phi_{n}^{(x_{L})}(\psi_{0}) &= \frac{1}{\sqrt{2}}\left[\chi_{n}^{(+)}(\psi_{0})+\chi_{n}^{(-)}(\psi_{0})\right],\\
\phi_{n}^{(x_{R})}(\psi_{0}) &= \frac{1}{\sqrt{2}}\left[\chi_{n}^{(+)}(\psi_{0})-\chi_{n}^{(-)}(\psi_{0})\right]. \tag*{(8)}
\end{aligned} ϕ n ( x L ) ( ψ 0 ) ϕ n ( x R ) ( ψ 0 ) = 2 1 [ χ n ( + ) ( ψ 0 ) + χ n ( − ) ( ψ 0 ) ] , = 2 1 [ χ n ( + ) ( ψ 0 ) − χ n ( − ) ( ψ 0 ) ] . (8)
The auxiliary functions
χ n ( ± ) ( ψ 0 ) = ⟨ d ± ∣ U ( τ ) [ ( 1 − ∣ d ± ⟩ ⟨ d ± ∣ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ , (9) \chi_{n}^{(\pm)}(\psi_{0}) = \langle d_{\pm}|U(\tau)\left[(1-|d_{\pm}\rangle\langle d_{\pm}|)U(\tau)\right]^{n-1}|\psi_{0}\rangle, \tag*{(9)} χ n ( ± ) ( ψ 0 ) = ⟨ d ± ∣ U ( τ ) [ ( 1 − ∣ d ± ⟩ ⟨ d ± ∣ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ , (9)
represent two distinct single-target first-detection amplitudes, with projective measurements performed on the states ∣ d ± ⟩ = ( ∣ 1 ⟩ ± ∣ N ⟩ ) / 2 |d_{\pm}\rangle= (|1\rangle\pm|N\rangle)/\sqrt{2} ∣ d ± ⟩ = ( ∣1 ⟩ ± ∣ N ⟩) / 2 [50 , 53 ] . Eq. (8) is the mapping mentioned before. The dual-target splitting problem in Eq. (5) is mapped to two auxiliary single-target detection problems: one involving the auxiliary target state ∣ d + ⟩ |d_{+}\rangle ∣ d + ⟩ , and the other involving a second auxiliary target state ∣ d − ⟩ |d_{-}\rangle ∣ d − ⟩ . The mapping is intrinsically quantum, relying on auxiliary target states that arise from superposition and therefore have no classical counterpart. Moreover, Eq. (8) makes it possible to analyze the splitting dynamics using methods already developed for single-target detection problems [50 , 51 , 53 , 58 ] . In particular, we employ this mapping to obtain exact expressions for the splitting probabilities for arbitrary Hamiltonians possessing parity symmetry, arbitrary values of τ \tau τ and N N N , and an arbitrary initial state ∣ ψ 0 ⟩ |\psi_{0}\rangle ∣ ψ 0 ⟩ .
(3) Nearest-neighbor hopping Hamiltonian: We apply Eq. (8) to a quantum walk with nearest-neighbor hopping Hamiltonian in Eq. (2) such that H ∣ E ℓ ⟩ = E ℓ ∣ E ℓ ⟩ H|E_{\ell}\rangle= E_{\ell}|E_{\ell}\rangle H ∣ E ℓ ⟩ = E ℓ ∣ E ℓ ⟩ , with E ℓ E_{\ell} E ℓ and ∣ E ℓ ⟩ |E_{\ell}\rangle ∣ E ℓ ⟩ given in Eq. (55) (ℓ ∈ { 1 , 2 , … , N } \ell\in\{1, 2, \ldots, N\} ℓ ∈ { 1 , 2 , … , N } ). Whenever measurements are made at the left and the right endpoints of the chain, we denote the splitting probabilities by P L ( ψ 0 ) P_{L}(\psi_{0}) P L ( ψ 0 ) and P R ( ψ 0 ) P_{R}(\psi_{0}) P R ( ψ 0 ) corresponding to ∣ D 1 ⟩ = ∣ x L ⟩ = ∣ 1 ⟩ |D_{1}\rangle= |x_{L}\rangle= |1\rangle ∣ D 1 ⟩ = ∣ x L ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ |D_{2}\rangle= |x_{R}\rangle= |N\rangle ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ , respectively. For generic values of τ \tau τ , both P L ( ψ 0 ) P_{L}(\psi_{0}) P L ( ψ 0 ) and P R ( ψ 0 ) P_{R}(\psi_{0}) P R ( ψ 0 ) vary smoothly with τ \tau τ , satisfying the normalization condition P L ( ψ 0 ) + P R ( ψ 0 ) = 1 P_{L}(\psi_{0}) + P_{R}(\psi_{0}) = 1 P L ( ψ 0 ) + P R ( ψ 0 ) = 1 , just as the classical case in Eq. (1) . However, at certain special values of τ \tau τ , this smooth behavior breaks down and the splitting probabilities develop point-wise discontinuous jumps. At these resonant points, the normalization condition is violated, P L ( ψ 0 ) + P R ( ψ 0 ) < 1 P_{L}(\psi_{0}) + P_{R}(\psi_{0}) < 1 P L ( ψ 0 ) + P R ( ψ 0 ) < 1 , due to the formation of a dark state in the system. We prove that resonances occur when
( E ℓ − E ℓ ′ ) τ = 0 m o d 2 π and ⟨ ψ 0 ∣ D ⟩ ≠ 0 , (10) (E_{\ell}-E_{\ell'})\tau= 0 \ \mathrm{mod}\ 2\pi\quad\text{and} \quad\langle\psi_{0}|D\rangle\ne0, \tag*{(10)} ( E ℓ − E ℓ ′ ) τ = 0 mod 2 π and ⟨ ψ 0 ∣ D ⟩ = 0 , (10)
where ∣ E ℓ ⟩ |E_{\ell}\rangle ∣ E ℓ ⟩ and ∣ E ℓ ′ ⟩ |E_{\ell'}\rangle ∣ E ℓ ′ ⟩ are two distinct energy eigenstates that possess the same parity under reflection, and ∣ D ⟩ ∼ [ sin ( π ℓ / ( N + 1 ) ) ∣ E ℓ ′ ⟩ − sin ( π ℓ ′ / ( N + 1 ) ) ∣ E ℓ ⟩ ] |D\rangle\sim[\sin(\pi\ell/(N+1))|E_{\ell'}\rangle- \sin(\pi\ell'/(N+1))|E_{\ell}\rangle] ∣ D ⟩ ∼ [ sin ( π ℓ / ( N + 1 )) ∣ E ℓ ′ ⟩ − sin ( π ℓ ′ / ( N + 1 )) ∣ E ℓ ⟩] is the dark state mentioned above.
(4) Flat profile in the Zeno limit: In the regime of small but finite τ \tau τ , the splitting probabilities for H H H in Eq. (2) take a simple form determined solely by the projections of the initial state onto the two target states
P L ( ψ 0 ) ≃ 1 2 + 1 2 [ ∣ ⟨ x L ∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ x R ∣ ψ 0 ⟩ ∣ 2 ] , P R ( ψ 0 ) ≃ 1 2 − 1 2 [ ∣ ⟨ x L ∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ x R ∣ ψ 0 ⟩ ∣ 2 ] . (11) P_L(\psi_0) \simeq\frac{1}{2}+\frac{1}{2}\left[\lvert\langle x_L \mid\psi_0\rangle\rvert^2-\lvert\langle x_R \mid\psi_0\rangle\rvert^2\right],\qquad P_R(\psi_0) \simeq\frac{1}{2}-\frac{1}{2}\left[\lvert\langle x_L \mid\psi_0\rangle\rvert^2-\lvert\langle x_R \mid\psi_0\rangle\rvert^2\right]. \tag*{(11)} P L ( ψ 0 ) ≃ 2 1 + 2 1 [ ∣⟨ x L ∣ ψ 0 ⟩ ∣ 2 − ∣⟨ x R ∣ ψ 0 ⟩ ∣ 2 ] , P R ( ψ 0 ) ≃ 2 1 − 2 1 [ ∣⟨ x L ∣ ψ 0 ⟩ ∣ 2 − ∣⟨ x R ∣ ψ 0 ⟩ ∣ 2 ] . (11)
For example, if the initial condition is localized at a lattice site, ∣ ψ 0 ⟩ = ∣ x 0 ⟩ \lvert\psi_0\rangle=\lvert x_0\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ , x 0 ∈ { 1 , 2 , … , N } x_0\in\{1,2,\ldots,N\} x 0 ∈ { 1 , 2 , … , N } , Eq. (11) implies that P L ( x 0 ) ≃ P R ( x 0 ) ≃ 1 / 2 P_L(x_0)\simeq P_R(x_0)\simeq1/2 P L ( x 0 ) ≃ P R ( x 0 ) ≃ 1/2 for 1 < x 0 < N 1<x_0<N 1 < x 0 < N . Remarkably, for all initial positions distinct from the boundaries, both P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) are equal to 1 / 2 1/2 1/2 , independent of x 0 x_0 x 0 and N N N (see Figure 3 ). This flat profile is in sharp contrast to the classical random walk, where the splitting probabilities vary linearly with x 0 x_0 x 0 , as shown in Eq. (1) .
Figure 3. Splitting probability P L ( x 0 ) P_L(x_0) P L ( x 0 ) as a function of the initial position x 0 x_0 x 0 for the nearest-neighbor tight-binding chain of size N = 50 N=50 N = 50 in Eq. (2) . Panel (a) shows the theoretical prediction for the Zeno limit in Eq. (68) together with numerical simulations for γ = 1 \gamma=1 γ = 1 and τ = 0.1 \tau=0.1 τ = 0.1 ; see Appendix § 15 for details on the numerical simulations. Panels (b,c) display simulation data of P L ( x 0 ) P_L(x_0) P L ( x 0 ) for τ = 1 \tau=1 τ = 1 and τ = 2 \tau=2 τ = 2 , respectively, compared with the theoretical expression in Eq. (61) . These values of the sampling time are chosen such that they lie below and above the critical value γ τ c = π / 2 \gamma\tau_c=\pi/2 γ τ c = π /2 (≈ 1.57 \approx1.57 ≈ 1.57 ). As evident, P L ( x 0 ) P_L(x_0) P L ( x 0 ) for τ = 0.1 \tau=0.1 τ = 0.1 and τ = 1 \tau=1 τ = 1 remains essentially flat and equal to 1 / 2 1/2 1/2 , except close to the boundary sites. In contrast, for τ = 2 \tau=2 τ = 2 , P L ( x 0 ) P_L(x_0) P L ( x 0 ) develops pronounced oscillations as a function of x 0 x_0 x 0 . In the large-N N N limit, these oscillations are characterized by the wavenumber in Eq. (86) . The splitting probabilities are defined only for integer values of x 0 x_0 x 0 ; lines are included merely as a visual aid.
(5) Flat-to-fluctuating transition at τ c \tau_c τ c : In the limit of large system size, a phase-transition-like behavior emerges, in which both P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) undergo a sharp qualitative change as functions of τ \tau τ (see Figure 6 ). This nonanalytic transition, as was also pointed out in the Introduction, occurs at a critical sampling time
Figure 6. Illustration of the flat-to-fluctuating transition in P L ( x 0 ) P_L(x_0) P L ( x 0 ) at the critical sampling time γ τ c = π / 2 \gamma\tau_c = \pi/2 γ τ c = π /2 , using its expression in Eq. (61) for x 0 = 11 x_0 = 11 x 0 = 11 and x 0 = 30 x_0 = 30 x 0 = 30 . Each panel shows results for three different system sizes, illustrating that P L ( x 0 ) P_L(x_0) P L ( x 0 ) becomes essentially independent of N N N once the system is sufficiently large. In both panels, P L ( x 0 ) P_L(x_0) P L ( x 0 ) is essentially 1 / 2 1/2 1/2 for τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , independent of the initial position and the sampling time. The critical point γ τ c = π / 2 \gamma\tau_c = \pi/2 γ τ c = π /2 is indicated by the dashed vertical line. Beyond this point, for τ > τ c \tau> \tau_c τ > τ c , the flat behavior breaks down, and we find the emergence of a fluctuating regime, characterized by structures of pronounced peaks and dips. In this regime, P L ( x 0 ) P_L(x_0) P L ( x 0 ) deviates from 1 / 2 1/2 1/2 and the overall profile depends both on τ \tau τ and x 0 x_0 x 0 . For all plots, only the smooth part of the splitting probabilities is shown; the sharp resonant discontinuities found for special values of τ \tau τ are omitted for clarity (see Figure 9 for resonances). Since P L ( x 0 ) + P R ( x 0 ) = 1 P_L(x_0) + P_R(x_0) = 1 P L ( x 0 ) + P R ( x 0 ) = 1 , the same transition can also be found for P R ( x 0 ) P_R(x_0) P R ( x 0 ) .
Δ E τ c = 2 π , (12) \Delta E\,\tau_c=2\pi, \tag*{(12)} Δ E τ c = 2 π , (12)
with Δ E ≃ 4 γ \Delta E\simeq4\gamma Δ E ≃ 4 γ is the energy bandwidth for N ≫ 1 N\gg1 N ≫ 1 for H H H in Eq. (2) . In the regime 0 < τ < τ c 0<\tau<\tau_c 0 < τ < τ c , both splitting probabilities are essentially equal to 1 / 2 1/2 1/2 , independent of τ \tau τ . Above this critical sampling time, τ ≥ τ c \tau\geq\tau_c τ ≥ τ c , both P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) deviate from 1 / 2 1/2 1/2 and have a fluctuating dependence on τ \tau τ .
(6) Emergence of oscillations for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c : We also study the dependence of P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) on the initial position x 0 x_0 x 0 . For τ ≥ τ c \tau\geq\tau_c τ ≥ τ c , they exhibit pronounced oscillations as functions of x 0 x_0 x 0 (see Figure 3 ). In the large-N N N limit, the dominant wavenumber characterizing these oscillations is given by
Q max ( τ ) = 2 sin − 1 ( r τ c τ ) , r = 1 , 2 , … . (13) Q_{\max}(\tau)=2\sin^{-1}\left(\frac{r\tau_c}{\tau}\right),\qquad r=1,2,\ldots. \tag*{(13)} Q m a x ( τ ) = 2 sin − 1 ( τ r τ c ) , r = 1 , 2 , … . (13)
This equation gives real Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) only for τ ≥ r τ c \tau\geq r\tau_c τ ≥ r τ c , indicating that the minimum sampling time at which oscillations appear is τ c \tau_c τ c in Eq. (12) . Close to τ c \tau_c τ c , the wavenumber for r = 1 r=1 r = 1 scales as Q max ( τ ) ≃ π − 2 ( τ − τ c ) / τ c Q_{\max}(\tau)\simeq\pi-2\sqrt{(\tau-\tau_c)/\tau_c} Q m a x ( τ ) ≃ π − 2 ( τ − τ c ) / τ c . We show that the oscillations are governed by phase-matching events between different momentum modes of the quantum walker, which can occur only for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c .
(7) Coherent initial state: We consider the initial state ∣ ψ 0 ⟩ = ( ∣ x 0 ⟩ + e i Φ ∣ x 0 + d 0 ⟩ ) / 2 \lvert\psi_0\rangle=(\lvert x_0\rangle+e^{i\Phi}\lvert x_0+d_0\rangle)/\sqrt{2} ∣ ψ 0 ⟩ = (∣ x 0 ⟩ + e i Φ ∣ x 0 + d 0 ⟩) / 2 , and study the effect of the phase difference Φ \Phi Φ and the separation d 0 d_0 d 0 on the splitting probability. For the nearest-neighbor hopping Hamiltonian, we show
P L ( ψ 0 ) = { 1 / 2 + K d 0 ( x 0 , τ ) sin Φ , d 0 odd , 1 / 2 + K d 0 ( x 0 , τ ) cos Φ , d 0 even , (14) P_L(\psi_0)=
\begin{cases}
1/2+K_{d_0}(x_0,\tau)\sin\Phi, & d_0\ \text{odd},\\
1/2+K_{d_0}(x_0,\tau)\cos\Phi, & d_0\ \text{even},
\end{cases} \tag*{(14)} P L ( ψ 0 ) = { 1/2 + K d 0 ( x 0 , τ ) sin Φ , 1/2 + K d 0 ( x 0 , τ ) cos Φ , d 0 odd , d 0 even , (14)
where K d 0 ( x 0 , τ ) K_{d_0}(x_0,\tau) K d 0 ( x 0 , τ ) is given in Eq. (129) [similarly for P R ( ψ 0 ) P_R(\psi_0) P R ( ψ 0 ) ]. Thus, the splitting probability has sin Φ \sin\Phi sin Φ or cos Φ \cos\Phi cos Φ form, depending on whether the separation d 0 d_0 d 0 is odd or even. Interestingly, the separation d 0 d_0 d 0 also plays an important role in determining whether the flat-to-fluctuating transition persists for coherent initial states. As shown in Figure 10 , for even d 0 d_0 d 0 , the transition at τ c \tau_c τ c persists, whereas for odd d 0 d_0 d 0 , it becomes less pronounced, particularly for small separations. However, as d 0 d_0 d 0 increases, the flat-to-fluctuating transition is recovered also for odd d 0 d_0 d 0 .
Figure 10. Splitting probability P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) for the initial state ∣ ψ 0 ⟩ = ( ∣ x 0 ⟩ + e i Φ ∣ x 0 + d 0 ⟩ ) / 2 |\psi_0\rangle= (|x_0\rangle+ e^{i\Phi}|x_0+d_0\rangle)/\sqrt{2} ∣ ψ 0 ⟩ = ( ∣ x 0 ⟩ + e i Φ ∣ x 0 + d 0 ⟩) / 2 , calculated using Eqs. (61) –(63) . We have plotted P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) as a function of the relative phase Φ \Phi Φ for (a) N = 50 N=50 N = 50 , γ = 1 \gamma=1 γ = 1 , τ = 1 \tau=1 τ = 1 , x 0 = 10 x_0=10 x 0 = 10 , and (b) N = 50 N=50 N = 50 , γ = 1 \gamma=1 γ = 1 , τ = 1.2 \tau=1.2 τ = 1.2 , x 0 = 17 x_0=17 x 0 = 17 . In both panels, we have shown the results for three separations, d 0 = 1 , 2 , 3 d_0=1,2,3 d 0 = 1 , 2 , 3 . For odd d 0 d_0 d 0 , P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) in both panels exhibits the sin Φ \sin\Phi sin Φ dependence predicted by Eq. (128) . For even d 0 d_0 d 0 , by contrast, P L ( ψ 0 ) ≈ 1 / 2 P_L(\psi_0)\approx1/2 P L ( ψ 0 ) ≈ 1/2 , essentially independent of Φ \Phi Φ , since the prefactor K d 0 ( x 0 , τ ) K_{d_0}(x_0,\tau) K d 0 ( x 0 , τ ) multiplying the cos Φ \cos\Phi cos Φ term in Eq. (128) is vanishingly small, as discussed in the text. To see this more explicitly, in panel (c) we plot P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) as a function of γ τ \gamma\tau γ τ for Φ = 1 \Phi=1 Φ = 1 , N = 50 N=50 N = 50 , x 0 = 10 x_0=10 x 0 = 10 and three even separations, d 0 = 2 , 4 , 6 d_0=2,4,6 d 0 = 2 , 4 , 6 . In all cases, P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) remains equal to 1 / 2 1/2 1/2 , independent of τ \tau τ , for τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , where τ c \tau_c τ c is given in Eq. (12) . Thus, the flat behavior below the critical sampling time, previously found for localized initial conditions in Figure 6 , persists also for ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ with even separation d 0 d_0 d 0 . On the other hand, panel (d) shows the corresponding results for three odd separations, d 0 = 1 , 3 , 5 d_0=1,3,5 d 0 = 1 , 3 , 5 . In this case, P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) generally depends on τ \tau τ already for τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , with the strongest dependence observed for the smallest separation d 0 = 1 d_0=1 d 0 = 1 . As d 0 d_0 d 0 increases, this dependence becomes markedly weaker, and the flat-to-fluctuating behavior is recovered; see d 0 = 5 d_0=5 d 0 = 5 .
In what follows, we provide a rigorous derivation of all the results outlined above. We first formulate the monitored splitting problem in a general setting and then derive the corresponding renewal equations. We subsequently use this framework to obtain the conditions for the special sampling times and to analyze the nonanalytic transition.
We begin by deriving the quantum renewal relation in Eq. (6) for a general discrete-state, continuous-time quantum system with Hamiltonian H H H , general initial condition ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ and two arbitrary orthogonal detection states ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ .
First dual-target measurement at t = τ t=\tau t = τ
Starting from ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ , the system evolves with the unitary operator in Eq. (3) for the duration τ \tau τ . Before we perform the first measurement, the state of the system at time t = ( τ − ϵ ) t=(\tau-\epsilon) t = ( τ − ϵ ) with ϵ → 0 + \epsilon\to0^+ ϵ → 0 + is
∣ ψ ( τ − ) ⟩ = U ( τ ) ∣ ψ 0 ⟩ . (15) \lvert\psi(\tau^-)\rangle=U(\tau)\lvert\psi_0\rangle. \tag*{(15)} ∣ ψ ( τ − )⟩ = U ( τ ) ∣ ψ 0 ⟩ . (15)
Now the first measurement is performed at the ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ target state. The probability of detecting the system in this state is
G 1 ( D 1 ) ( ψ 0 ) = ∣ ⟨ D 1 ∣ ψ ( τ − ) ⟩ ∣ 2 = ∣ ⟨ D 1 ∣ U ( τ ) ∣ ψ 0 ⟩ ∣ 2 . (16) \mathcal{G}_{1}^{(D_1)}(\psi_0)=\left|\langle D_1\lvert\psi(\tau^-)\rangle\right|^2=\left|\langle D_1\lvert U(\tau)\lvert\psi_0\rangle\right|^2 . \tag*{(16)} G 1 ( D 1 ) ( ψ 0 ) = ⟨ D 1 ∣ ψ ( τ − )⟩ 2 = ∣ ⟨ D 1 ∣ U ( τ ) ∣ ψ 0 ⟩ ∣ 2 . (16)
We use G n ( D α ) ( ψ 0 ) \mathcal{G}_{n}^{(D_\alpha)}(\psi_0) G n ( D α ) ( ψ 0 ) with α ∈ { 1 , 2 } \alpha\in\{1,2\} α ∈ { 1 , 2 } to denote the conditional probability that the first successful detection occurs at target ∣ D α ⟩ \lvert D_\alpha\rangle ∣ D α ⟩ on the n n n th measurement attempt, given that no detection occurred at either target in the earlier measurements.
In Eq. (16) , a successful detection of ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ ends the process. However, if no detection occurs, then the projection postulate of quantum mechanics implies that the state ∣ ψ ( τ − ) ⟩ \lvert\psi(\tau^-)\rangle ∣ ψ ( τ − )⟩ is projected onto the subspace orthogonal to ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ [78 ] . The component of ∣ ψ ( τ − ) ⟩ \lvert\psi(\tau^-)\rangle ∣ ψ ( τ − )⟩ along ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ is removed. Thus, immediately after a failed detection of ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ , the state of the system at time t = ( τ + ϵ ) t=(\tau+\epsilon) t = ( τ + ϵ ) becomes
∣ ψ ( τ + ) ⟩ = N ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) ∣ ψ ( τ − ) ⟩ = N ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) U ( τ ) ∣ ψ 0 ⟩ , (17) \lvert\psi(\tau^+)\rangle=\mathcal{N}\left(1-\lvert D_1\rangle\langle D_1\rvert\right)\lvert\psi(\tau^-)\rangle=\mathcal{N}\left(1-\lvert D_1\rangle\langle D_1\rvert\right)U(\tau)\lvert\psi_0\rangle, \tag*{(17)} ∣ ψ ( τ + )⟩ = N ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) ∣ ψ ( τ − )⟩ = N ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) U ( τ ) ∣ ψ 0 ⟩ , (17)
where N = 1 / 1 − G 1 ( D 1 ) ( ψ 0 ) \mathcal{N}=1/\sqrt{1-\mathcal{G}_{1}^{(D_1)}(\psi_0)} N = 1/ 1 − G 1 ( D 1 ) ( ψ 0 ) gives the correct normalization ⟨ ψ ( τ + ) ∣ ψ ( τ + ) ⟩ = 1 \langle\psi(\tau^+)\lvert\psi(\tau^+)\rangle=1 ⟨ ψ ( τ + ) ∣ ψ ( τ + )⟩ = 1 . With ∣ ψ ( τ + ) ⟩ \lvert\psi(\tau^+)\rangle ∣ ψ ( τ + )⟩ , we next perform an immediate projective measurement of the second target state ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ . The probability of detection of this state is
G 1 ( D 2 ) ( ψ 0 ) = ∣ ⟨ D 2 ∣ ψ ( τ + ) ⟩ ∣ 2 = ∣ ⟨ D 2 ∣ ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) U ( τ ) ∣ ψ 0 ⟩ ∣ 2 1 − G 1 ( D 1 ) ( ψ 0 ) . (18) \mathcal{G}_{1}^{(D_2)}(\psi_0)=\left|\langle D_2\lvert\psi(\tau^+)\rangle\right|^2=\frac{\left|\langle D_2\lvert\left(1-\lvert D_1\rangle\langle D_1\rvert\right)U(\tau)\lvert\psi_0\rangle\right|^2}{1-\mathcal{G}_{1}^{(D_1)}(\psi_0)} . \tag*{(18)} G 1 ( D 2 ) ( ψ 0 ) = ⟨ D 2 ∣ ψ ( τ + )⟩ 2 = 1 − G 1 ( D 1 ) ( ψ 0 ) ∣ ⟨ D 2 ∣ ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) U ( τ ) ∣ ψ 0 ⟩ ∣ 2 . (18)
Remember that this expression represents a conditional probability, conditioned on the failure of the preceding detection attempt of ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ . Moreover, as the two target states are orthogonal, ⟨ D 2 ∣ D 1 ⟩ = 0 \langle D_2\lvert D_1\rangle=0 ⟨ D 2 ∣ D 1 ⟩ = 0 , Eq. (18) then becomes
G 1 ( D 2 ) ( ψ 0 ) = ∣ ⟨ D 2 ∣ U ( τ ) ∣ ψ 0 ⟩ ∣ 2 1 − G 1 ( D 1 ) ( ψ 0 ) . (19) \mathcal{G}_{1}^{(D_2)}(\psi_0)=\frac{\left|\langle D_2\lvert U(\tau)\lvert\psi_0\rangle\right|^2}{1-\mathcal{G}_{1}^{(D_1)}(\psi_0)} . \tag*{(19)} G 1 ( D 2 ) ( ψ 0 ) = 1 − G 1 ( D 1 ) ( ψ 0 ) ∣ ⟨ D 2 ∣ U ( τ ) ∣ ψ 0 ⟩ ∣ 2 . (19)
Once again if the detection of ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ is successful, we terminate the process. Otherwise the state ∣ ψ ( τ + ) ⟩ \lvert\psi(\tau^+)\rangle ∣ ψ ( τ + )⟩ is projected onto the subspace complementary to ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ . Thus, after null detections of both target states during the first measurement cycle, the state becomes
∣ ψ ( τ + + ) ⟩ = ( 1 − ∣ D 2 ⟩ ⟨ D 2 ∣ ) ∣ ψ ( τ + ) ⟩ 1 − G 1 ( D 2 ) ( ψ 0 ) = ( 1 − D ^ ) U ( τ ) ∣ ψ 0 ⟩ ( 1 − G 1 ( D 1 ) ( ψ 0 ) ) ( 1 − G 1 ( D 2 ) ( ψ 0 ) ) , (20) \lvert\psi(\tau^{++})\rangle=\frac{\left(1-\lvert D_2\rangle\langle D_2\rvert\right)\lvert\psi(\tau^+)\rangle}{\sqrt{1-\mathcal{G}_{1}^{(D_2)}(\psi_0)}}=\frac{\left(1-\hat{D}\right)U(\tau)\lvert\psi_0\rangle}{\sqrt{\left(1-\mathcal{G}_{1}^{(D_1)}(\psi_0)\right)\left(1-\mathcal{G}_{1}^{(D_2)}(\psi_0)\right)}} , \tag*{(20)} ∣ ψ ( τ ++ )⟩ = 1 − G 1 ( D 2 ) ( ψ 0 ) ( 1 − ∣ D 2 ⟩ ⟨ D 2 ∣ ) ∣ ψ ( τ + )⟩ = ( 1 − G 1 ( D 1 ) ( ψ 0 ) ) ( 1 − G 1 ( D 2 ) ( ψ 0 ) ) ( 1 − D ^ ) U ( τ ) ∣ ψ 0 ⟩ , (20)
where ‘+ + ++ + + ’ in ∣ ψ ( τ + + ) ⟩ \lvert\psi(\tau^{++})\rangle ∣ ψ ( τ ++ )⟩ denotes that two measurements were performed at time τ \tau τ . Also, the operator D ^ \hat{D} D ^ stands for the projection in the detector subspace, and is defined as
D ^ = ∣ D 1 ⟩ ⟨ D 1 ∣ + ∣ D 2 ⟩ ⟨ D 2 ∣ . (21) \hat{D}=\lvert D_1\rangle\langle D_1\rvert+\lvert D_2\rangle\langle D_2\rvert. \tag*{(21)} D ^ = ∣ D 1 ⟩ ⟨ D 1 ∣ + ∣ D 2 ⟩ ⟨ D 2 ∣ . (21)
Eqs. (17) and (20) describe the state of the system. We will soon turn to the first-detection probabilities.
Second dual-target measurement at t = 2 τ t=2\tau t = 2 τ
The measurement at the end of the second time interval t = 2 τ t=2\tau t = 2 τ also proceeds in the same way. Just before this measurement, the state of the system is ∣ ψ ( 2 τ − ) ⟩ = U ( τ ) ∣ ψ ( τ + + ) ⟩ \lvert\psi(2\tau^-)\rangle=U(\tau)\lvert\psi(\tau^{++})\rangle ∣ ψ ( 2 τ − )⟩ = U ( τ ) ∣ ψ ( τ ++ )⟩ . The conditional probability that the system is detected at the ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ target, given that it was not detected in either of the target states at the end of the first measurement time interval, is
G 2 ( D 1 ) ( ψ 0 ) = ∣ ⟨ D 1 ∣ U ( τ ) ( 1 − D ^ ) U ( τ ) ∣ ψ 0 ⟩ ∣ 2 ( 1 − G 1 ( D 1 ) ( ψ 0 ) ) ( 1 − G 1 ( D 2 ) ( ψ 0 ) ) . (22) \mathcal{G}_{2}^{(D_1)}(\psi_0)=\frac{\left|\langle D_1\lvert U(\tau)\left(1-\hat{D}\right)U(\tau)\lvert\psi_0\rangle\right|^2}{\left(1-\mathcal{G}_{1}^{(D_1)}(\psi_0)\right)\left(1-\mathcal{G}_{1}^{(D_2)}(\psi_0)\right)} . \tag*{(22)} G 2 ( D 1 ) ( ψ 0 ) = ( 1 − G 1 ( D 1 ) ( ψ 0 ) ) ( 1 − G 1 ( D 2 ) ( ψ 0 ) ) ⟨ D 1 ∣ U ( τ ) ( 1 − D ^ ) U ( τ ) ∣ ψ 0 ⟩ 2 . (22)
Similarly, the probability of detection of ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ at time t = 2 τ t=2\tau t = 2 τ , conditioned on the system not being detected in any of the target states before, is
G 2 ( D 2 ) ( ψ 0 ) = ∣ ⟨ D 2 ∣ U ( τ ) ( 1 − D ^ ) U ( τ ) ∣ ψ 0 ⟩ ∣ 2 ( 1 − G 1 ( D 1 ) ( ψ 0 ) ) ( 1 − G 2 ( D 1 ) ( ψ 0 ) ) ( 1 − G 1 ( D 2 ) ( ψ 0 ) ) . (23) \mathcal{G}^{(D_2)}_2(\psi_0)=\frac{\left|\langle D_2|U(\tau)\left(1-\hat{D}\right)U(\tau)|\psi_0\rangle\right|^2}{\left(1-\mathcal{G}^{(D_1)}_1(\psi_0)\right)\left(1-\mathcal{G}^{(D_1)}_2(\psi_0)\right)\left(1-\mathcal{G}^{(D_2)}_1(\psi_0)\right)}. \tag*{(23)} G 2 ( D 2 ) ( ψ 0 ) = ( 1 − G 1 ( D 1 ) ( ψ 0 ) ) ( 1 − G 2 ( D 1 ) ( ψ 0 ) ) ( 1 − G 1 ( D 2 ) ( ψ 0 ) ) ⟨ D 2 ∣ U ( τ ) ( 1 − D ^ ) U ( τ ) ∣ ψ 0 ⟩ 2 . (23)
Continuing this iteration for general n n n , we obtain the conditional probability G n ( D α ) ( ψ 0 ) \mathcal{G}^{(D_\alpha)}_n(\psi_0) G n ( D α ) ( ψ 0 ) (recall α ∈ { 1 , 2 } \alpha\in\{1,2\} α ∈ { 1 , 2 } ) that the system is first detected at ∣ D α ⟩ \lvert D_\alpha\rangle ∣ D α ⟩ at the n n n -th measurement attempt as
G n ( D 1 ) ( ψ 0 ) = ∣ ⟨ D 1 ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ ∣ 2 ∏ j = 1 n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] , (24) \mathcal{G}^{(D_1)}_n(\psi_0)=\frac{\left|\langle D_1|U(\tau)\left[\left(1-\hat{D}\right)U(\tau)\right]^{n-1}|\psi_0\rangle\right|^2}{\prod_{j=1}^{n-1}\left[\left(1-\mathcal{G}^{(D_1)}_j(\psi_0)\right)\left(1-\mathcal{G}^{(D_2)}_j(\psi_0)\right)\right]}, \tag*{(24)} G n ( D 1 ) ( ψ 0 ) = ∏ j = 1 n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] ⟨ D 1 ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ 2 , (24)
G n ( D 2 ) ( ψ 0 ) = ∣ ⟨ D 2 ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ ∣ 2 ( 1 − G n ( D 1 ) ( ψ 0 ) ) ∏ j = 1 n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] . (25) \mathcal{G}^{(D_2)}_n(\psi_0)=\frac{\left|\langle D_2|U(\tau)\left[\left(1-\hat{D}\right)U(\tau)\right]^{n-1}|\psi_0\rangle\right|^2}{\left(1-\mathcal{G}^{(D_1)}_n(\psi_0)\right)\prod_{j=1}^{n-1}\left[\left(1-\mathcal{G}^{(D_1)}_j(\psi_0)\right)\left(1-\mathcal{G}^{(D_2)}_j(\psi_0)\right)\right]}. \tag*{(25)} G n ( D 2 ) ( ψ 0 ) = ( 1 − G n ( D 1 ) ( ψ 0 ) ) ∏ j = 1 n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] ⟨ D 2 ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ 2 . (25)
We next use these conditional probabilities to obtain the splitting probabilities of the target states.
Quantum splitting probability
Remember that F n ( D 1 ) ( ψ 0 ) F^{(D_1)}_n(\psi_0) F n ( D 1 ) ( ψ 0 ) and F n ( D 2 ) ( ψ 0 ) F^{(D_2)}_n(\psi_0) F n ( D 2 ) ( ψ 0 ) are the first-detection probabilities of the ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ target states, respectively, on the n n n -th measurement attempt. To obtain them, we recall that both G n ( D 1 ) ( ψ 0 ) \mathcal{G}^{(D_1)}_n(\psi_0) G n ( D 1 ) ( ψ 0 ) and G n ( D 2 ) ( ψ 0 ) \mathcal{G}^{(D_2)}_n(\psi_0) G n ( D 2 ) ( ψ 0 ) in Eqs. (24) and (25) are the conditional detection probabilities, conditioned on the fact that no successful detection has occurred in any of the preceding ( n − 1 ) (n-1) ( n − 1 ) measurement attempts. The unconditional first-detection probabilities, F n ( D 1 ) ( ψ 0 ) F^{(D_1)}_n(\psi_0) F n ( D 1 ) ( ψ 0 ) and F n ( D 2 ) ( ψ 0 ) F^{(D_2)}_n(\psi_0) F n ( D 2 ) ( ψ 0 ) , can be obtained from the conditional probabilities by using the relation
F n ( D 1 ) ( ψ 0 ) = G n ( D 1 ) ( ψ 0 ) ∏ j = 1 n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] , F^{(D_1)}_n(\psi_0)=\mathcal{G}^{(D_1)}_n(\psi_0)\prod_{j=1}^{n-1}\left[\left(1-\mathcal{G}^{(D_1)}_j(\psi_0)\right)\left(1-\mathcal{G}^{(D_2)}_j(\psi_0)\right)\right], F n ( D 1 ) ( ψ 0 ) = G n ( D 1 ) ( ψ 0 ) j = 1 ∏ n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] ,
F n ( D 2 ) ( ψ 0 ) = G n ( D 2 ) ( ψ 0 ) ( 1 − G n ( D 1 ) ( ψ 0 ) ) ∏ j = 1 n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] . (26) F^{(D_2)}_n(\psi_0)=\mathcal{G}^{(D_2)}_n(\psi_0)\left(1-\mathcal{G}^{(D_1)}_n(\psi_0)\right)\prod_{j=1}^{n-1}\left[\left(1-\mathcal{G}^{(D_1)}_j(\psi_0)\right)\left(1-\mathcal{G}^{(D_2)}_j(\psi_0)\right)\right]. \tag*{(26)} F n ( D 2 ) ( ψ 0 ) = G n ( D 2 ) ( ψ 0 ) ( 1 − G n ( D 1 ) ( ψ 0 ) ) j = 1 ∏ n − 1 [ ( 1 − G j ( D 1 ) ( ψ 0 ) ) ( 1 − G j ( D 2 ) ( ψ 0 ) ) ] . (26)
The product term ∏ j = 1 n − 1 [ ⋯ ] \prod_{j=1}^{n-1}[\cdots] ∏ j = 1 n − 1 [ ⋯ ] in the first line gives the probability that both ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ were not detected in any of the earlier ( n − 1 ) (n-1) ( n − 1 ) measurements. To understand it, let us analyze our measurement process. Starting from ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ , we evolve the system for a time interval τ \tau τ . At the first attempt, n = 1 n=1 n = 1 , the probability of first detecting the system at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ is G 1 ( D 1 ) ( ψ 0 ) \mathcal{G}^{(D_1)}_1(\psi_0) G 1 ( D 1 ) ( ψ 0 ) given in Eq. (22) . In computer simulations, we generate a uniform random number. If this random number is smaller than G 1 ( D 1 ) ( ψ 0 ) \mathcal{G}^{(D_1)}_1(\psi_0) G 1 ( D 1 ) ( ψ 0 ) , the first detection occurs at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ . Otherwise, the measurement at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ fails with probability [ 1 − G 1 ( D 1 ) ( ψ 0 ) ] \left[1-\mathcal{G}^{(D_1)}_1(\psi_0)\right] [ 1 − G 1 ( D 1 ) ( ψ 0 ) ] . Then, a new random number is generated to test detection of ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ . This detection is successful with probability G 1 ( D 2 ) ( ψ 0 ) \mathcal{G}^{(D_2)}_1(\psi_0) G 1 ( D 2 ) ( ψ 0 ) . Otherwise both measurements at n = 1 n=1 n = 1 fail, the probability of which is [ 1 − G 1 ( D 1 ) ( ψ 0 ) ] [ 1 − G 1 ( D 2 ) ( ψ 0 ) ] \left[1-\mathcal{G}^{(D_1)}_1(\psi_0)\right]\left[1-\mathcal{G}^{(D_2)}_1(\psi_0)\right] [ 1 − G 1 ( D 1 ) ( ψ 0 ) ] [ 1 − G 1 ( D 2 ) ( ψ 0 ) ] . We then evolve the system for another τ \tau τ duration and proceed to the n = 2 n=2 n = 2 measurement at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ target. In this way, we continue the process until a successful measurement is recorded, which then defines the random first-detection event for this realization. A second realization is obtained by returning to the same initial state ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ and repeating the measurement protocol with a new set of random numbers. By performing the procedure over many independent realizations, we obtain the first-detection probabilities F n ( D 1 ) ( ψ 0 ) F^{(D_1)}_n(\psi_0) F n ( D 1 ) ( ψ 0 ) and F n ( D 2 ) ( ψ 0 ) F^{(D_2)}_n(\psi_0) F n ( D 2 ) ( ψ 0 ) . Combining Eq. (26) with Eqs. (24) –(25) gives us the first-detection probability at the n n n -th measurement in the ∣ D α ⟩ \lvert D_\alpha\rangle ∣ D α ⟩ target as
ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ , and F n ( D α ) ( ψ 0 ) = ∣ ϕ n ( D α ) ( ψ 0 ) ∣ 2 (27) \phi^{(D_\alpha)}_n(\psi_0)=\langle D_\alpha|U(\tau)\left[\left(1-\hat{D}\right)U(\tau)\right]^{n-1}|\psi_0\rangle,\quad\text{and}\quad F^{(D_\alpha)}_n(\psi_0)=\left|\phi^{(D_\alpha)}_n(\psi_0)\right|^2 \tag*{(27)} ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ , and F n ( D α ) ( ψ 0 ) = ϕ n ( D α ) ( ψ 0 ) 2 (27)
This result was outlined before in Eq. (5) . One can now plug Eq. (27) in Eq. (4) to get the splitting probabilities of detecting ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ or ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ target states.
Quantum renewal equation
Starting from Eq. (5) , we now derive the quantum renewal relation for the first-detection amplitude ϕ n ( D α ) ( ψ 0 ) \phi_n^{(D_\alpha)}(\psi_0) ϕ n ( D α ) ( ψ 0 ) in Eq. (6) . This renewal approach is not only instructive from a theoretical perspective, but also serves to simplify the solution, as we will demonstrate in the next section.
Let us look at ϕ n ( D α ) ( ψ 0 ) \phi_n^{(D_\alpha)}(\psi_0) ϕ n ( D α ) ( ψ 0 ) for the first few values of n n n . Using Eq. (5) for n = 1 n=1 n = 1 , we find
ϕ 1 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) ∣ ψ 0 ⟩ , \phi_1^{(D_\alpha)}(\psi_0)=\langle D_\alpha|U(\tau)|\psi_0\rangle, ϕ 1 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( τ ) ∣ ψ 0 ⟩ ,
and for n = 2 n=2 n = 2
ϕ 2 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( 2 τ ) ∣ ψ 0 ⟩ − ϕ 1 ( D 1 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 1 ⟩ − ϕ 1 ( D 2 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 2 ⟩ . \phi_2^{(D_\alpha)}(\psi_0)=\langle D_\alpha|U(2\tau)|\psi_0\rangle-\phi_1^{(D_1)}(\psi_0)\langle D_\alpha|U(\tau)|D_1\rangle-\phi_1^{(D_2)}(\psi_0)\langle D_\alpha|U(\tau)|D_2\rangle. ϕ 2 ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( 2 τ ) ∣ ψ 0 ⟩ − ϕ 1 ( D 1 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 1 ⟩ − ϕ 1 ( D 2 ) ( ψ 0 ) ⟨ D α ∣ U ( τ ) ∣ D 2 ⟩ .
One can clearly see an emerging renewal structure. In Appendix A, we use induction to show that the renewal structure is true even for general n n n
ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( n τ ) ∣ ψ 0 ⟩ − ∑ β = 1 2 ∑ l = 1 n − 1 ϕ l ( D β ) ( ψ 0 ) ⟨ D α ∣ U ( ( n − l ) τ ) ∣ D β ⟩ . (28) \phi_n^{(D_\alpha)}(\psi_0)=\langle D_\alpha|U(n\tau)|\psi_0\rangle-\sum_{\beta=1}^{2}\sum_{l=1}^{n-1}\phi_l^{(D_\beta)}(\psi_0)\langle D_\alpha|U((n-l)\tau)|D_\beta\rangle. \tag*{(28)} ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ U ( n τ ) ∣ ψ 0 ⟩ − β = 1 ∑ 2 l = 1 ∑ n − 1 ϕ l ( D β ) ( ψ 0 ) ⟨ D α ∣ U (( n − l ) τ ) ∣ D β ⟩ . (28)
This relation was quoted above in Eq. (6) . The amplitude ϕ n ( D α ) ( ψ 0 ) \phi_n^{(D_\alpha)}(\psi_0) ϕ n ( D α ) ( ψ 0 ) of detecting the system in ∣ D α ⟩ |D_\alpha\rangle ∣ D α ⟩ target at the n n n -th attempt consists of two contributions: the direct propagation amplitude ⟨ D α ∣ U ( n τ ) ∣ ψ 0 ⟩ \langle D_\alpha|U(n\tau)|\psi_0\rangle ⟨ D α ∣ U ( n τ ) ∣ ψ 0 ⟩ from the initial state ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ to ∣ D α ⟩ |D_\alpha\rangle ∣ D α ⟩ in time n τ n\tau n τ , and a set of correction terms subtracting all histories where the system was already detected earlier at ∣ D β ⟩ |D_\beta\rangle ∣ D β ⟩ (with β ∈ { 1 , 2 } \beta\in\{1,2\} β ∈ { 1 , 2 } ) at some earlier step l < n l<n l < n , and then propagated from ∣ D β ⟩ |D_\beta\rangle ∣ D β ⟩ to ∣ D α ⟩ |D_\alpha\rangle ∣ D α ⟩ in the remaining time ( n − l ) τ (n-l)\tau ( n − l ) τ . This recursive relation captures the essential renewal nature of the quantum splitting problem. Later, we will exploit its convolution form using generating-function techniques to calculate the splitting probabilities.
Splitting probability and the order of measurements
Recall that in our detection protocol explained in Sec. II II.A, after each time duration τ \tau τ , a projective measurement is first performed at ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ and, if no detection occurs, subsequently at ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ . One may equally consider the opposite ordering, where ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ is measured first and ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ second. The renewal equation (28) is, however, manifestly invariant under the interchange ∣ D 1 ⟩ ↔ ∣ D 2 ⟩ |D_1\rangle\leftrightarrow|D_2\rangle ∣ D 1 ⟩ ↔ ∣ D 2 ⟩ , indicating that it is independent of the measurement order. As mentioned before, this invariance is due to the orthogonality of ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ , and the fact that the two measurements are performed immediately one after the other.
Generating function
Given the convolution structure of ϕ n ( D α ) ( ψ 0 ) \phi_n^{(D_\alpha)}(\psi_0) ϕ n ( D α ) ( ψ 0 ) in Eq. (28) , one may solve it using the generating function approach. For this, we perform the Z-transform (discrete Laplace transform) as
ϕ ~ D α ( z ) = ∑ n = 1 ∞ z n ϕ n ( D α ) ( ψ 0 ) . (29) \widetilde{\phi}_{D_\alpha}(z)=\sum_{n=1}^{\infty}z^n\phi_n^{(D_\alpha)}(\psi_0). \tag*{(29)} ϕ D α ( z ) = n = 1 ∑ ∞ z n ϕ n ( D α ) ( ψ 0 ) . (29)
We have not explicitly shown the initial state dependence in the generating function ϕ ~ D α ( z ) \widetilde{\phi}_{D_\alpha}(z) ϕ D α ( z ) to keep the notation compact. Z-transforming Eq. (28) gives
\begin{pmatrix}
1+\widetilde{U}_{11}(z) & \widetilde{U}_{12}(z)\\
\widetilde{U}_{21}(z) & 1+\widetilde{U}_{22}(z)
\end{pmatrix}
\begin{pmatrix}
\widetilde{\phi}_{D_1}(z)\\
\widetilde{\phi}_{D_2}(z)
\end{pmatrix}
=
\begin{pmatrix}
\langle D_1|\widetilde{U}(z)|\psi_0\rangle\\
\langle D_2|\widetilde{U}(z)|\psi_0\rangle
\end{pmatrix}. \tag*{(30)}\end{pmatrix}
where we have used the notation
U ~ ( z ) = ∑ n = 1 ∞ z n U ( n τ ) = z e − i H τ 1 − z e − i H τ , (31) \widetilde{U}(z)=\sum_{n=1}^{\infty}z^nU(n\tau)=\frac{ze^{-iH\tau}}{1-ze^{-iH\tau}}, \tag*{(31)} U ( z ) = n = 1 ∑ ∞ z n U ( n τ ) = 1 − z e − i H τ z e − i H τ , (31)
U ~ α β ( z ) = ⟨ D α ∣ U ~ ( z ) ∣ D β ⟩ , with α , β ∈ { 1 , 2 } . (32) \widetilde{U}_{\alpha\beta}(z)=\langle D_\alpha|\widetilde{U}(z)|D_\beta\rangle,\quad\text{with }\alpha,\beta\in\{1,2\}. \tag*{(32)} U α β ( z ) = ⟨ D α ∣ U ( z ) ∣ D β ⟩ , with α , β ∈ { 1 , 2 } . (32)
The idea is to solve Eq. (30) to obtain generating functions ϕ ~ D 1 ( z ) \tilde{\phi}_{D_1}(z) ϕ ~ D 1 ( z ) and ϕ ~ D 2 ( z ) \tilde{\phi}_{D_2}(z) ϕ ~ D 2 ( z ) . The splitting probability P D α ( ψ 0 ) P_{D_\alpha}(\psi_0) P D α ( ψ 0 ) in Eq. (4) can be expressed in terms of these generating functions. To see this, we first write P D α ( ψ 0 ) P_{D_\alpha}(\psi_0) P D α ( ψ 0 ) as
P D α ( ψ 0 ) = ∑ n = 1 ∞ ∣ ϕ n ( D α ) ∣ 2 = ∑ n = 1 ∞ ∑ l = 1 ∞ ϕ n ( D α ) [ ϕ l ( D α ) ] ∗ δ n , l , (33) P_{D_\alpha}(\psi_0)=\sum_{n=1}^{\infty}\left|\phi_n^{(D_\alpha)}\right|^2=\sum_{n=1}^{\infty}\sum_{l=1}^{\infty}\phi_n^{(D_\alpha)}\left[\phi_l^{(D_\alpha)}\right]^*\delta_{n,l}, \tag*{(33)} P D α ( ψ 0 ) = n = 1 ∑ ∞ ϕ n ( D α ) 2 = n = 1 ∑ ∞ l = 1 ∑ ∞ ϕ n ( D α ) [ ϕ l ( D α ) ] ∗ δ n , l , (33)
where * stands for the complex conjugate and δ n , l \delta_{n,l} δ n , l is the Kronecker delta that takes value one if n = l n=l n = l , and zero otherwise. Using
δ n , l = 1 2 π ∫ 0 2 π d ω e i ω ( n − l ) , (34) \delta_{n,l}=\frac{1}{2\pi}\int_0^{2\pi}d\omega\,e^{i\omega(n-l)}, \tag*{(34)} δ n , l = 2 π 1 ∫ 0 2 π d ω e iω ( n − l ) , (34)
along with the definition of the generating function in Eq. (29) , one can simplify Eq. (33) to
P D α ( ψ 0 ) = 1 2 π ∫ 0 2 π d ω ∣ ϕ ~ D α ( e i ω ) ∣ 2 , with ϕ ~ D α ( z ) = [ 1 + U ~ α ˉ α ˉ ( z ) ] ⟨ D α ∣ U ~ ( z ) ∣ ψ 0 ⟩ − U ~ α ˉ α ( z ) ⟨ D α ˉ ∣ U ~ ( z ) ∣ ψ 0 ⟩ [ 1 + U ~ 11 ( z ) ] [ 1 + U ~ 22 ( z ) ] − U ~ 12 ( z ) U ~ 21 ( z ) , (35) \begin{aligned}
P_{D_\alpha}(\psi_0)&=\frac{1}{2\pi}\int_0^{2\pi}d\omega\,\left|\tilde{\phi}_{D_\alpha}\left(e^{i\omega}\right)\right|^2,\\
\text{with }\tilde{\phi}_{D_\alpha}(z)&=\frac{\left[1+\tilde{U}_{\bar{\alpha}\bar{\alpha}}(z)\right]\langle D_\alpha|\tilde{U}(z)|\psi_0\rangle-\tilde{U}_{\bar{\alpha}\alpha}(z)\langle D_{\bar{\alpha}}|\tilde{U}(z)|\psi_0\rangle}{\left[1+\tilde{U}_{11}(z)\right]\left[1+\tilde{U}_{22}(z)\right]-\tilde{U}_{12}(z)\tilde{U}_{21}(z)}, \tag*{(35)}
\end{aligned} P D α ( ψ 0 ) with ϕ ~ D α ( z ) = 2 π 1 ∫ 0 2 π d ω ϕ ~ D α ( e iω ) 2 , = [ 1 + U ~ 11 ( z ) ] [ 1 + U ~ 22 ( z ) ] − U ~ 12 ( z ) U ~ 21 ( z ) [ 1 + U ~ α ˉ α ˉ ( z ) ] ⟨ D α ∣ U ~ ( z ) ∣ ψ 0 ⟩ − U ~ α ˉ α ( z ) ⟨ D α ˉ ∣ U ~ ( z ) ∣ ψ 0 ⟩ , (35)
where recall α ˉ = 2 \bar{\alpha}=2 α ˉ = 2 if α = 1 \alpha=1 α = 1 and vice versa. Eq. (35) provides us the exact expression of the splitting probability for arbitrary Hamiltonian H H H , orthogonal target states ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ , ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ , the initial condition ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ and the sampling time τ \tau τ .
Mapping of the splitting problem to two separate single-target detection problems
In the summary section III, we noted that for a finite one-dimensional lattice chain whose Hamiltonian possesses reflection symmetry about the center of the chain, and with target states localized at the left and the right endpoints of the chain, the dual-target splitting problem can be mapped onto a pair of single-target first detection problems, each with its own auxiliary target. In the Letter [43 ] , we derived this mapping using an induction method. Here we provide an alternate derivation based on our renewal theory.
The reflection/parity operator is defined as
P ∣ x ⟩ = ∣ N + 1 − x ⟩ , (36) \mathcal{P}|x\rangle=|N+1-x\rangle, \tag*{(36)} P ∣ x ⟩ = ∣ N + 1 − x ⟩ , (36)
corresponding to reflection about the center of the chain. For Hamiltonians satisfying the parity symmetry, the commutation between H H H and P \mathcal{P} P vanishes
[ H , P ] = 0 , ⟹ P H P = H . (37) [H,\mathcal{P}]=0,\ \Longrightarrow\ \mathcal{P}H\mathcal{P}=H. \tag*{(37)} [ H , P ] = 0 , ⟹ P H P = H . (37)
To obtain the second relation, we have used the identity P − 1 = P \mathcal{P}^{-1}=\mathcal{P} P − 1 = P . Eq. (37) also implies P U ~ ( z ) P = U ~ ( z ) \mathcal{P}\tilde{U}(z)\mathcal{P}=\tilde{U}(z) P U ~ ( z ) P = U ~ ( z ) . We also fix the detection states localized at the endpoints of the chain, ∣ D 1 ⟩ = ∣ x L ⟩ = ∣ 1 ⟩ |D_1\rangle=|x_L\rangle=|1\rangle ∣ D 1 ⟩ = ∣ x L ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ |D_2\rangle=|x_R\rangle=|N\rangle ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ . Eq. (32) now gives
U ~ 11 ( z ) = ⟨ 1 ∣ U ~ ( z ) ∣ 1 ⟩ = ⟨ N ∣ P U ~ ( z ) P ∣ N ⟩ = U ~ 22 ( z ) , (38) \tilde{U}_{11}(z)=\langle1|\tilde{U}(z)|1\rangle=\langle N|\mathcal{P}\tilde{U}(z)\mathcal{P}|N\rangle=\tilde{U}_{22}(z), \tag*{(38)} U ~ 11 ( z ) = ⟨ 1∣ U ~ ( z ) ∣1 ⟩ = ⟨ N ∣ P U ~ ( z ) P ∣ N ⟩ = U ~ 22 ( z ) , (38)
U ~ 12 ( z ) = ⟨ 1 ∣ U ~ ( z ) ∣ N ⟩ = ⟨ N ∣ P U ~ ( z ) P ∣ 1 ⟩ = U ~ 21 ( z ) . (39) \tilde{U}_{12}(z)=\langle1|\tilde{U}(z)|N\rangle=\langle N|\mathcal{P}\tilde{U}(z)\mathcal{P}|1\rangle=\tilde{U}_{21}(z). \tag*{(39)} U ~ 12 ( z ) = ⟨ 1∣ U ~ ( z ) ∣ N ⟩ = ⟨ N ∣ P U ~ ( z ) P ∣1 ⟩ = U ~ 21 ( z ) . (39)
The generating functions in Eq. (35) then simplify to
ϕ ~ x L ( z ) = 1 2 [ ⟨ d + ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + U ~ 11 ( z ) + U ~ 12 ( z ) + ⟨ d − ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + U ~ 11 ( z ) − U ~ 12 ( z ) ] , (40) \tilde{\phi}_{x_L}(z)=\frac{1}{\sqrt{2}}\left[\frac{\langle d_+|\tilde{U}(z)|\psi_0\rangle}{1+\tilde{U}_{11}(z)+\tilde{U}_{12}(z)}+\frac{\langle d_-|\tilde{U}(z)|\psi_0\rangle}{1+\tilde{U}_{11}(z)-\tilde{U}_{12}(z)}\right], \tag*{(40)} ϕ ~ x L ( z ) = 2 1 [ 1 + U ~ 11 ( z ) + U ~ 12 ( z ) ⟨ d + ∣ U ~ ( z ) ∣ ψ 0 ⟩ + 1 + U ~ 11 ( z ) − U ~ 12 ( z ) ⟨ d − ∣ U ~ ( z ) ∣ ψ 0 ⟩ ] , (40)
ϕ ~ x R ( z ) = 1 2 [ ⟨ d + ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + U ~ 11 ( z ) + U ~ 12 ( z ) − ⟨ d − ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + U ~ 11 ( z ) − U ~ 12 ( z ) ] , (41) \tilde{\phi}_{x_R}(z)=\frac{1}{\sqrt{2}}\left[\frac{\langle d_+|\tilde{U}(z)|\psi_0\rangle}{1+\tilde{U}_{11}(z)+\tilde{U}_{12}(z)}-\frac{\langle d_-|\tilde{U}(z)|\psi_0\rangle}{1+\tilde{U}_{11}(z)-\tilde{U}_{12}(z)}\right], \tag*{(41)} ϕ ~ x R ( z ) = 2 1 [ 1 + U ~ 11 ( z ) + U ~ 12 ( z ) ⟨ d + ∣ U ~ ( z ) ∣ ψ 0 ⟩ − 1 + U ~ 11 ( z ) − U ~ 12 ( z ) ⟨ d − ∣ U ~ ( z ) ∣ ψ 0 ⟩ ] , (41)
where
∣ d ± ⟩ = ∣ 1 ⟩ ± ∣ N ⟩ 2 . (42) |d_\pm\rangle=\frac{|1\rangle\pm|N\rangle}{\sqrt{2}}. \tag*{(42)} ∣ d ± ⟩ = 2 ∣1 ⟩ ± ∣ N ⟩ . (42)
Notice that P ∣ d ± ⟩ = ± ∣ d ± ⟩ \mathcal{P}|d_{\pm}\rangle=\pm|d_{\pm}\rangle P ∣ d ± ⟩ = ± ∣ d ± ⟩ , which gives ⟨ d + ∣ U ~ ( z ) ∣ d − ⟩ = ⟨ d − ∣ U ~ ( z ) ∣ d + ⟩ = 0 \langle d_{+}|\widetilde{U}(z)|d_{-}\rangle=\langle d_{-}|\widetilde{U}(z)|d_{+}\rangle=0 ⟨ d + ∣ U ( z ) ∣ d − ⟩ = ⟨ d − ∣ U ( z ) ∣ d + ⟩ = 0 . With this result, we obtain
U ~ 11 ( z ) + U ~ 12 ( z ) = ⟨ d + ∣ U ~ ( z ) ∣ d + ⟩ , and U ~ 11 ( z ) − U ~ 12 ( z ) = ⟨ d − ∣ U ~ ( z ) ∣ d − ⟩ . (43) \widetilde{U}_{11}(z)+\widetilde{U}_{12}(z)=\langle d_{+}|\widetilde{U}(z)|d_{+}\rangle,\quad\text{and}\quad\widetilde{U}_{11}(z)-\widetilde{U}_{12}(z)=\langle d_{-}|\widetilde{U}(z)|d_{-}\rangle. \tag*{(43)} U 11 ( z ) + U 12 ( z ) = ⟨ d + ∣ U ( z ) ∣ d + ⟩ , and U 11 ( z ) − U 12 ( z ) = ⟨ d − ∣ U ( z ) ∣ d − ⟩ . (43)
These relations allow us to recast Eqs. (40) –(41) as
ϕ ~ x L ( z ) = 1 2 [ ⟨ d + ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + ⟨ d + ∣ U ~ ( z ) ∣ d + ⟩ + ⟨ d − ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + ⟨ d − ∣ U ~ ( z ) ∣ d − ⟩ ] , (44) \widetilde{\phi}_{x_L}(z)=\frac{1}{\sqrt{2}}\left[\frac{\langle d_{+}|\widetilde{U}(z)|\psi_0\rangle}{1+\langle d_{+}|\widetilde{U}(z)|d_{+}\rangle}+\frac{\langle d_{-}|\widetilde{U}(z)|\psi_0\rangle}{1+\langle d_{-}|\widetilde{U}(z)|d_{-}\rangle}\right], \tag*{(44)} ϕ x L ( z ) = 2 1 [ 1 + ⟨ d + ∣ U ( z ) ∣ d + ⟩ ⟨ d + ∣ U ( z ) ∣ ψ 0 ⟩ + 1 + ⟨ d − ∣ U ( z ) ∣ d − ⟩ ⟨ d − ∣ U ( z ) ∣ ψ 0 ⟩ ] , (44)
ϕ ~ x R ( z ) = 1 2 [ ⟨ d + ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + ⟨ d + ∣ U ~ ( z ) ∣ d + ⟩ − ⟨ d − ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + ⟨ d − ∣ U ~ ( z ) ∣ d − ⟩ ] . (45) \widetilde{\phi}_{x_R}(z)=\frac{1}{\sqrt{2}}\left[\frac{\langle d_{+}|\widetilde{U}(z)|\psi_0\rangle}{1+\langle d_{+}|\widetilde{U}(z)|d_{+}\rangle}-\frac{\langle d_{-}|\widetilde{U}(z)|\psi_0\rangle}{1+\langle d_{-}|\widetilde{U}(z)|d_{-}\rangle}\right]. \tag*{(45)} ϕ x R ( z ) = 2 1 [ 1 + ⟨ d + ∣ U ( z ) ∣ d + ⟩ ⟨ d + ∣ U ( z ) ∣ ψ 0 ⟩ − 1 + ⟨ d − ∣ U ( z ) ∣ d − ⟩ ⟨ d − ∣ U ( z ) ∣ ψ 0 ⟩ ] . (45)
Following earlier studies [53 , 58 ] , we identify the terms
χ ~ ± ( z ) = ⟨ d ± ∣ U ~ ( z ) ∣ ψ 0 ⟩ 1 + ⟨ d ± ∣ U ~ ( z ) ∣ d ± ⟩ , (46) \widetilde{\chi}_{\pm}(z)=\frac{\langle d_{\pm}|\widetilde{U}(z)|\psi_0\rangle}{1+\langle d_{\pm}|\widetilde{U}(z)|d_{\pm}\rangle}, \tag*{(46)} χ ± ( z ) = 1 + ⟨ d ± ∣ U ( z ) ∣ d ± ⟩ ⟨ d ± ∣ U ( z ) ∣ ψ 0 ⟩ , (46)
on the right hand side of Eqs. (44) and (45) as the generating functions associated with two different single-target detection problems, that of the first detection of the quantum walker on the target state ∣ d ± ⟩ |d_{\pm}\rangle ∣ d ± ⟩ . These single-target detection amplitudes in the n n n -domain are denoted by χ n ( ± ) ( ψ 0 ) \chi_n^{(\pm)}(\psi_0) χ n ( ± ) ( ψ 0 ) and are given by Eq. (9) [50 , 51 , 53 , 58 ] . Recasting Eqs. (44) and (45) in the n n n -domain, we obtain
ϕ n ( x L ) ( ψ 0 ) = 1 2 [ χ n ( + ) ( ψ 0 ) + χ n ( − ) ( ψ 0 ) ] , ϕ n ( x R ) ( ψ 0 ) = 1 2 [ χ n ( + ) ( ψ 0 ) − χ n ( − ) ( ψ 0 ) ] . (47) \begin{aligned}
\phi_n^{(x_L)}(\psi_0)&=\frac{1}{\sqrt{2}}\left[\chi_n^{(+)}(\psi_0)+\chi_n^{(-)}(\psi_0)\right],\\
\phi_n^{(x_R)}(\psi_0)&=\frac{1}{\sqrt{2}}\left[\chi_n^{(+)}(\psi_0)-\chi_n^{(-)}(\psi_0)\right]. \tag*{(47)}
\end{aligned} ϕ n ( x L ) ( ψ 0 ) ϕ n ( x R ) ( ψ 0 ) = 2 1 [ χ n ( + ) ( ψ 0 ) + χ n ( − ) ( ψ 0 ) ] , = 2 1 [ χ n ( + ) ( ψ 0 ) − χ n ( − ) ( ψ 0 ) ] . (47)
Eq. (47) is the mapping announced previously in Eq. (8) . Thus, the detection amplitudes for the splitting problem with two boundary targets can be expressed as linear combinations of the amplitudes χ n ( ± ) ( ψ 0 ) \chi_n^{(\pm)}(\psi_0) χ n ( ± ) ( ψ 0 ) , associated with two separate single-target problems: one involving detection at the auxiliary target state ∣ d + ⟩ |d_{+}\rangle ∣ d + ⟩ , and the other involving detection at another target state ∣ d − ⟩ |d_{-}\rangle ∣ d − ⟩ . A useful consequence of this mapping is that methods developed for the single-target first-detection problem can be applied directly to the present dual-target splitting problem. Moreover, this mapping is purely quantum and has no analogue in classical systems, since ∣ d ± ⟩ |d_{\pm}\rangle ∣ d ± ⟩ cannot be interpreted as physical states in the classical probabilistic sense. Indeed, the detection of a particle in a coherent superposition of the two detector states, with a relative phase between them, as in ∣ d ± ⟩ |d_{\pm}\rangle ∣ d ± ⟩ , has no classical analogue.
The mapping in Eq. (47) also provides a useful physical interpretation of the splitting process. The detection amplitudes ϕ n ( x L ) ( ψ 0 ) \phi_n^{(x_L)}(\psi_0) ϕ n ( x L ) ( ψ 0 ) and ϕ n ( x R ) ( ψ 0 ) \phi_n^{(x_R)}(\psi_0) ϕ n ( x R ) ( ψ 0 ) can be viewed as an interference effect between a pair of single-target detection processes. As we show later, this interference is responsible for several characteristic features of the splitting problem, including the flat-to-fluctuating transition and the emergence of spatial oscillations in the splitting probabilities above a critical sampling time.
Single-target detection probability
For later use, we evaluate the total detection probabilities associated with the auxiliary single-target detection problems. It can be expressed in terms of the generating functions as [52 ]
ρ ± ( ψ 0 ) = 1 2 π ∫ 0 2 π d ω ∣ χ ~ ± ( e i ω ) ∣ 2 . (48) \rho_{\pm}(\psi_0)=\frac{1}{2\pi}\int_{0}^{2\pi}d\omega\,\left|\widetilde{\chi}_{\pm}\left(e^{i\omega}\right)\right|^2. \tag*{(48)} ρ ± ( ψ 0 ) = 2 π 1 ∫ 0 2 π d ω χ ± ( e iω ) 2 . (48)
Unlike in the classical random walk, these probabilities can be less than one in the monitored quantum setting even for finite systems [52 , 58 ] . The integration in Eq. (48) was carried out in terms of the eigenstates of the unitary operator U ( τ ) U(\tau) U ( τ ) in Eq. (3) [58 ] . The eigenvalues of U ( τ ) U(\tau) U ( τ ) take the form e − i λ ℓ e^{-i\lambda_{\ell}} e − i λ ℓ , where
λ ℓ = E ℓ τ mod 2 π , (49) \lambda_{\ell}=E_{\ell}\tau\ \operatorname{mod} 2\pi, \tag*{(49)} λ ℓ = E ℓ τ mod 2 π , (49)
are the so-called quasienergies. Recall that H ∣ E ℓ ⟩ = E ℓ ∣ E ℓ ⟩ H|E_{\ell}\rangle=E_{\ell}|E_{\ell}\rangle H ∣ E ℓ ⟩ = E ℓ ∣ E ℓ ⟩ , with the index ℓ \ell ℓ labeling all energy eigenstates.
For generic sampling times, each distinct energy eigenvalue E ℓ E_\ell E ℓ produces a distinct phase factor, and therefore a distinct quasienergy λ ℓ \lambda_\ell λ ℓ . Therefore, the number of distinct quasienergies is equal to the number of distinct energy eigenvalues for most sampling times. At special values of τ \tau τ , however, two or more different energies { E ℓ , E ℓ ′ } \{E_\ell,E_{\ell'}\} { E ℓ , E ℓ ′ } can lead to the same phase factor, e − i E ℓ τ = e − i E ℓ ′ τ e^{-iE_\ell\tau}=e^{-iE_{\ell'}\tau} e − i E ℓ τ = e − i E ℓ ′ τ , and hence the same quasienergy. Consequently, the number of distinct quasienergies is reduced relative to the number of distinct energies. Let us assume that g ℓ g_\ell g ℓ is the degeneracy associated with the λ ℓ \lambda_\ell λ ℓ quasienergy and we denote these degenerate states as { ∣ E ℓ , m ⟩ } \{|E_{\ell,m}\rangle\} { ∣ E ℓ , m ⟩} , with m = 1 , 2 , … , g ℓ m=1,2,\ldots,g_\ell m = 1 , 2 , … , g ℓ . These states satisfy U ( τ ) ∣ E ℓ , m ⟩ = e − i E ℓ τ ∣ E ℓ , m ⟩ U(\tau)|E_{\ell,m}\rangle=e^{-iE_\ell\tau}|E_{\ell,m}\rangle U ( τ ) ∣ E ℓ , m ⟩ = e − i E ℓ τ ∣ E ℓ , m ⟩ . Then, the unitary operator can be expressed as
U ( τ ) = ∑ ℓ e − i λ ℓ L ℓ , with L ℓ = ∑ m = 1 g ℓ ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ . (50) U(\tau)=\sum_\ell e^{-i\lambda_\ell}L_\ell,\qquad\text{with}\quad L_\ell=\sum_{m=1}^{g_\ell}|E_{\ell,m}\rangle\langle E_{\ell,m}|. \tag*{(50)} U ( τ ) = ℓ ∑ e − i λ ℓ L ℓ , with L ℓ = m = 1 ∑ g ℓ ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣. (50)
Here, the summation runs over all distinct values of λ ℓ \lambda_\ell λ ℓ . Also, note that both the quasienergies λ ℓ \lambda_\ell λ ℓ , and the corresponding degeneracies g ℓ g_\ell g ℓ are, in general, functions of τ \tau τ , although we do not make this dependence explicit in the notation.
In the quasienergy basis, ρ ± ( ψ 0 ) \rho_\pm(\psi_0) ρ ± ( ψ 0 ) take the form [58 ]
ρ ± ( ψ 0 ) = ∑ ℓ ′ ∣ ∑ m = 1 g ℓ ⟨ d ± ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ ψ 0 ⟩ ∣ 2 ∑ m = 1 g ℓ ∣ ⟨ d ± ∣ E ℓ , m ⟩ ∣ 2 . (51) \rho_\pm(\psi_0)=\sum_\ell' \frac{\left|\sum_{m=1}^{g_\ell}\langle d_\pm|E_{\ell,m}\rangle\langle E_{\ell,m}|\psi_0\rangle\right|^2}{\sum_{m=1}^{g_\ell}|\langle d_\pm|E_{\ell,m}\rangle|^2}. \tag*{(51)} ρ ± ( ψ 0 ) = ℓ ∑ ′ ∑ m = 1 g ℓ ∣ ⟨ d ± ∣ E ℓ , m ⟩ ∣ 2 ∣ ∑ m = 1 g ℓ ⟨ d ± ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ ψ 0 ⟩ ∣ 2 . (51)
For ρ + ( ψ 0 ) \rho_+(\psi_0) ρ + ( ψ 0 ) , the primed summation is restricted only to those distinct quasienergy states that have a non-vanishing overlap with ∣ d + ⟩ |d_+\rangle ∣ d + ⟩ [and similarly for ρ − ( ψ 0 ) \rho_-(\psi_0) ρ − ( ψ 0 ) ].
Splitting probability
We now turn to the splitting probability for the quantum walker on a one-dimensional chain, with parity symmetric Hamiltonian and target states localized at the left and the right endpoints, ∣ D 1 ⟩ = ∣ x L ⟩ = ∣ 1 ⟩ |D_1\rangle=|x_L\rangle=|1\rangle ∣ D 1 ⟩ = ∣ x L ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ |D_2\rangle=|x_R\rangle=|N\rangle ∣ D 2 ⟩ = ∣ x R ⟩ = ∣ N ⟩ . As above, we denote these probabilities by P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) and P R ( ψ 0 ) P_R(\psi_0) P R ( ψ 0 ) , corresponding to the target states being the left and the right endpoints. Using our mapping (8) in Eq. (35) , we get
P L ( ψ 0 ) = 1 2 ρ + ( ψ 0 ) + 1 2 ρ − ( ψ 0 ) + ζ + − ( ψ 0 ) , P R ( ψ 0 ) = 1 2 ρ + ( ψ 0 ) + 1 2 ρ − ( ψ 0 ) − ζ + − ( ψ 0 ) . (52) \begin{aligned}
P_L(\psi_0)&=\frac{1}{2}\rho_+(\psi_0)+\frac{1}{2}\rho_-(\psi_0)+\zeta_{+-}(\psi_0),\\
P_R(\psi_0)&=\frac{1}{2}\rho_+(\psi_0)+\frac{1}{2}\rho_-(\psi_0)-\zeta_{+-}(\psi_0). \tag*{(52)}
\end{aligned} P L ( ψ 0 ) P R ( ψ 0 ) = 2 1 ρ + ( ψ 0 ) + 2 1 ρ − ( ψ 0 ) + ζ +− ( ψ 0 ) , = 2 1 ρ + ( ψ 0 ) + 2 1 ρ − ( ψ 0 ) − ζ +− ( ψ 0 ) . (52)
where ρ ± ( ψ 0 ) \rho_\pm(\psi_0) ρ ± ( ψ 0 ) are given in Eq. (51) and ζ + − ( ψ 0 ) \zeta_{+-}(\psi_0) ζ +− ( ψ 0 ) represents the interference effect between the two detection amplitudes associated with ∣ d + ⟩ |d_+\rangle ∣ d + ⟩ and ∣ d − ⟩ |d_-\rangle ∣ d − ⟩ . Its expression reads
ζ + − ( ψ 0 ) = Re [ 1 2 π ∫ 0 2 π d ω χ ~ + ( e i ω ) ( χ ~ − ( e i ω ) ) ∗ ] . (53) \zeta_{+-}(\psi_0)=\operatorname{Re}\left[\frac{1}{2\pi}\int_0^{2\pi}d\omega\,\widetilde{\chi}_+\left(e^{i\omega}\right)\left(\widetilde{\chi}_-\left(e^{i\omega}\right)\right)^*\right]. \tag*{(53)} ζ +− ( ψ 0 ) = Re [ 2 π 1 ∫ 0 2 π d ω χ + ( e iω ) ( χ − ( e iω ) ) ∗ ] . (53)
where χ ~ ± ( z ) \widetilde{\chi}_\pm(z) χ ± ( z ) are given in Eq. (46) . Also, Re [ ⋯ ] \operatorname{Re}[\,\cdots\,] Re [ ⋯ ] denotes the real part of [ ⋯ ] [\,\cdots\,] [ ⋯ ] , while ( ⋯ ) ∗ (\cdots)^* ( ⋯ ) ∗ indicates the complex conjugate of ( ⋯ ) (\cdots) ( ⋯ ) .
Eqs. (52) , together with Eqs. (51) and (53) , determine the splitting probabilities for any one-dimensional tight-binding Hamiltonian possessing reflection symmetry with initial condition ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ , and projective measurements performed at the boundary sites at regular stroboscopic time intervals τ \tau τ . We next examine these expressions for the nearest-neighbor hopping Hamiltonian.
Nearest-neighbor hopping Hamiltonian
We now apply our general results to the nearest-neighbor hopping Hamiltonian in Eq. (2) , which we rewrite here for continuity
H = − γ ∑ x = 1 N − 1 [ ∣ x ⟩ ⟨ x + 1 ∣ + ∣ x + 1 ⟩ ⟨ x ∣ ] , x ∈ { 1 , 2 , … , N } . (54) H=-\gamma\sum_{x=1}^{N-1}\left[|x\rangle\langle x+1|+|x+1\rangle\langle x|\right],\qquad x\in\{1,2,\ldots,N\}. \tag*{(54)} H = − γ x = 1 ∑ N − 1 [ ∣ x ⟩ ⟨ x + 1∣ + ∣ x + 1 ⟩ ⟨ x ∣ ] , x ∈ { 1 , 2 , … , N } . (54)
We briefly recall basic properties of this Hamiltonian. Observe that this Hamiltonian satisfies the parity symmetry, [ H , P ] = 0 [H,\mathcal{P}] = 0 [ H , P ] = 0 with P \mathcal{P} P in Eq. (36) . The eigenvalues and the eigenstates, denoted by E ℓ E_\ell E ℓ and ∣ E ℓ ⟩ |E_\ell\rangle ∣ E ℓ ⟩ for ℓ ∈ { 1 , 2 , … , N } \ell\in\{1,2,\ldots,N\} ℓ ∈ { 1 , 2 , … , N } , are
E ℓ = − 2 γ cos ( k ℓ ) , ∣ E ℓ ⟩ = 2 N + 1 ∑ x = 1 N sin ( k ℓ x ) ∣ x ⟩ , k ℓ = π ℓ N + 1 . (55) E_\ell= -2\gamma\cos(k_\ell), \qquad|E_\ell\rangle= \sqrt{\frac{2}{N+1}}\sum_{x=1}^{N}\sin(k_\ell x)|x\rangle, \qquad k_\ell= \frac{\pi\ell}{N+1}. \tag*{(55)} E ℓ = − 2 γ cos ( k ℓ ) , ∣ E ℓ ⟩ = N + 1 2 x = 1 ∑ N sin ( k ℓ x ) ∣ x ⟩ , k ℓ = N + 1 π ℓ . (55)
Equivalently, the site basis state can be expanded in the energy eigenstates as
∣ x ⟩ = 2 N + 1 ∑ ℓ = 1 N sin ( k ℓ x ) ∣ E ℓ ⟩ , (56) |x\rangle= \sqrt{\frac{2}{N+1}}\sum_{\ell=1}^{N}\sin(k_\ell x)|E_\ell\rangle, \tag*{(56)} ∣ x ⟩ = N + 1 2 ℓ = 1 ∑ N sin ( k ℓ x ) ∣ E ℓ ⟩ , (56)
while the auxiliary single-target states are
∣ d + ⟩ = 2 N + 1 ∑ ℓ = 1 ℓ o d d N sin ( k ℓ x ) ∣ E ℓ ⟩ , (57) |d_+\rangle= \frac{2}{\sqrt{N+1}}\sum_{\substack{\ell=1 \\ \ell\ \mathrm{odd}}}^{N}\sin(k_\ell x)|E_\ell\rangle, \tag*{(57)} ∣ d + ⟩ = N + 1 2 ℓ = 1 ℓ odd ∑ N sin ( k ℓ x ) ∣ E ℓ ⟩ , (57)
∣ d − ⟩ = 2 N + 1 ∑ ℓ = 1 ℓ e v e n N sin ( k ℓ x ) ∣ E ℓ ⟩ . (58) |d_-\rangle= \frac{2}{\sqrt{N+1}}\sum_{\substack{\ell=1 \\ \ell\ \mathrm{even}}}^{N}\sin(k_\ell x)|E_\ell\rangle. \tag*{(58)} ∣ d − ⟩ = N + 1 2 ℓ = 1 ℓ even ∑ N sin ( k ℓ x ) ∣ E ℓ ⟩ . (58)
The summation in the first line involves only odd values of ℓ \ell ℓ , while in the second line, it only involves even values. We next use our result in Eqs. (52) to compute the splitting probabilities of this model.
Generic sampling times and non-degenerate λ ℓ \lambda_\ell λ ℓ
For generic sampling times, we explained before that the number of distinct quasienergies is equal to the number of distinct energies. For H H H in Eq. (54) , all energies in Eq. (55) are non-degenerate, and consequently, all N N N quasienergies are distinct. For example, in Figure 2 (a), we have shown that for N = 5 N = 5 N = 5 , τ = 1 \tau= 1 τ = 1 and γ = 1 \gamma= 1 γ = 1 , all five phase factors are distinct, yielding five distinct quasienergies. At special values of τ \tau τ , however, two or more different energies { E ℓ , E ℓ ′ } \{E_\ell,E_{\ell'}\} { E ℓ , E ℓ ′ } can lead to the same phase factor, e − i E ℓ τ = e − i E ℓ ′ τ e^{-iE_\ell\tau} = e^{-iE_{\ell'}\tau} e − i E ℓ τ = e − i E ℓ ′ τ , and hence the same quasienergy. This is illustrated again in Figure 2 (b) for γ = 1 \gamma= 1 γ = 1 and τ = 2 π / 3 \tau= 2\pi/\sqrt{3} τ = 2 π / 3 , where three phase factors merge, leaving only three distinct quasienergies. For such special sampling times, we therefore obtain a degeneracy in the quasienergy spectrum. We first consider the nondegenerate case and return to the degenerate case later.
Figure 2. The nearest-neighbor hopping Hamiltonian in Eq. (54) , with N = 5 N = 5 N = 5 , has five eigenvalues given in Eq. (55) , namely { E 1 = − 3 γ , E 2 = − γ , E 3 = 0 , E 4 = γ , E 5 = 3 γ } \{E_1 = -\sqrt{3}\gamma, E_2 = -\gamma, E_3 = 0, E_4 = \gamma, E_5 = \sqrt{3}\gamma\} { E 1 = − 3 γ , E 2 = − γ , E 3 = 0 , E 4 = γ , E 5 = 3 γ } . These eigenvalues generate five distinct phases { e − i E ℓ τ } \{e^{-iE_\ell\tau}\} { e − i E ℓ τ } shown in red in panel (a) for γ = 1 \gamma= 1 γ = 1 and τ = 1 \tau= 1 τ = 1 . Consequently, we get five distinct quasienergies for this case. By contrast, for γ = 1 \gamma= 1 γ = 1 and τ = 2 π / 3 \tau= 2\pi/\sqrt{3} τ = 2 π / 3 , we have merging of three phases, e − i E 1 τ = e − i E 3 τ = e − i E 5 τ = 1 e^{-iE_1\tau} = e^{-iE_3\tau} = e^{-iE_5\tau} = 1 e − i E 1 τ = e − i E 3 τ = e − i E 5 τ = 1 , indicated by the arrow in panel (b). As a result, there are only three distinct quasienergies for τ = 2 π / 3 \tau= 2\pi/\sqrt{3} τ = 2 π / 3 . As explained in the paper, the merging of phases is responsible for the discontinuous-jumps in the splitting probabilities (see Figure 9 ) and gives rise to the resonance condition in Eq. (10) .
For generic τ \tau τ , the quasienergy degeneracy factor appearing in the expressions of ρ ± ( x 0 ) \rho_{\pm}(x_0) ρ ± ( x 0 ) in Eq. (51) is unity, i.e., g ℓ = 1 g_{\ell}=1 g ℓ = 1 for all ℓ \ell ℓ . So the quasienergy state ∣ E ℓ , 1 ⟩ \lvert E_{\ell,1}\rangle ∣ E ℓ , 1 ⟩ can be simply written as ∣ E ℓ ⟩ \lvert E_{\ell}\rangle ∣ E ℓ ⟩ . This yields
ρ ± ( ψ 0 ) = ∑ ℓ = 1 N ′ ∣ ⟨ d ± ∣ E ℓ ⟩ ⟨ E ℓ ∣ ψ 0 ⟩ ∣ 2 ∣ ⟨ d ± ∣ E ℓ ⟩ ∣ 2 . (59) \rho_{\pm}(\psi_0)=\sum_{\ell=1}^{N}{}'\frac{\left|\langle d_{\pm}\vert E_{\ell}\rangle\langle E_{\ell}\vert\psi_0\rangle\right|^2}{\left|\langle d_{\pm}\vert E_{\ell}\rangle\right|^2}. \tag*{(59)} ρ ± ( ψ 0 ) = ℓ = 1 ∑ N ′ ∣ ⟨ d ± ∣ E ℓ ⟩ ∣ 2 ∣ ⟨ d ± ∣ E ℓ ⟩ ⟨ E ℓ ∣ ψ 0 ⟩ ∣ 2 . (59)
Recall that the primed summation is restricted to only those values of ℓ \ell ℓ for which the overlap ⟨ d ± ∣ E ℓ ⟩ \langle d_{\pm}\vert E_{\ell}\rangle ⟨ d ± ∣ E ℓ ⟩ is non-zero. Using Eqs. (57) and (58) , we note that ⟨ d + ∣ E ℓ ⟩ = 0 \langle d_{+}\vert E_{\ell}\rangle=0 ⟨ d + ∣ E ℓ ⟩ = 0 when ℓ \ell ℓ is even, while ⟨ d − ∣ E ℓ ⟩ = 0 \langle d_{-}\vert E_{\ell}\rangle=0 ⟨ d − ∣ E ℓ ⟩ = 0 when ℓ \ell ℓ is odd. Eq. (59) then becomes
ρ + ( ψ 0 ) = ∑ ℓ = 1 ℓ odd N ∣ ⟨ E ℓ ∣ ψ 0 ⟩ ∣ 2 , ρ − ( ψ 0 ) = ∑ ℓ = 1 ℓ even N ∣ ⟨ E ℓ ∣ ψ 0 ⟩ ∣ 2 . (60) \rho_{+}(\psi_0)=\sum_{\substack{\ell=1\\ \ell\ \text{odd}}}^{N}\left|\langle E_{\ell}\vert\psi_0\rangle\right|^2,\qquad
\rho_{-}(\psi_0)=\sum_{\substack{\ell=1\\ \ell\ \text{even}}}^{N}\left|\langle E_{\ell}\vert\psi_0\rangle\right|^2. \tag*{(60)} ρ + ( ψ 0 ) = ℓ = 1 ℓ odd ∑ N ∣ ⟨ E ℓ ∣ ψ 0 ⟩ ∣ 2 , ρ − ( ψ 0 ) = ℓ = 1 ℓ even ∑ N ∣ ⟨ E ℓ ∣ ψ 0 ⟩ ∣ 2 . (60)
Plugging them in Eq. (52)
P L ( ψ 0 ) = 1 2 + ζ + − ( ψ 0 ) , P R ( ψ 0 ) = 1 2 − ζ + − ( ψ 0 ) , (61) \begin{aligned}
P_L(\psi_0)&=\frac{1}{2}+\zeta_{+-}(\psi_0),\\
P_R(\psi_0)&=\frac{1}{2}-\zeta_{+-}(\psi_0), \tag*{(61)}
\end{aligned} P L ( ψ 0 ) P R ( ψ 0 ) = 2 1 + ζ +− ( ψ 0 ) , = 2 1 − ζ +− ( ψ 0 ) , (61)
where recall that ζ + − ( ψ 0 ) \zeta_{+-}(\psi_0) ζ +− ( ψ 0 ) represents the interference effect between the two auxiliary single-target detection processes. Using Eq. (53) , it can be written as
ζ + − ( ψ 0 ) = Re [ ∮ ∣ z ∣ = 1 I ( ψ 0 , z ) d z ] , (62) \zeta_{+-}(\psi_0)=\operatorname{Re}\left[\oint_{\lvert z\rvert=1}\mathcal{I}(\psi_0,z)\,\mathrm{d}z\right], \tag*{(62)} ζ +− ( ψ 0 ) = Re [ ∮ ∣ z ∣ = 1 I ( ψ 0 , z ) d z ] , (62)
involving the real part of a contour integral of I ( ψ 0 , z ) \mathcal{I}(\psi_0,z) I ( ψ 0 , z ) around the unit circle in the complex plane. The integrand I ( ψ 0 , z ) = χ + ( z ) χ − ( 1 / z ) ∗ / 2 π i z \mathcal{I}(\psi_0,z)=\chi_{+}(z)\chi_{-}(1/z)^*/2\pi i z I ( ψ 0 , z ) = χ + ( z ) χ − ( 1/ z ) ∗ /2 π i z can be explicitly written using Eq. (46) as
I ( ψ 0 , z ) = N + 1 8 π i z [ ∑ ℓ ∈ o d d sin ( k ℓ ) ⟨ E ℓ ∣ ψ 0 ⟩ ( e 2 i γ τ cos ( k ℓ ) 1 − z e 2 i γ τ cos ( k ℓ ) ) ∑ ℓ ∈ o d d sin 2 ( k ℓ ) ( 1 1 − z e 2 i γ τ cos ( k ℓ ) ) × ∑ ℓ ∈ e v e n sin ( k ℓ ) ⟨ ψ 0 ∣ E ℓ ⟩ ( e − 2 i γ τ cos ( k ℓ ) z − e − 2 i γ τ cos ( k ℓ ) ) ∑ ℓ ∈ e v e n sin 2 ( k ℓ ) ( 1 z − e − 2 i γ τ cos ( k ℓ ) ) ] , (63) \mathcal{I}(\psi_0,z)=\frac{N+1}{8\pi i z}
\left[
\frac{\sum_{\ell\in\mathrm{odd}}\sin(k_{\ell})\langle E_{\ell}\vert\psi_0\rangle\left(\frac{e^{2i\gamma\tau\cos(k_{\ell})}}{1-ze^{2i\gamma\tau\cos(k_{\ell})}}\right)}
{\sum_{\ell\in\mathrm{odd}}\sin^2(k_{\ell})\left(\frac{1}{1-ze^{2i\gamma\tau\cos(k_{\ell})}}\right)}
\right.
\left.
\times
\frac{\sum_{\ell\in\mathrm{even}}\sin(k_{\ell})\langle\psi_0\vert E_{\ell}\rangle\left(\frac{e^{-2i\gamma\tau\cos(k_{\ell})}}{z-e^{-2i\gamma\tau\cos(k_{\ell})}}\right)}
{\sum_{\ell\in\mathrm{even}}\sin^2(k_{\ell})\left(\frac{1}{z-e^{-2i\gamma\tau\cos(k_{\ell})}}\right)}
\right], \tag*{(63)} I ( ψ 0 , z ) = 8 π i z N + 1 ∑ ℓ ∈ odd sin 2 ( k ℓ ) ( 1 − z e 2 iγ τ c o s ( k ℓ ) 1 ) ∑ ℓ ∈ odd sin ( k ℓ ) ⟨ E ℓ ∣ ψ 0 ⟩ ( 1 − z e 2 iγ τ c o s ( k ℓ ) e 2 iγ τ c o s ( k ℓ ) ) × ∑ ℓ ∈ even sin 2 ( k ℓ ) ( z − e − 2 iγ τ c o s ( k ℓ ) 1 ) ∑ ℓ ∈ even sin ( k ℓ ) ⟨ ψ 0 ∣ E ℓ ⟩ ( z − e − 2 iγ τ c o s ( k ℓ ) e − 2 iγ τ c o s ( k ℓ ) ) , (63)
where remember k ℓ = π ℓ / ( N + 1 ) k_{\ell}=\pi\ell/(N+1) k ℓ = π ℓ / ( N + 1 ) . Once again, we emphasize that Eqs. (61) –(63) are valid for the nearest-neighbor hopping Hamiltonian at generic sampling times, and the special sampling times at which quasienergy degeneracies occur will be treated separately later. Since ζ + − ( ψ 0 ) \zeta_{+-}(\psi_0) ζ +− ( ψ 0 ) depends on both the sampling time and the initial condition, it gives rise to the rich dependence of P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) and P R ( ψ 0 ) P_R(\psi_0) P R ( ψ 0 ) on τ \tau τ and ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ , which we discuss next.
Universal Zeno Features
For the Hamiltonian in Eq. (54) , we discussed in Eq. (11) that, in the Zeno limit, the splitting probability depends only on the projection of the initial state onto the two targets. To prove this result, let us consider the 0 < γ τ ≪ 1 0<\gamma\tau\ll1 0 < γ τ ≪ 1 limit of Eq. (61) . As evident, one then needs to evaluate the interference term ζ + − ( ψ 0 ) \zeta_{+-}(\psi_0) ζ +− ( ψ 0 ) in this expression, for which we turn to Eqs. (62) –(63) , and simplify them for γ τ ≪ 1 \gamma\tau\ll1 γ τ ≪ 1 . In particular, the function I ( ψ 0 , z ) \mathcal{I}(\psi_0,z) I ( ψ 0 , z ) in Eq. (63) becomes
I ( ψ 0 , z ) ≃ N + 1 8 π i z [ ∑ ℓ ∈ o d d sin ( k ℓ ) ⟨ E ℓ ∣ ψ 0 ⟩ ∑ ℓ ∈ o d d sin 2 ( k ℓ ) ] [ ∑ ℓ ∈ e v e n sin ( k ℓ ) ⟨ ψ 0 ∣ E ℓ ⟩ ∑ ℓ ∈ e v e n sin 2 ( k ℓ ) ] . (64) \mathcal{I}(\psi_0,z)\simeq\frac{N+1}{8\pi i z}
\left[
\frac{\sum_{\ell\in\mathrm{odd}}\sin(k_{\ell})\langle E_{\ell}\vert\psi_0\rangle}
{\sum_{\ell\in\mathrm{odd}}\sin^2(k_{\ell})}
\right]
\left[
\frac{\sum_{\ell\in\mathrm{even}}\sin(k_{\ell})\langle\psi_0\vert E_{\ell}\rangle}
{\sum_{\ell\in\mathrm{even}}\sin^2(k_{\ell})}
\right]. \tag*{(64)} I ( ψ 0 , z ) ≃ 8 π i z N + 1 [ ∑ ℓ ∈ odd sin 2 ( k ℓ ) ∑ ℓ ∈ odd sin ( k ℓ ) ⟨ E ℓ ∣ ψ 0 ⟩ ] [ ∑ ℓ ∈ even sin 2 ( k ℓ ) ∑ ℓ ∈ even sin ( k ℓ ) ⟨ ψ 0 ∣ E ℓ ⟩ ] . (64)
We next use
∑ ℓ ∈ o d d sin 2 ( k ℓ ) = ∑ ℓ ∈ e v e n sin 2 ( k ℓ ) = N + 1 4 , (65) \sum_{\ell\in\mathrm{odd}}\sin^2(k_{\ell})=\sum_{\ell\in\mathrm{even}}\sin^2(k_{\ell})=\frac{N+1}{4}, \tag*{(65)} ℓ ∈ odd ∑ sin 2 ( k ℓ ) = ℓ ∈ even ∑ sin 2 ( k ℓ ) = 4 N + 1 , (65)
and recast Eq. (64) as
I ( ψ 0 , z ) ≃ ⟨ d + ∣ ψ 0 ⟩ ⟨ ψ 0 ∣ d − ⟩ 2 π i z ≃ 1 4 π i z [ ∣ ⟨ 1 ∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 − 2 i Im ( ⟨ 1 ∣ ψ 0 ⟩ ⟨ ψ 0 ∣ N ⟩ ) ] . (66) \mathcal{I}(\psi_0,z) \simeq\frac{\langle d_+|\psi_0\rangle\langle\psi_0|d_-\rangle}{2\pi i z} \simeq\frac{1}{4\pi i z}\left[\left|\langle1|\psi_0\rangle\right|^2-\left|\langle N|\psi_0\rangle\right|^2-2i\operatorname{Im}\left(\langle1|\psi_0\rangle\langle\psi_0|N\rangle\right)\right]. \tag*{(66)} I ( ψ 0 , z ) ≃ 2 π i z ⟨ d + ∣ ψ 0 ⟩ ⟨ ψ 0 ∣ d − ⟩ ≃ 4 π i z 1 [ ∣ ⟨ 1∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 − 2 i Im ( ⟨ 1∣ ψ 0 ⟩ ⟨ ψ 0 ∣ N ⟩ ) ] . (66)
where ∣ d ± ⟩ |d_\pm\rangle ∣ d ± ⟩ are given in Eqs. (57) –(58) . Plugging the form of I ( ψ 0 , z ) \mathcal{I}(\psi_0,z) I ( ψ 0 , z ) in Eq. (62) gives
ζ + − ( ψ 0 ) ≃ 1 2 ( ∣ ⟨ 1 ∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 ) . (67) \zeta_{+-}(\psi_0) \simeq\frac{1}{2}\left(\left|\langle1|\psi_0\rangle\right|^2-\left|\langle N|\psi_0\rangle\right|^2\right). \tag*{(67)} ζ +− ( ψ 0 ) ≃ 2 1 ( ∣ ⟨ 1∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 ) . (67)
Inserting this result in Eq. (61) , we obtain the Zeno-limit expressions of the splitting probabilities outlined in Eq. (11) for the one-dimensional tight-binding chain with nearest-neighbor hopping. Interestingly, in the Zeno limit, the splitting probabilities are completely independent of the detailed structure of the initial state ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ in the interior of the chain and depend only on its projections onto the two boundary target sites.
To illustrate this result further, let us consider a localized initial condition, ∣ ψ 0 ⟩ = ∣ x 0 ⟩ |\psi_0\rangle=|x_0\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ . In this case, Eq. (11) immediately gives
P L ( x 0 ) ≃ { 1 , for x 0 = 1 , 0 , for x 0 = N , 1 2 , otherwise , P R ( x 0 ) ≃ { 0 , for x 0 = 1 , 1 , for x 0 = N , 1 2 , otherwise . (68) \begin{aligned}
P_L(x_0)&\simeq\begin{cases} 1, & \text{for }x_0=1,\\ 0, & \text{for }x_0=N,\\ \frac{1}{2}, & \text{otherwise}, \end{cases}
&
P_R(x_0)&\simeq\begin{cases} 0, & \text{for }x_0=1,\\ 1, & \text{for }x_0=N,\\ \frac{1}{2}, & \text{otherwise}. \end{cases} \tag*{(68)}
\end{aligned} P L ( x 0 ) ≃ ⎩ ⎨ ⎧ 1 , 0 , 2 1 , for x 0 = 1 , for x 0 = N , otherwise , P R ( x 0 ) ≃ ⎩ ⎨ ⎧ 0 , 1 , 2 1 , for x 0 = 1 , for x 0 = N , otherwise . (68)
In Figure 3 (a), we compare our theoretical result with the numerical simulations. The simulation is performed for N = 50 N=50 N = 50 , γ = 1 \gamma=1 γ = 1 and τ = 0.1 \tau=0.1 τ = 0.1 . It is in excellent agreement with our theory. We refer to Appendix § 15 for details on the numerical simulations. Remarkably, for small sampling time, both splitting probabilities are equal to 1 / 2 1/2 1/2 , independent of the initial position x 0 x_0 x 0 , as long as x 0 x_0 x 0 is in the interior of the tight-binding chain. At the boundary sites, the walker is detected by the corresponding detector with probability one and by the opposite detector with probability zero; see Figure 3 (a). Notably, this behavior is completely universal, and holds irrespective of the system size N N N .
To understand the flat behavior physically, recall that for small values of τ \tau τ , the amplitudes of detection at the boundary sites, starting from anywhere in the interior of the chain, scale as ∼ τ \sim\tau ∼ τ . The corresponding detection probability between the successive measurements then scales as ∼ τ 2 \sim\tau^2 ∼ τ 2 [79 ] . Since τ \tau τ is small, this probability is very small. Moreover, each unsuccessful detection at the boundaries collapses the wave function into the interior of the chain. Therefore, the walker is restricted mostly in the interior, and with only a small probability of order τ 2 \tau^2 τ 2 does it get detected at the boundary sites. As a result, the walker typically takes a large number of measurement attempts to be detected unless its initial position already coincides with one of the detectors. Any initial asymmetry present in the system is effectively erased at such long detection times, leading to the flat splitting probability with a value equal to 1 / 2 1/2 1/2 .
Singular case τ = 0 \tau=0 τ = 0
In experiments, such as those conducted on waveguides [69 , 70 , 71 , 80 ] , and quantum computers [72 , 73 , 74 ] , the sampling time τ \tau τ can be small, though it remains finite (0 < γ τ ≪ 1 0<\gamma\tau\ll1 0 < γ τ ≪ 1 ). Now, we turn to the τ = 0 \tau=0 τ = 0 case to show that the Zeno limit is singular. In other words, we cannot recover the τ = 0 \tau=0 τ = 0 result by using Eq. (68) for 0 < γ τ ≪ 1 0<\gamma\tau\ll1 0 < γ τ ≪ 1 and plugging τ = 0 \tau=0 τ = 0 in this expression.
For τ = 0 \tau=0 τ = 0 , we show in Appendix § 16 that the splitting probabilities are given by
P L ( ψ 0 ) = ∣ ⟨ 1 ∣ ψ 0 ⟩ ∣ 2 , P R ( ψ 0 ) = ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 , (69) P_{L}(\psi_{0})=\lvert\langle1\vert\psi_{0}\rangle\rvert^{2},\qquad P_{R}(\psi_{0})=\lvert\langle N\vert\psi_{0}\rangle\rvert^{2}, \tag*{(69)} P L ( ψ 0 ) = ∣⟨ 1∣ ψ 0 ⟩ ∣ 2 , P R ( ψ 0 ) = ∣⟨ N ∣ ψ 0 ⟩ ∣ 2 , (69)
which, for ∣ ψ 0 ⟩ = ∣ x 0 ⟩ \lvert\psi_{0}\rangle=\lvert x_{0}\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ , becomes
P L ( x 0 ) = δ x 0 , 1 , P R ( x 0 ) = δ x 0 , N . (70) P_{L}(x_{0})=\delta_{x_{0},1},\qquad P_{R}(x_{0})=\delta_{x_{0},N}. \tag*{(70)} P L ( x 0 ) = δ x 0 , 1 , P R ( x 0 ) = δ x 0 , N . (70)
This result is clearly different from Eq. (68) . For τ = 0 \tau=0 τ = 0 , the system is never detected if x 0 x_{0} x 0 is not at the boundary. These results admit a simple physical interpretation based on quantum Zeno physics [79 , 81 ] . For wave functions initially localized in the interior of the chain, continuous measurements at the boundaries always yield a failed detection. Each failed measurement collapses the wave function in the interior of the chain, restricting the walker completely in this part. As a result, the walker is never detected at the boundary sites, and both splitting probabilities vanish. Combining Eq. (70) with Eq. (68) , we thus find that both P L ( x 0 ) P_{L}(x_{0}) P L ( x 0 ) and P R ( x 0 ) P_{R}(x_{0}) P R ( x 0 ) exhibit a universal discontinuous jump at τ = 0 \tau=0 τ = 0 , changing from 0 to 1 / 2 1/2 1/2 on slightly increasing τ \tau τ .
Importantly, the jump at τ = 0 \tau=0 τ = 0 is not merely a theoretical construct, but also leaves an experimental fingerprint in quantum computers. A finite number of measurement attempts leads to the broadening of the otherwise discontinuous jumps [72 , 73 , 74 ] . As explained above, the 1 / 2 1/2 1/2 value of the splitting probabilities in Eq. (68) crucially depends on the walker typically taking a large number of measurements attempts for a successful detection. We later show that putting a cap on the number of measurements leads to deviations from this value. A finite number of measurements leads to a broadening effect of the otherwise pointwise discontinuous jump, an effect discussed later. In particular, the limit of an infinite number of measurements used to detect the quantum walker and the γ τ ≪ 1 \gamma\tau\ll1 γ τ ≪ 1 limit do not commute.
Flat-to-Fluctuating Transition and Oscillations in the Splitting Probability
For localized initial states, ∣ ψ 0 ⟩ = ∣ x 0 ⟩ |\psi_0\rangle=|x_0\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ , we further examine how the splitting probabilities depend on the initial position x 0 x_0 x 0 for general sampling times for the nearest-neighbor hopping Hamiltonian. More general initial states will be considered later. For the former case, we use Eq. (61) for P L ( x 0 ) P_L(x_0) P L ( x 0 ) , along with the interference term ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) in Eqs. (62) –(63) . Following Eq. (55) , we also use ⟨ E s ∣ x 0 ⟩ = 2 / ( N + 1 ) sin ( k s x 0 ) \langle E_s|x_0\rangle=\sqrt{2/(N+1)}\sin(k_sx_0) ⟨ E s ∣ x 0 ⟩ = 2/ ( N + 1 ) sin ( k s x 0 ) (s ∈ { 1 , 2 , … , N } s\in\{1,2,\ldots,N\} s ∈ { 1 , 2 , … , N } ) in this expression of ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) . Figure 3 shows the plot of P L ( x 0 ) P_L(x_0) P L ( x 0 ) as a function of x 0 x_0 x 0 for τ = 1 \tau=1 τ = 1 and τ = 2 \tau=2 τ = 2 , along with the near-Zeno limit case (τ = 0.1 \tau=0.1 τ = 0.1 ). In all cases, the hopping amplitude is fixed at γ = 1 \gamma=1 γ = 1 and N = 50 N=50 N = 50 . In Figure 3 (b), we see that the profile for τ = 1 \tau=1 τ = 1 remains essentially flat and equal to 1 / 2 1/2 1/2 , closely resembling the profile in the near-Zeno limit in Figure 3 (a). Deviations occur only when initial position is close to the endpoints of the chain. In contrast, for τ = 2 \tau=2 τ = 2 in Figure 3 (c), this behavior changes qualitatively: P L ( x 0 ) P_L(x_0) P L ( x 0 ) loses its flat behavior and develops oscillations with respect to x 0 x_0 x 0 . Similar behavior is obtained for P R ( x 0 ) P_R(x_0) P R ( x 0 ) .
From Eq. (68) , the oscillations in the splitting probabilities arise due to ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) . It is therefore useful to analyze its Fourier transform
ζ ~ ( Q , N ) = ∑ x 0 = 1 N e − i Q x 0 ζ + − ( x 0 ) , (71) \widetilde{\zeta}(Q,N)=\sum_{x_0=1}^{N}e^{-iQx_0}\zeta_{+-}(x_0), \tag*{(71)} ζ ( Q , N ) = x 0 = 1 ∑ N e − i Q x 0 ζ +− ( x 0 ) , (71)
where Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with j ∈ { 1 , 2 , … , N } j\in\{1,2,\ldots,N\} j ∈ { 1 , 2 , … , N } being the wavenumber. Plugging the expression for ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) from Eqs. (62) –(63) in Eq. (71) , we numerically obtain this Fourier transform. The resulting heat map of the power spectrum, ∣ ζ ~ ( Q , N ) ∣ 2 |\widetilde{\zeta}(Q,N)|^2 ∣ ζ ( Q , N ) ∣ 2 , as a function of Q Q Q and τ \tau τ , for N = 50 N=50 N = 50 and γ = 1 \gamma=1 γ = 1 , is shown in Figure 4 . To construct the heat map, we randomly generate a ( Q , τ ) (Q,\tau) ( Q , τ ) pair, evaluate ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) for this pair numerically from Eqs. (62) –(63) for different x 0 x_0 x 0 , and then substitute these values into Eq. (71) to obtain the corresponding value of ∣ ζ ~ ( Q , N ) ∣ 2 |\widetilde{\zeta}(Q,N)|^2 ∣ ζ ( Q , N ) ∣ 2 . We then repeat this procedure for 10 4 10^4 1 0 4 randomly sampled ( Q , τ ) (Q,\tau) ( Q , τ ) pairs yielding the heat map in Figure 4 .
Figure 4. Heat map of the power spectrum ∣ ζ ~ ( Q , N ) ∣ 2 |\widetilde{\zeta}(Q,N)|^{2} ∣ ζ ( Q , N ) ∣ 2 for the nearest-neighbor hopping Hamiltonian, shown as a function of the discrete wavenumber Q ∈ { π / ( N + 1 ) , 2 π / ( N + 1 ) , … , N π / ( N + 1 ) } Q \in\{\pi/(N+1),2\pi/(N+1),\ldots,N\pi/(N+1)\} Q ∈ { π / ( N + 1 ) , 2 π / ( N + 1 ) , … , N π / ( N + 1 )} and the sampling time τ \tau τ , for N = 50 N=50 N = 50 and γ = 1 \gamma=1 γ = 1 . To construct the plot, we randomly sample a pair ( Q , τ ) (Q,\tau) ( Q , τ ) , evaluate ζ + − ( x 0 ) \zeta_{+-}(x_{0}) ζ +− ( x 0 ) numerically from Eqs. (62) –(63) , and then use Eq. (71) to obtain the corresponding value of ∣ ζ ~ ( Q , N ) ∣ 2 |\widetilde{\zeta}(Q,N)|^{2} ∣ ζ ( Q , N ) ∣ 2 . Repeating this procedure for 10 4 10^{4} 1 0 4 randomly sampled ( Q , τ ) (Q,\tau) ( Q , τ ) pairs yields the heat map. The dashed black line in the plot marks γ τ c = π / 2 ( ≈ 1.57 ) \gamma\tau_{c}=\pi/2(\approx1.57) γ τ c = π /2 ( ≈ 1.57 ) . For τ ≥ τ c \tau\geq\tau_{c} τ ≥ τ c , spectral branches emerge at which the value of the power spectrum is pronounced. As shown in the paper, in the large-N N N limit, these branches are given by Q max ( τ ) = 2 sin − 1 ( r τ c / τ ) Q_{\max}(\tau)=2\sin^{-1}(r\tau_{c}/\tau) Q m a x ( τ ) = 2 sin − 1 ( r τ c / τ ) , with r = 1 , 2 , … r=1,2,\ldots r = 1 , 2 , … . The white dashed curves show the predictions for r = 1 r=1 r = 1 and r = 2 r=2 r = 2 . These dominant wavenumbers govern the oscillations in ζ + − ( x 0 ) \zeta_{+-}(x_{0}) ζ +− ( x 0 ) and, through Eq. (61) , the oscillatory dependence of the splitting probability on the initial position x 0 x_{0} x 0 , as observed for τ ≥ τ c \tau\geq\tau_{c} τ ≥ τ c in Figure 3 . Observe that, since N = 50 N=50 N = 50 is finite, the allowed wavenumbers Q Q Q are discrete, and the spectral branches therefore appear as separated bright points rather than continuous lines.
Interestingly, for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c , where τ c \tau_c τ c is given in Eq. (12) , we observe pronounced branches in Figure 4 , depicted by white dashed lines, along which the value of ∣ ζ ~ ( Q , N ) ∣ 2 |\widetilde{\zeta}(Q,N)|^2 ∣ ζ ( Q , N ) ∣ 2 is significantly enhanced relative to the surrounding region. By contrast, for τ < τ c \tau<\tau_c τ < τ c , no such branches are observed. Physically, the enhanced Fourier modes imply that ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) develops oscillations as a function of the initial position x 0 x_0 x 0 only for τ > τ c \tau>\tau_c τ > τ c , which, according to Eq. (68) , directly leads to oscillations in the splitting probabilities. This provides a Fourier-space signature of the oscillatory behavior observed in Figure 3 , and identifies τ c \tau_c τ c as the threshold for the emergence of oscillations. Furthermore, we see in Figure 4 that the dominant wavenumber, Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) , at which the power spectrum is maximal depends on the sampling time; see the white dashed lines. This dependence was stated earlier in Eq. (13) . We now derive this result.
We first recast the expression of ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) in Eqs. (62) –(63) by using ⟨ E s ∣ x 0 ⟩ = 2 / ( N + 1 ) sin ( k s x 0 ) \langle E_s|x_0\rangle=\sqrt{2/(N+1)}\sin(k_sx_0) ⟨ E s ∣ x 0 ⟩ = 2/ ( N + 1 ) sin ( k s x 0 ) , as mentioned before. We then get
ζ + − ( x 0 ) = ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) sin ( k s x 0 ) sin ( k s ′ x 0 ) , (72) \zeta_{+-}(x_0)=\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\sin(k_sx_0)\sin(k_{s'}x_0), \tag*{(72)} ζ +− ( x 0 ) = s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) sin ( k s x 0 ) sin ( k s ′ x 0 ) , (72)
where k s k_s k s is the momentum mode given in Eq. (55) , and C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is defined as
C s , s ′ ( τ , N ) = Re [ 1 4 π ∫ 0 2 π d η μ s ( η ) ν s ′ ( η ) ] , (73) C_{s,s'}(\tau,N)=\operatorname{Re}\left[\frac{1}{4\pi}\int_{0}^{2\pi}d\eta\ \mu_s(\eta)\nu_{s'}(\eta)\right], \tag*{(73)} C s , s ′ ( τ , N ) = Re [ 4 π 1 ∫ 0 2 π d η μ s ( η ) ν s ′ ( η ) ] , (73)
μ s ( η ) = sin ( k s ) e 2 i γ τ cos ( k s ) sin 2 ( k s ) + [ 1 − e i η e 2 i γ τ cos ( k s ) ] ∑ m ≠ s m o d d [ sin 2 ( k m ) 1 − e i η e 2 i γ τ cos ( k m ) ] , (74) \mu_s(\eta)=\frac{\sin(k_s)e^{2i\gamma\tau\cos(k_s)}}{\sin^2(k_s)+\left[1-e^{i\eta}e^{2i\gamma\tau\cos(k_s)}\right]\sum_{\substack{m\ne s\\m\ \mathrm{odd}}}\left[\frac{\sin^2(k_m)}{1-e^{i\eta}e^{2i\gamma\tau\cos(k_m)}}\right]}, \tag*{(74)} μ s ( η ) = sin 2 ( k s ) + [ 1 − e i η e 2 iγ τ c o s ( k s ) ] ∑ m = s m odd [ 1 − e i η e 2 iγ τ c o s ( k m ) s i n 2 ( k m ) ] sin ( k s ) e 2 iγ τ c o s ( k s ) , (74)
ν s ′ ( η ) = sin ( k s ′ ) e − 2 i γ τ cos ( k s ′ ) sin 2 ( k s ′ ) + [ 1 − e − i η e − 2 i γ τ cos ( k s ′ ) ] ∑ m ≠ s ′ m e v e n [ sin 2 ( k m ) 1 − e − i η e − 2 i γ τ cos ( k m ) ] . (75) \nu_{s'}(\eta)=\frac{\sin(k_{s'})e^{-2i\gamma\tau\cos(k_{s'})}}{\sin^2(k_{s'})+\left[1-e^{-i\eta}e^{-2i\gamma\tau\cos(k_{s'})}\right]\sum_{\substack{m\ne s'\\m\ \mathrm{even}}}\left[\frac{\sin^2(k_m)}{1-e^{-i\eta}e^{-2i\gamma\tau\cos(k_m)}}\right]}. \tag*{(75)} ν s ′ ( η ) = sin 2 ( k s ′ ) + [ 1 − e − i η e − 2 iγ τ c o s ( k s ′ ) ] ∑ m = s ′ m even [ 1 − e − i η e − 2 iγ τ c o s ( k m ) s i n 2 ( k m ) ] sin ( k s ′ ) e − 2 iγ τ c o s ( k s ′ ) . (75)
Performing the Fourier transform of Eq. (72) , we get
ζ ~ ( Q , N ) = ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) 4 [ e − i ( Q − k s + k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q − k s + k s ′ ) ] sin [ 1 2 ( Q − k s + k s ′ ) ] + e − i ( Q + k s − k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q + k s − k s ′ ) ] sin [ 1 2 ( Q + k s − k s ′ ) ] − e − i ( Q − k s − k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q − k s − k s ′ ) ] sin [ 1 2 ( Q − k s − k s ′ ) ] − e − i ( Q + k s + k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q + k s + k s ′ ) ] sin [ 1 2 ( Q + k s + k s ′ ) ] ] . (76) \begin{aligned}
\widetilde{\zeta}(Q,N)={}&\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}\frac{C_{s,s'}(\tau,N)}{4}\Bigg[
e^{-i\frac{(Q-k_s+k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q-k_s+k_{s'})\right]}{\sin\left[\frac{1}{2}(Q-k_s+k_{s'})\right]}\\
&+e^{-i\frac{(Q+k_s-k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q+k_s-k_{s'})\right]}{\sin\left[\frac{1}{2}(Q+k_s-k_{s'})\right]}\\
&-e^{-i\frac{(Q-k_s-k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q-k_s-k_{s'})\right]}{\sin\left[\frac{1}{2}(Q-k_s-k_{s'})\right]}\\
&-e^{-i\frac{(Q+k_s+k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q+k_s+k_{s'})\right]}{\sin\left[\frac{1}{2}(Q+k_s+k_{s'})\right]}
\Bigg]. \tag*{(76)}
\end{aligned} ζ ( Q , N ) = s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N 4 C s , s ′ ( τ , N ) [ e − i 2 ( Q − k s + k s ′ ) ( N + 1 ) sin [ 2 1 ( Q − k s + k s ′ ) ] sin [ 2 N ( Q − k s + k s ′ ) ] + e − i 2 ( Q + k s − k s ′ ) ( N + 1 ) sin [ 2 1 ( Q + k s − k s ′ ) ] sin [ 2 N ( Q + k s − k s ′ ) ] − e − i 2 ( Q − k s − k s ′ ) ( N + 1 ) sin [ 2 1 ( Q − k s − k s ′ ) ] sin [ 2 N ( Q − k s − k s ′ ) ] − e − i 2 ( Q + k s + k s ′ ) ( N + 1 ) sin [ 2 1 ( Q + k s + k s ′ ) ] sin [ 2 N ( Q + k s + k s ′ ) ] ] . (76)
Eq. (76) contains terms of the form ∼ sin ( N y ) / sin ( y ) \sim\sin(Ny)/\sin(y) ∼ sin ( N y ) / sin ( y ) , which become highly oscillatory for large N N N and develop sharp peaks of order N N N as y → 0 y\to0 y → 0 . These sharply peaked contributions simplify ζ ~ ( Q , N ) \widetilde{\zeta}(Q,N) ζ ( Q , N ) in Eq. (76) considerably in the large-N N N limit. In Appendix § 17 , we show that, for N ≫ 1 N\gg1 N ≫ 1 , ζ ~ ( Q , N ) \widetilde{\zeta}(Q,N) ζ ( Q , N ) with Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) takes different asymptotic forms depending on whether j j j is odd or even. Here we present the calculation for odd j j j and relegate the even-j j j case to Appendix § 17 . In the appendix, we show that for odd j j j and N ≫ 1 N\gg1 N ≫ 1 , Eq. (76) reduces to
ζ ~ ( Q , N ) ≃ T 1 ( Q , N ) + T 2 ( Q , N ) , (77) \widetilde{\zeta}(Q,N)\simeq T_1(Q,N)+T_2(Q,N), \tag*{(77)} ζ ( Q , N ) ≃ T 1 ( Q , N ) + T 2 ( Q , N ) , (77)
with
T 1 ( Q , N ) = N 4 ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) δ j , ∣ s − s ′ ∣ , (78) T_1(Q,N)=\frac{N}{4}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\delta_{j,\lvert s-s'\rvert}, \tag*{(78)} T 1 ( Q , N ) = 4 N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) δ j , ∣ s − s ′ ∣ , (78)
T 2 ( Q , N ) = − N 4 ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ . (79) T_2(Q,N)=-\frac{N}{4}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\delta_{j,N+1-\lvert N+1-s-s'\rvert}. \tag*{(79)} T 2 ( Q , N ) = − 4 N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ . (79)
Remember our goal is to determine the dominant wavenumber Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) at which the power spectrum ∣ ζ ~ ( Q , N ) ∣ 2 \lvert\widetilde{\zeta}(Q,N)\rvert^2 ∣ ζ ( Q , N ) ∣ 2 , and equivalently ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ , attains its maximum. Since ζ ~ ( Q , N ) \widetilde{\zeta}(Q,N) ζ ( Q , N ) in Eq. (77) is real and is composed of the two terms, T 1 ( Q , N ) T_1(Q,N) T 1 ( Q , N ) and T 2 ( Q , N ) T_2(Q,N) T 2 ( Q , N ) , it is useful to examine their contributions separately. In Figure 5 (a), we plot ∣ T 1 ( Q , N ) ∣ \lvert T_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ , ∣ T 2 ( Q , N ) ∣ \lvert T_2(Q,N)\rvert ∣ T 2 ( Q , N )∣ , and ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ as functions of Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with odd j j j for N = 200 N=200 N = 200 , γ = 1 \gamma=1 γ = 1 and τ = 2 \tau=2 τ = 2 . We observe that the maximum of ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ occurs at the same Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) as the maximum of ∣ T 1 ( Q , N ) ∣ \lvert T_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ . This value is indicated by the dashed line in Figure 5 (a). We have verified for other system sizes and sampling times that the maxima of ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ and ∣ T 1 ( Q , N ) ∣ \lvert T_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ likewise occur at the same Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) . We therefore proceed to determine Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) by analyzing T 1 ( Q , N ) T_1(Q,N) T 1 ( Q , N ) .
Figure 5. (a) The absolute value of the Fourier transform ζ ~ ( Q , N ) \widetilde{\zeta}(Q,N) ζ ( Q , N ) in Eq. (76) is plotted as a function of Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) for odd j j j , N = 200 N=200 N = 200 , γ = 1 \gamma=1 γ = 1 , and τ = 2 \tau=2 τ = 2 , along with the two terms ∣ T 1 ( Q , N ) ∣ |T_1(Q,N)| ∣ T 1 ( Q , N ) ∣ and ∣ T 2 ( Q , N ) ∣ |T_2(Q,N)| ∣ T 2 ( Q , N ) ∣ in Eqs. (78) and (79) . The maximum of ∣ ζ ~ ( Q , N ) ∣ |\widetilde{\zeta}(Q,N)| ∣ ζ ( Q , N ) ∣ occurs at the same Q Q Q -value as the maximum of ∣ T 1 ( Q , N ) ∣ |T_1(Q,N)| ∣ T 1 ( Q , N ) ∣ ; denoted by Q max ( τ ) ( ≈ 1.81 ) Q_{\max}^{(\tau)}(\approx1.81) Q m a x ( τ ) ( ≈ 1.81 ) and depicted by the dashed line. We show in Eq. (86) that for large N N N , Q max ( τ ) = 2 sin − 1 ( π r / 2 γ τ ) Q_{\max}^{(\tau)}=2\sin^{-1}(\pi r/2\gamma\tau) Q m a x ( τ ) = 2 sin − 1 ( π r /2 γ τ ) , where r r r is a positive integer. (b) The functions ∣ μ s ( η ) ∣ 2 |\mu_s(\eta)|^2 ∣ μ s ( η ) ∣ 2 and ∣ ν s ′ ( η ) ∣ 2 |\nu_{s'}(\eta)|^2 ∣ ν s ′ ( η ) ∣ 2 in Eqs. (74) and (75) , respectively, are plotted as functions of η \eta η for s = 19 s=19 s = 19 and s ′ = 60 s'=60 s ′ = 60 , with the same parameters as in panel (a). These functions determine C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) through Eq. (73) . Both functions are sharply peaked at η μ s p e a k ≈ 2.4 \eta_{\mu_s}^{\mathrm{peak}}\approx2.4 η μ s peak ≈ 2.4 and η ν s ′ p e a k ≈ 3.9 \eta_{\nu_{s'}}^{\mathrm{peak}}\approx3.9 η ν s ′ peak ≈ 3.9 , respectively, consistent with the peak locations given in Eqs. (80) and (81) . (c) The function C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) in Eq. (73) is plotted with respect to the even index s ′ s' s ′ for five representative fixed odd values of s s s , using the same parameters as in panel (a). In the large-N N N limit, we showed that C s , s ′ ( τ , N ) ≤ 2 / N C_{s,s'}(\tau,N)\leq2/N C s , s ′ ( τ , N ) ≤ 2/ N , where the value 2 / N 2/N 2/ N is indicated by the green dashed line. For the finite system shown here, N = 200 N=200 N = 200 , the bound is not saturated. However, the maximal value of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) approaches 2 / N 2/N 2/ N as N N N increases.
From the expression for T 1 ( Q , N ) T_1(Q,N) T 1 ( Q , N ) in Eq. (78) , we see that each pair ( s , s ′ ) (s,s') ( s , s ′ ) , with odd s s s and even s ′ s' s ′ is weighted by C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) . Hence, the dominant contributions to T 1 ( Q , N ) T_1(Q,N) T 1 ( Q , N ) arise from those ( s , s ′ ) (s,s') ( s , s ′ ) pairs for which C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is maximal or close to its maximum value. Let us next determine the maximal value of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) .
Maximum of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) : According to Eq. (73) , C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) involves integral of the product of two functions, μ s ( η ) \mu_s(\eta) μ s ( η ) and ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) , defined in Eqs. (74) and (75) , respectively. As functions of η ∈ [ 0 , 2 π ] \eta\in[0,2\pi] η ∈ [ 0 , 2 π ] with fixed s s s and s ′ s' s ′ , we show in Appendix § 17 that, for large N N N , these two functions are sharply peaked around
η μ s peak ≃ − 2 γ τ cos ( k s ) ( m o d 2 π ) , (80) \eta^{\mu_s\text{ peak}}\simeq-2\gamma\tau\cos(k_s)\pmod{2\pi}, \tag*{(80)} η μ s peak ≃ − 2 γ τ cos ( k s ) ( mod 2 π ) , (80)
η ν s ′ peak ≃ − 2 γ τ cos ( k s ′ ) ( m o d 2 π ) . (81) \eta^{\nu_{s'}\text{ peak}}\simeq-2\gamma\tau\cos(k_{s'})\pmod{2\pi}. \tag*{(81)} η ν s ′ peak ≃ − 2 γ τ cos ( k s ′ ) ( mod 2 π ) . (81)
Once again we emphasize s s s is odd and s ′ s' s ′ is even. We have illustrated Eqs. (80) –(81) in Figure 5 (b) for s = 19 s=19 s = 19 , s ′ = 60 s'=60 s ′ = 60 , N = 200 N=200 N = 200 , γ = 1 \gamma=1 γ = 1 and τ = 2 \tau=2 τ = 2 , where we use Eqs. (74) and (75) to plot the absolute values of μ s ( η ) \mu_s(\eta) μ s ( η ) and ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) as functions of η \eta η . We see both of them are peaked at η μ s p e a k ≈ 2.4 \eta_{\mu_s^{\mathrm{peak}}}\approx2.4 η μ s peak ≈ 2.4 and η ν s ′ p e a k ≈ 3.9 \eta_{\nu_{s'}^{\mathrm{peak}}}\approx3.9 η ν s ′ peak ≈ 3.9 , consistent with Eqs. (80) –(81) . From Eq. (73) , it then follows that the integration over η \eta η gives a pronounced contribution to C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) only when the locations of the two peaks overlap, i.e., η μ s p e a k ≈ η ν s ′ p e a k \eta_{\mu_s^{\mathrm{peak}}}\approx\eta_{\nu_{s'}^{\mathrm{peak}}} η μ s peak ≈ η ν s ′ peak . When the two peaks are well separated, one function is negligible in the region where the other is large, so their product, and consequently its contribution to the integral, is negligible. For example, for the ( s , s ′ ) (s,s') ( s , s ′ ) pair mentioned above, we saw that the peak locations are well-separated and one finds C s , s ′ ( τ , N ) ∼ 10 − 5 C_{s,s'}(\tau,N)\sim10^{-5} C s , s ′ ( τ , N ) ∼ 1 0 − 5 . By contrast, for ( s , s ′ ) = ( 33 , 150 ) (s,s')=(33,150) ( s , s ′ ) = ( 33 , 150 ) , one gets η μ s p e a k ≈ η ν s ′ p e a k ≈ 2.8 \eta_{\mu_s^{\mathrm{peak}}}\approx\eta_{\nu_{s'}^{\mathrm{peak}}}\approx2.8 η μ s peak ≈ η ν s ′ peak ≈ 2.8 , and correspondingly C s , s ′ ( τ , N ) ∼ 10 − 3 C_{s,s'}(\tau,N)\sim10^{-3} C s , s ′ ( τ , N ) ∼ 1 0 − 3 . Thus, pronounced values of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) arise when the peak locations coincide. Using Eqs. (80) –(81) , this condition can be written as
∣ E s − E s ′ ∣ τ = 2 π r , r = 1 , 2 , … , (82) \lvert E_s-E_{s'}\rvert\tau=2\pi r,\qquad r=1,2,\ldots, \tag*{(82)} ∣ E s − E s ′ ∣ τ = 2 π r , r = 1 , 2 , … , (82)
such that ∣ E s ⟩ \lvert E_s\rangle ∣ E s ⟩ has even parity while ∣ E s ′ ⟩ \lvert E_{s'}\rangle ∣ E s ′ ⟩ has odd parity (corresponding to s s s being odd and s ′ s' s ′ being even, as proven later). The associated C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is derived in Appendix D to be
C s , s ′ ( τ , N ) ≃ 4 N sin ( k s ) sin ( k s ′ ) [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] 2 + 16 α s ( τ ) 2 [ sin 2 ( k s ) − sin 2 ( k s ′ ) ] 2 . (83) C_{s,s'}(\tau,N)\simeq\frac{4}{N}\frac{\sin(k_s)\sin(k_{s'})\left[\sin^2(k_s)+\sin^2(k_{s'})\right]}{\left[\sin^2(k_s)+\sin^2(k_{s'})\right]^2+16\alpha_s(\tau)^2\left[\sin^2(k_s)-\sin^2(k_{s'})\right]^2}. \tag*{(83)} C s , s ′ ( τ , N ) ≃ N 4 [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] 2 + 16 α s ( τ ) 2 [ sin 2 ( k s ) − sin 2 ( k s ′ ) ] 2 sin ( k s ) sin ( k s ′ ) [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] . (83)
with α s ( τ ) \alpha_s(\tau) α s ( τ ) given in Eq. (187) . We further maximize C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) in Eq. (83) by noting that the second term in its denominator is nonnegative, and C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is maximized when this term vanishes, i.e., sin 2 ( k s ) = sin 2 ( k s ′ ) \sin^2(k_s)=\sin^2(k_{s'}) sin 2 ( k s ) = sin 2 ( k s ′ ) . Since k s ≠ k s ′ k_s\ne k_{s'} k s = k s ′ (remember s ≠ s ′ s\ne s' s = s ′ ), we obtain the second condition
k s + k s ′ = π . (84) k_s+k_{s'}=\pi. \tag*{(84)} k s + k s ′ = π . (84)
The corresponding maximum value is C s , s ′ ( τ , N ) ≃ 2 / N C_{s,s'}(\tau,N)\simeq2/N C s , s ′ ( τ , N ) ≃ 2/ N . Eqs. (82) –(84) give the conditions for maximizing C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) with respect to s s s and s ′ s' s ′ in the large-N N N limit. In Figure 5 (c), we have used the exact expression of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) in Eq. (73) to plot it as a function of the even index s ′ s' s ′ for several representative fixed odd values of s s s , for N = 200 N=200 N = 200 , γ = 1 \gamma=1 γ = 1 and τ = 2 \tau=2 τ = 2 . For all pairs ( s , s ′ ) (s,s') ( s , s ′ ) , C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) remains below its large-N N N maximum value of 2 / N 2/N 2/ N , indicated by the green dashed line in Figure 5 (c). Notice that for finite system sizes, such as N = 200 N=200 N = 200 in Figure 5 (c), the values of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) remain strictly below 2 / N 2/N 2/ N . However, as the system size increases, the maximal value of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) approaches 2 / N 2/N 2/ N .
We also emphasize that although the maximal C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) decays as 2 / N 2/N 2/ N with the system size, the function T 1 ( Q , N ) \mathcal{T}_1(Q,N) T 1 ( Q , N ) remains finite because of the overall N N N prefactor in its expression in Eq. (78) .
Determination of Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) : Having obtained the conditions under which C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is maximal, we now return to the determination of Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) . As discussed before and illustrated in Figure 5 (a), the maxima of ∣ T 1 ( Q , N ) ∣ \lvert\mathcal{T}_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ and ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ occur at the same Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) . From Eq. (78) , ∣ T 1 ( Q , N ) ∣ \lvert\mathcal{T}_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ is maximum when C s , s ′ ( N , τ ) C_{s,s'}(N,\tau) C s , s ′ ( N , τ ) is maximum, and the corresponding wavenumber is Q max ( τ ) = ∣ k s − k s ′ ∣ Q_{\max}(\tau)=\lvert k_s-k_{s'}\rvert Q m a x ( τ ) = ∣ k s − k s ′ ∣ , where the momentum modes k s k_s k s and k s ′ k_{s'} k s ′ satisfy Eqs. (82) –(84) for N ≫ 1 N\gg1 N ≫ 1 . In particular, Eq. (82) can be expressed as
2 γ τ ∣ cos ( k s ) − cos ( k s ′ ) ∣ = 2 π r , 2\gamma\tau\lvert\cos(k_s)-\cos(k_{s'})\rvert=2\pi r, 2 γ τ ∣ cos ( k s ) − cos ( k s ′ )∣ = 2 π r ,
⟹ 4 γ τ sin ( ∣ k s − k s ′ ∣ 2 ) sin ( k s + k s ′ 2 ) = 2 π r . (85) \Longrightarrow4\gamma\tau\sin\left(\frac{\lvert k_s-k_{s'}\rvert}{2}\right)\sin\left(\frac{k_s+k_{s'}}{2}\right)=2\pi r. \tag*{(85)} ⟹ 4 γ τ sin ( 2 ∣ k s − k s ′ ∣ ) sin ( 2 k s + k s ′ ) = 2 π r . (85)
Using Eq. (84) along with Q max ( τ ) = ∣ k s − k s ′ ∣ Q_{\max}(\tau)=\lvert k_s-k_{s'}\rvert Q m a x ( τ ) = ∣ k s − k s ′ ∣ , we obtain
Q max ( τ ) = 2 sin − 1 ( π τ c τ ) , r = 1 , 2 , … (86) Q_{\max}(\tau)=2\sin^{-1}\left(\frac{\pi\tau_c}{\tau}\right),\qquad r=1,2,\ldots \tag*{(86)} Q m a x ( τ ) = 2 sin − 1 ( τ π τ c ) , r = 1 , 2 , … (86)
While this result was derived from Eqs. (77) –(79) for Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with odd j j j , we show in Appendix D that Eq. (86) also holds for even j j j in the large-N N N limit. Thus for both odd and even j j j , the dominant wavenumber is given by Eq. (86) in the large-N N N limit. The result was quoted before in Eq. (13) . Different values of r r r correspond to different branches observed in Figure 4 . The predictions for r = 1 r=1 r = 1 and r = 2 r=2 r = 2 in Eq. (86) are shown by the white dashed curves in the figure, and they show excellent agreement with Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) in the heat map of ∣ ζ ~ ( Q , N ) ∣ 2 \lvert\widetilde{\zeta}(Q,N)\rvert^2 ∣ ζ ( Q , N ) ∣ 2 in Figure 4 . Since the system size N = 50 N=50 N = 50 is finite, the allowed wavenumbers Q Q Q are discrete, and the spectral branches in Figure 4 therefore appear as separated bright points rather than continuous lines. Notice that Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) is real only for τ ≥ r τ c \tau\geq r\tau_c τ ≥ r τ c . Thus, the minimum sampling time for observing a dominant wavenumber is τ = τ c \tau=\tau_c τ = τ c .
Phase-matched momentum pairs
It is also interesting to look at the interpretation of the condition in Eq. (82) in terms of phase-matched momentum pairs. In the large-N N N limit, the allowed momenta become effectively continuous over the interval [ 0 , π ] [0,\pi] [ 0 , π ] ; see Eq. (55) . One can therefore drop the discrete indices s s s and s ′ s' s ′ and denote the corresponding momenta simply by k k k and k ′ k' k ′ . Then, the dispersion relation in Eq. (55) for the nearest-neighbor Hamiltonian becomes E k = − 2 γ cos k E_k = -2\gamma\cos k E k = − 2 γ cos k . Combining this with the phase-matching condition in Eq. (82) , we get
∣ cos k − cos k ′ ∣ = π r γ τ , r = 1 , 2 , … (87) \lvert\cos k-\cos k'\rvert=\frac{\pi r}{\gamma\tau},\qquad r=1,2,\ldots \tag*{(87)} ∣ cos k − cos k ′ ∣ = γ τ π r , r = 1 , 2 , … (87)
Let us consider k ′ < k k' < k k ′ < k (so cos k ′ > cos k \cos k' > \cos k cos k ′ > cos k ) and r = 1 r = 1 r = 1 . For a given k k k , the allowed momenta that satisfy the condition in Eq. (82) [equivalently Eq. (87) ] are
k ′ ( k , τ ) = cos − 1 ( cos k + π γ τ ) . (88) k'(k,\tau)=\cos^{-1}\left(\cos k+\frac{\pi}{\gamma\tau}\right). \tag*{(88)} k ′ ( k , τ ) = cos − 1 ( cos k + γ τ π ) . (88)
From these possible values of k ′ k' k ′ , the one satisfying Eq. (84) gives the dominant contribution to the oscillations, as explained above in Eq. (86) .
Let us now determine when such a phase-matched wave number k ′ k' k ′ exists, which can then be chosen to satisfy Eq. (84) . For k ′ k' k ′ in Eq. (88) to be real, the argument of the inverse cosine in Eq. (88) must lie within the interval [ − 1 , 1 ] [-1,1] [ − 1 , 1 ] , namely − 1 ≤ cos k + π γ τ ≤ 1 -1 \leq\cos k + \frac{\pi}{\gamma\tau} \leq1 − 1 ≤ cos k + γ τ π ≤ 1 . The lower bound is readily satisfied since cos k ≥ − 1 \cos k \geq-1 cos k ≥ − 1 and π / ( γ τ ) > 0 \pi/(\gamma\tau) > 0 π / ( γ τ ) > 0 . On the other hand, the upper bound gives cos k ≤ ( 1 − π γ τ ) \cos k \leq\left(1-\frac{\pi}{\gamma\tau}\right) cos k ≤ ( 1 − γ τ π ) , which implies k ≥ cos − 1 ( 1 − π γ τ ) k \geq\cos^{-1}\left(1-\frac{\pi}{\gamma\tau}\right) k ≥ cos − 1 ( 1 − γ τ π ) . Therefore, the range of k k k for which there exists a k ′ k' k ′ in Eq. (88) satisfying the phase-matching condition in Eq. (82) is
k min ( τ ) ≤ k ≤ π , k min ( τ ) = cos − 1 ( 1 − π γ τ ) . (89) k_{\min}(\tau)\leq k\leq\pi,\qquad k_{\min}(\tau)=\cos^{-1}\left(1-\frac{\pi}{\gamma\tau}\right). \tag*{(89)} k m i n ( τ ) ≤ k ≤ π , k m i n ( τ ) = cos − 1 ( 1 − γ τ π ) . (89)
However k min ( τ ) k_{\min}(\tau) k m i n ( τ ) is real only for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c , where τ c = π / 2 γ \tau_c = \pi/2\gamma τ c = π /2 γ is given in Eq. (12) . This means that a physically meaningful range of k k k exists only for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c . This can also be seen directly from Eq. (88) . For τ < τ c \tau< \tau_c τ < τ c , the requirement that k ′ k' k ′ be real implies cos k < − 1 \cos k < -1 cos k < − 1 , which is not possible for real values of k k k . Hence, no real phase-matched pair ( k , k ′ ) (k,k') ( k , k ′ ) exists for τ < τ c \tau< \tau_c τ < τ c . Only for τ ≥ τ c \tau\geq\tau_c τ ≥ τ c can a pair of momentum modes satisfying the phase-matching condition in Eq. (82) exist. As explained above, these phase-matched pairs, when combined with Eq. (84) , give pronounced contributions to the interference term ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) , which in turn gives rise to the oscillations of the splitting probabilities with x 0 x_0 x 0 .
Flat-to-fluctuating transition
So far, we studied the behavior of the splitting probability with respect to the initial position. In § 2 , we pointed out that, for large system sizes, the splitting probability undergoes a flat-to-fluctuating transition as the sampling time is varied. To see this, we plot P L ( x 0 ) P_L(x_0) P L ( x 0 ) as a function of γ τ \gamma\tau γ τ in Figure 6 for two different choices of the initial position, x 0 = 11 x_0=11 x 0 = 11 and x 0 = 30 x_0=30 x 0 = 30 . Moreover, in each panel, we show the splitting probability for three different system sizes, N = 200 N=200 N = 200 (blue), N = 300 N=300 N = 300 (red) and N = 400 N=400 N = 400 (green) for a fixed x 0 x_0 x 0 . We observe the data collapse across all panels, indicating that the splitting probability is essentially independent of N N N once the system is sufficiently large. For smaller N N N , we do not get the data collapse, see Figure 7 . For N ≫ 1 N\gg1 N ≫ 1 , P L ( x 0 ) P_L(x_0) P L ( x 0 ) , as shown in Figure 6 (a) and Figure 6 (b), remain essentially equal to 1 / 2 1/2 1/2 , independent of the sampling time till a critical sampling time γ τ c \gamma\tau_c γ τ c , shown by the dashed vertical line in Figure 6 . In this regime, the splitting probability is also independent of the initial position, as also seen before in Figure 3 (b). Beyond this critical sampling time, τ > τ c \tau>\tau_c τ > τ c , we see a sharp change in the splitting dynamics, with a fluctuating pattern of pronounced peaks and dips emerging in the probability as a function of γ τ \gamma\tau γ τ . In this regime, P L ( x 0 ) P_L(x_0) P L ( x 0 ) deviates from its flat value of 1 / 2 1/2 1/2 , and develops a clear dependence on the initial position. Since P L ( x 0 ) + P R ( x 0 ) = 1 P_L(x_0)+P_R(x_0)=1 P L ( x 0 ) + P R ( x 0 ) = 1 [see Eq. (61)], the same behavior can also be found for P R ( x 0 ) P_R(x_0) P R ( x 0 ) .
Figure 7. Emergence of the flat-to-fluctuating transition in the splitting probabilities P L ( x 0 ) P_L(x_0) P L ( x 0 ) (left panel) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) (right panel) for initial position x 0 = 11 x_0=11 x 0 = 11 and system sizes N = 15 , 20 , 25 , 30 N=15,20,25,30 N = 15 , 20 , 25 , 30 . Even at these moderate sizes, the transition is visible: the flat regime (γ τ ≤ π / 2 \gamma\tau\le\pi/2 γ τ ≤ π /2 ) shows probabilities near 1 / 2 1/2 1/2 , while beyond the critical point (γ τ > π / 2 \gamma\tau> \pi/2 γ τ > π /2 ), fluctuations appear with dependence on the values of N N N .
The flat-to-fluctuating transition emerges in the limit of large system size. We later demonstrate in Figure 9 that the splitting probabilities for small system size (N = 5 N=5 N = 5 ) do not exhibit the transition. However, somewhat surprisingly, one does not need to go to very large systems: even for moderate sizes such as N = 15 , 20 , 25 , 30 N=15,20,25,30 N = 15 , 20 , 25 , 30 , the transition is already clearly visible, as shown in Figure 7 . We already mentioned that the critical sampling time in Eq. (12) is related to the width of the energy band and also emerges in single-target problems [53 , 54 , 82 ] . We next provide an understanding of the transition based on the poles associated with the interference term in Eq. (53) . Such pole analysis has been instrumental in the context of other first hitting time problems [52 , 53 , 54 , 55 , 74 ] . Here, we leverage it to gain some understanding on the flat-to-fluctuating transition, including the location of the transition point given before in Eq. (12) .
Figure 9. Quantum splitting probabilities P R ( x 0 ) P_R(x_0) P R ( x 0 ) and P L ( x 0 ) P_L(x_0) P L ( x 0 ) are shown as functions of the measurement time interval τ \tau τ for a one-dimensional chain of size N = 5 N=5 N = 5 , H H H in Eq. (2) , and three different localized initial conditions: ∣ x 0 ⟩ = ∣ 3 ⟩ , ∣ 4 ⟩ , ∣ 5 ⟩ |x_0\rangle=|3\rangle, |4\rangle, |5\rangle ∣ x 0 ⟩ = ∣3 ⟩ , ∣4 ⟩ , ∣5 ⟩ (from the symmetry in Eq. (109) , these also give ∣ x 0 ⟩ = ∣ 1 ⟩ , ∣ 2 ⟩ |x_0\rangle=|1\rangle, |2\rangle ∣ x 0 ⟩ = ∣1 ⟩ , ∣2 ⟩ ). In all cases, the detectors are situated at the boundaries, ∣ D 1 ⟩ = ∣ 1 ⟩ |D_1\rangle=|1\rangle ∣ D 1 ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ 5 ⟩ |D_2\rangle=|5\rangle ∣ D 2 ⟩ = ∣5 ⟩ , and the hopping parameter is taken to be γ = 1 \gamma=1 γ = 1 . The black solid lines correspond to the theoretical predictions, given in Eq. (111) for ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩ , Eqs. (115) –(116) for x 0 = 4 x_0=4 x 0 = 4 , and Eq. (118) for ∣ x 0 ⟩ = ∣ 5 ⟩ |x_0\rangle=|5\rangle ∣ x 0 ⟩ = ∣5 ⟩ . The red and blue symbols denote the data obtained from the numerical simulations, with blue symbols depicting the resonances at special values of τ \tau τ . Details of the simulations are given in Appendix § 15 . Analytic predictions of the resonances for ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩ and ∣ x 0 ⟩ = ∣ 4 ⟩ |x_0\rangle=|4\rangle ∣ x 0 ⟩ = ∣4 ⟩ are summarized in Table I. In contrast, for ∣ x 0 ⟩ = ∣ 5 ⟩ |x_0\rangle=|5\rangle ∣ x 0 ⟩ = ∣5 ⟩ , no resonances occur, since the overlap conditions in Eqs. (102) and (103) are not satisfied.
Pole analysis
From Eq. (61) , it is clear that the deviation from the flat profile originates from the term ζ + ( x 0 ) \zeta_+(x_0) ζ + ( x 0 ) , which itself arises from the interference effect between the two auxiliary single-target detection problems with target states ∣ d ± ⟩ |d_\pm\rangle ∣ d ± ⟩ (see Eq. (53) ). We now demonstrate that the behavior of this interference term changes qualitatively at τ c \tau_c τ c . For τ < τ c \tau<\tau_c τ < τ c , ζ + ( x 0 ) \zeta_+(x_0) ζ + ( x 0 ) remains negligible in the large-system limit, whereas for τ > τ c \tau>\tau_c τ > τ c , it generally becomes appreciable, leading to deviations of the splitting probabilities from 1 / 2 1/2 1/2 via Eq. (61) .
Recall that in Eq. (62) , ζ + ( x 0 ) \zeta_+(x_0) ζ + ( x 0 ) is given by the real part of a contour integral of a complex function I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) in Eq. (63) . It is therefore natural to expect that the analytic structure of I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) controls the behavior of ζ + ( x 0 ) \zeta_+(x_0) ζ + ( x 0 ) . We therefore use Eq. (63) to numerically find the singularities of I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) in the complex plane. Our analysis shows that the only singularities present are simple poles, whose positions depend on the system parameters. Let z j ∗ z_j^* z j ∗ denote these poles, with j j j indexing them. According to the residue theorem, only the poles enclosed within the unit circle ∣ z ∣ = 1 |z|=1 ∣ z ∣ = 1 contribute to the contour integral in Eq. (62) , yielding
∮ ∣ z ∣ = 1 I ( x 0 , z ) d z = 2 π i ∑ ∣ z j ∗ ∣ < 1 Residue [ I ( x 0 , z ) ; z j ∗ ] . (90) \oint_{|z|=1} \mathcal{I}(x_0,z)\,\mathrm{d}z=2\pi i\sum_{|z_j^*|<1}\operatorname{Residue}\left[\mathcal{I}(x_0,z);z_j^*\right]. \tag*{(90)} ∮ ∣ z ∣ = 1 I ( x 0 , z ) d z = 2 π i ∣ z j ∗ ∣ < 1 ∑ Residue [ I ( x 0 , z ) ; z j ∗ ] . (90)
Here, Residue [ I ( x 0 , z ) ; z j ∗ ] \operatorname{Residue}\left[\mathcal{I}(x_0,z);z_j^*\right] Residue [ I ( x 0 , z ) ; z j ∗ ] denotes the residue of the function I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) at the pole z = z j ∗ z=z_j^* z = z j ∗ . The summation in Eq. (90) runs over all poles z j ∗ z_j^* z j ∗ that satisfy the condition ∣ z j ∗ ∣ < 1 |z_j^*|<1 ∣ z j ∗ ∣ < 1 , i.e., all singularities of I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) enclosed within the unit circle.
Combining Eqs. (62) and (90)
ζ + − ( x 0 ) = − 2 π Im [ ∑ ∣ z j ∗ ∣ < 1 Residue [ I ( x 0 , z ) ; z j ∗ ] ] . (91) \zeta_{+-}(x_0)=-2\pi\operatorname{Im}\left[\sum_{\lvert z_j^*\rvert<1}\operatorname{Residue}\left[\mathcal{I}(x_0,z);z_j^*\right]\right]. \tag*{(91)} ζ +− ( x 0 ) = − 2 π Im ∣ z j ∗ ∣ < 1 ∑ Residue [ I ( x 0 , z ) ; z j ∗ ] . (91)
In Figure 8 (a)-Figure 8 (c), we employ Eq. (63) to numerically obtain the poles of I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) for three different sampling times, (a) γ τ = 1 \gamma\tau=1 γ τ = 1 , (b) γ τ = π / 2 \gamma\tau=\pi/2 γ τ = π /2 , and (c) γ τ = 3.75 \gamma\tau=3.75 γ τ = 3.75 , corresponding to τ < τ c \tau<\tau_c τ < τ c in Figure 8 (a), τ = τ c \tau=\tau_c τ = τ c in Figure 8 (b), and τ > τ c \tau>\tau_c τ > τ c in Figure 8 (c). We choose N = 4000 N=4000 N = 4000 and x 0 = 11 x_0=11 x 0 = 11 for all panels. Since only poles lying inside or on the unit circle contribute to the residue in Eq. (91) , only these poles are displayed. For γ τ = 1 \gamma\tau=1 γ τ = 1 in Figure 8 (a), all poles except the one at the origin lie close to the unit circle and form a single arc between the points a ∗ ( τ ) a_*(\tau) a ∗ ( τ ) and b ∗ ( τ ) b_*(\tau) b ∗ ( τ ) , marked by green dots. By varying τ \tau τ , we identify these endpoints as a ∗ ( τ ) = exp ( − i E max τ ) a_*(\tau)=\exp(-iE_{\max}\tau) a ∗ ( τ ) = exp ( − i E m a x τ ) and b ∗ ( τ ) = exp ( − i E min τ ) b_*(\tau)=\exp(-iE_{\min}\tau) b ∗ ( τ ) = exp ( − i E m i n τ ) (E max ≈ 2 γ E_{\max}\approx2\gamma E m a x ≈ 2 γ and E min ≈ − 2 γ E_{\min}\approx-2\gamma E m i n ≈ − 2 γ being the maximum and the minimum energies of H H H in Eq. (54) for N ≫ 1 N\gg1 N ≫ 1 ). Increasing τ \tau τ further keeps the qualitative feature the same: poles continue to remain along the arc between the points a ∗ ( τ ) a_*(\tau) a ∗ ( τ ) and b ∗ ( τ ) b_*(\tau) b ∗ ( τ ) . The green arrows indicate the direction of the movement of these points as τ \tau τ is increased. At the critical sampling time, τ c = 2 π / Δ E \tau_c=2\pi/\Delta E τ c = 2 π /Δ E , a ∗ ( τ ) = b ∗ ( τ ) = − 1 a_*(\tau)=b_*(\tau)=-1 a ∗ ( τ ) = b ∗ ( τ ) = − 1 . The two endpoints meet and the arc therefore closes in Figure 8 (b). For τ > τ c \tau>\tau_c τ > τ c in Figure 8 (c), the points a ∗ ( τ ) a_*(\tau) a ∗ ( τ ) and b ∗ ( τ ) b_*(\tau) b ∗ ( τ ) move past each other, and the distribution of the poles changes.
Figure 8. Distribution of poles (z j ∗ z_j^* z j ∗ ) and corresponding residues of the integrand I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) , defined in Eqs. (63) , in the complex z z z -plane, which determines the behavior of the interference term ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) in Eq. (62) . We show the results for three different sampling times with ∣ ψ 0 ⟩ = ∣ x 0 ⟩ = ∣ 11 ⟩ |\psi_0\rangle=|x_0\rangle=|11\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ = ∣11 ⟩ and N = 4000 N=4000 N = 4000 . Panels (a-c) in the top show the poles (in magenta) for (a) τ < τ c \tau<\tau_c τ < τ c , (b) τ = τ c \tau=\tau_c τ = τ c , and (c) for τ > τ c \tau>\tau_c τ > τ c . Only poles located inside or on the unit circle will contribute to Eq. (62) and are thus shown. For γ τ = 1 \gamma\tau=1 γ τ = 1 in (a), all poles, except the one at the origin, lie close to the unit circle. They are distributed along a single arc between the points a ∗ = exp ( − i E max τ ) a_*=\exp(-iE_{\max}\tau) a ∗ = exp ( − i E m a x τ ) and b ∗ = exp ( − i E min τ ) b_*=\exp(-iE_{\min}\tau) b ∗ = exp ( − i E m i n τ ) , shown by green dots (the green arrows indicate the direction of the movement of these points as τ \tau τ is increased). At the critical point γ τ c = π / 2 \gamma\tau_c=\pi/2 γ τ c = π /2 in (b), the two arc endpoints coalesce at z = − 1 z=-1 z = − 1 , signaling the onset of the transition. For τ > τ c \tau>\tau_c τ > τ c , the points a ∗ a_* a ∗ and b ∗ b_* b ∗ move past each other, and the distribution of the poles change. To characterize this change, we plot the density of poles, ρ ( θ ) \rho(\theta) ρ ( θ ) , along the angular coordinate θ ∈ [ − π , π ] \theta\in[-\pi,\pi] θ ∈ [ − π , π ] in the complex plane in panels (d-f). For τ < τ c \tau<\tau_c τ < τ c , ρ ( θ ) \rho(\theta) ρ ( θ ) , shown in magenta, exhibits a U-shaped profile, bounded between E min τ E_{\min}\tau E m i n τ and E max τ E_{\max}\tau E m a x τ depicted by green lines. At τ = τ c \tau=\tau_c τ = τ c , the profile expands over the full range of θ \theta θ , with Δ E τ = 2 π \Delta E\tau=2\pi Δ E τ = 2 π . For τ > τ c \tau>\tau_c τ > τ c , the profile changes qualitatively and develops discontinuities at θ = E max τ − 2 π \theta=E_{\max}\tau-2\pi θ = E m a x τ − 2 π and θ = E min τ + 2 π \theta=E_{\min}\tau+2\pi θ = E m i n τ + 2 π . The corresponding residues of I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) are symmetrically distributed about the origin for τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , but are asymmetrically distributed for τ > τ c \tau>\tau_c τ > τ c , see (g-i) for the imaginary part of the residues (in blue) of I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) . Due to the asymmetric distribution, the sum of the residues, in general, do not cancel for τ > τ c \tau>\tau_c τ > τ c , leading to an appreciable interference contribution ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) via Eq. (91) . Using this result in Eq. (61) , we obtain that the splitting probabilities are essentially equal to 1 / 2 1/2 1/2 for τ ≤ τ c \tau\leq\tau_c τ ≤ τ c and deviate from 1 / 2 1/2 1/2 for τ > τ c \tau>\tau_c τ > τ c .
To characterize this change, we study the density of poles, ρ ( θ ) \rho(\theta) ρ ( θ ) , along the angular coordinate θ ∈ [ − π , π ] \theta\in[-\pi,\pi] θ ∈ [ − π , π ] in the complex plane in Figure 8 (d)-Figure 8 (f). More specifically, ρ ( θ ) d θ \rho(\theta)d\theta ρ ( θ ) d θ gives the fraction of poles whose angular coordinates lie in the interval [ θ , θ + d θ ] [\theta,\theta+d\theta] [ θ , θ + d θ ] . We plot ρ ( θ ) \rho(\theta) ρ ( θ ) vs θ \theta θ for the three sampling times mentioned above. For τ < τ c \tau<\tau_c τ < τ c , ρ ( θ ) \rho(\theta) ρ ( θ ) , shown in Figure 8 (d), exhibits a characteristic U-shaped profile bounded between E min τ E_{\min}\tau E m i n τ and E max τ E_{\max}\tau E m a x τ , indicated by the green lines. As τ \tau τ increases, the two boundaries move apart and the support of ρ ( θ ) \rho(\theta) ρ ( θ ) expands, while its qualitative U-shaped form remains unchanged. At the critical point, Δ E τ c = 2 π \Delta E\tau_c=2\pi Δ E τ c = 2 π , the density ρ ( θ ) \rho(\theta) ρ ( θ ) spans the entire possible range of θ \theta θ , while still having the U-shape; see Figure 8 (e). This corresponds to the meeting of the two endpoints a ∗ ( τ ) a_*(\tau) a ∗ ( τ ) and b ∗ ( τ ) b_*(\tau) b ∗ ( τ ) in Figure 8 (b). For τ > τ c \tau>\tau_c τ > τ c , the U-shaped profile of ρ ( θ ) \rho(\theta) ρ ( θ ) changes qualitatively, as shown in Figure 8 (f), as it develops discontinuities at θ = ( E max τ − 2 π ) \theta=(E_{\max}\tau-2\pi) θ = ( E m a x τ − 2 π ) and θ = ( E min τ + 2 π ) \theta=(E_{\min}\tau+2\pi) θ = ( E m i n τ + 2 π ) , indicated by the green lines.
According to Eq. (91) , the qualitative change in the distribution of poles affects the imaginary part of the residue corresponding to I ( x 0 , z ) \mathcal{I}(x_0,z) I ( x 0 , z ) in Eq. (63) arising due to the poles. For τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , we find in Figure 8 (g)-Figure 8 (h) that the residues exhibit a symmetric distribution about the origin. As a result, their summed contribution is negligible, this, according to Eq. (91) , implies that the resulting ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) is negligible. For example, we obtain ζ + − ( x 0 ) ≈ − 0.004 \zeta_{+-}(x_0)\approx-0.004 ζ +− ( x 0 ) ≈ − 0.004 for γ τ = 1 \gamma\tau=1 γ τ = 1 in Figure 8 (g) and ζ + − ( x 0 ) ≈ 0.002 \zeta_{+-}(x_0)\approx0.002 ζ +− ( x 0 ) ≈ 0.002 for γ τ = π / 2 \gamma\tau=\pi/2 γ τ = π /2 in Figure 8 (h). Plugging this result in Eq. (61) then gives P L ( x 0 ) ≈ P R ( x 0 ) ≈ 1 / 2 P_L(x_0)\approx P_R(x_0)\approx1/2 P L ( x 0 ) ≈ P R ( x 0 ) ≈ 1/2 . By contrast, the residues for τ > τ c \tau>\tau_c τ > τ c are not symmetrically distributed, see Figure 8 (i). Summing all the residues using Eq. (91) , we find that ζ + − ( x 0 ) \zeta_{+-}(x_0) ζ +− ( x 0 ) is, in general, not negligible. For instance, from Figure 8 (i), γ τ = 3.75 \gamma\tau=3.75 γ τ = 3.75 , ζ + − ( x 0 ) ≈ − 0.09 \zeta_{+-}(x_0)\approx-0.09 ζ +− ( x 0 ) ≈ − 0.09 . This result signals that the probabilities P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) differ from the flat value of 1 / 2 1/2 1/2 in the τ > τ c \tau>\tau_c τ > τ c regime. This, in turn, leads to the flat-to-fluctuating transition in Figure 6 in the large-N N N limit.
Although the discussion above focuses on the large-N N N limit, the flat-to-fluctuating transition, as noted before in Figure 7 , is already clearly visible for moderate system sizes like N = 15 , 20 , 25 , 30 N=15,20,25,30 N = 15 , 20 , 25 , 30 . For moderate N N N , the poles can also display a radial rearrangement in the complex plane, in addition to the angular rearrangement discussed above [43 ] . In the regime τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , the poles remain close to the unit circle, as in Figure 8 (a) and Figure 8 (b), even for moderate N N N , whereas for τ > τ c \tau>\tau_c τ > τ c , some of the poles move appreciably into the interior of the unit disk, giving rise to an irregular radial arrangement. However, as the system size is increased, the radial irregularity becomes progressively less pronounced and the poles again accumulate close to the unit circle, as observed in Figure 8 (c). Consequently, in the τ > τ c \tau>\tau_c τ > τ c regime, the moderate-N N N splitting probabilities in Figure 7 exhibit more pronounced and irregular fluctuations than those observed for large system sizes in Figure 6 .
Resonance Conditions and Exceptional Sampling Times
We now turn to the finite-N N N case and focus on the special sampling times, as was mentioned above in § 2 . As shown in Figure 9 for N = 5 N=5 N = 5 and different initial positions, both P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) exhibit point-wise discontinuous jumps at these special sampling times, and satisfy P L ( x 0 ) + P R ( x 0 ) < 1 P_L(x_0)+P_R(x_0)<1 P L ( x 0 ) + P R ( x 0 ) < 1 . To account for these jumps, one must use the general expression given in Eq. (52) . In this section, we explicitly derive the resonance conditions, which was previously pointed out in Eq. (10) .
It is instructive to note that, using Eq. (55) , the energy eigenfunction ψ ℓ ( x ) = ⟨ x ∣ E ℓ ⟩ = 2 N + 1 sin ( π ℓ x N + 1 ) \psi_\ell(x)=\langle x|E_\ell\rangle=\sqrt{\frac{2}{N+1}}\sin\left(\frac{\pi\ell x}{N+1}\right) ψ ℓ ( x ) = ⟨ x ∣ E ℓ ⟩ = N + 1 2 sin ( N + 1 π ℓ x ) (recall ℓ ∈ { 1 , 2 , … , N } \ell\in\{1,2,\ldots,N\} ℓ ∈ { 1 , 2 , … , N } ) possesses definite parity with respect to reflection about the center of the chain. Under the transformation x → ( N + 1 − x ) x\to(N+1-x) x → ( N + 1 − x ) , the eigenfunctions transform as
ψ ℓ ( N + 1 − x ) = ( − 1 ) ℓ + 1 ψ ℓ ( x ) , (92) \psi_\ell(N+1-x)=(-1)^{\ell+1}\psi_\ell(x), \tag*{(92)} ψ ℓ ( N + 1 − x ) = ( − 1 ) ℓ + 1 ψ ℓ ( x ) , (92)
indicating that eigenfunctions with odd ℓ \ell ℓ are symmetric under reflection, while those with even ℓ \ell ℓ are antisymmetric. Substituting this property into Eqs. (57) and (58) , we find that the auxiliary target state ∣ d + ⟩ \lvert d_{+}\rangle ∣ d + ⟩ involves only symmetric eigenstates, whereas ∣ d − ⟩ \lvert d_{-}\rangle ∣ d − ⟩ involves antisymmetric eigenstates.
This identification allows us to recast the auxiliary probabilities ρ ± ( x 0 ) \rho_{\pm}(x_{0}) ρ ± ( x 0 ) in Eq. (51) in terms of the parity-specific sectors of the spectrum
ρ + ( x 0 ) = ∑ ℓ ∈ o d d ′ ∣ ∑ m = 1 G e ⟨ d + ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ ∣ 2 ∑ m = 1 G e ∣ ⟨ d + ∣ E ℓ , m ⟩ ∣ 2 (93) \rho_{+}(x_{0})=\sum_{\ell\in\mathrm{odd}}^{\prime}\frac{\left\lvert\sum_{m=1}^{\mathcal{G}^{e}}\langle d_{+}\mid E_{\ell,m}\rangle\langle E_{\ell,m}\mid x_{0}\rangle\right\rvert^{2}}{\sum_{m=1}^{\mathcal{G}^{e}}\left\lvert\langle d_{+}\mid E_{\ell,m}\rangle\right\rvert^{2}} \tag*{(93)} ρ + ( x 0 ) = ℓ ∈ odd ∑ ′ ∑ m = 1 G e ∣ ⟨ d + ∣ E ℓ , m ⟩ ∣ 2 ∑ m = 1 G e ⟨ d + ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ 2 (93)
ρ − ( x 0 ) = ∑ ℓ ∈ e v e n ′ ∣ ∑ m = 1 G e ⟨ d − ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ ∣ 2 ∑ m = 1 G e ∣ ⟨ d − ∣ E ℓ , m ⟩ ∣ 2 (94) \rho_{-}(x_{0})=\sum_{\ell\in\mathrm{even}}^{\prime}\frac{\left\lvert\sum_{m=1}^{\mathcal{G}^{e}}\langle d_{-}\mid E_{\ell,m}\rangle\langle E_{\ell,m}\mid x_{0}\rangle\right\rvert^{2}}{\sum_{m=1}^{\mathcal{G}^{e}}\left\lvert\langle d_{-}\mid E_{\ell,m}\rangle\right\rvert^{2}} \tag*{(94)} ρ − ( x 0 ) = ℓ ∈ even ∑ ′ ∑ m = 1 G e ∣ ⟨ d − ∣ E ℓ , m ⟩ ∣ 2 ∑ m = 1 G e ⟨ d − ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ 2 (94)
and similarly for the auxiliary generating functions χ ~ ± ( x 0 ) \widetilde{\chi}_{\pm}(x_{0}) χ ± ( x 0 ) in Eq. (46)
χ ~ + ( z ) = ∑ ℓ ∈ o d d ∑ m = 1 G e ⟨ d + ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ z e − i λ ℓ 1 − z e − i λ ℓ ∑ ℓ ∈ o d d ∑ m = 1 G e ∣ ⟨ d + ∣ E ℓ , m ⟩ ∣ 2 1 1 − z e − i λ ℓ (95) \widetilde{\chi}_{+}(z)=\frac{\sum_{\ell\in\mathrm{odd}}\sum_{m=1}^{\mathcal{G}^{e}}\langle d_{+}\mid E_{\ell,m}\rangle\langle E_{\ell,m}\mid x_{0}\rangle\frac{ze^{-i\lambda_{\ell}}}{1-ze^{-i\lambda_{\ell}}}}{\sum_{\ell\in\mathrm{odd}}\sum_{m=1}^{\mathcal{G}^{e}}\left\lvert\langle d_{+}\mid E_{\ell,m}\rangle\right\rvert^{2}\frac{1}{1-ze^{-i\lambda_{\ell}}}} \tag*{(95)} χ + ( z ) = ∑ ℓ ∈ odd ∑ m = 1 G e ∣ ⟨ d + ∣ E ℓ , m ⟩ ∣ 2 1 − z e − i λ ℓ 1 ∑ ℓ ∈ odd ∑ m = 1 G e ⟨ d + ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ 1 − z e − i λ ℓ z e − i λ ℓ (95)
χ ~ − ( z ) = ∑ ℓ ∈ e v e n ∑ m = 1 G e ⟨ d − ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ z e − i λ ℓ 1 − z e − i λ ℓ ∑ ℓ ∈ e v e n ∑ m = 1 G e ∣ ⟨ d − ∣ E ℓ , m ⟩ ∣ 2 1 1 − z e − i λ ℓ (96) \widetilde{\chi}_{-}(z)=\frac{\sum_{\ell\in\mathrm{even}}\sum_{m=1}^{\mathcal{G}^{e}}\langle d_{-}\mid E_{\ell,m}\rangle\langle E_{\ell,m}\mid x_{0}\rangle\frac{ze^{-i\lambda_{\ell}}}{1-ze^{-i\lambda_{\ell}}}}{\sum_{\ell\in\mathrm{even}}\sum_{m=1}^{\mathcal{G}^{e}}\left\lvert\langle d_{-}\mid E_{\ell,m}\rangle\right\rvert^{2}\frac{1}{1-ze^{-i\lambda_{\ell}}}} \tag*{(96)} χ − ( z ) = ∑ ℓ ∈ even ∑ m = 1 G e ∣ ⟨ d − ∣ E ℓ , m ⟩ ∣ 2 1 − z e − i λ ℓ 1 ∑ ℓ ∈ even ∑ m = 1 G e ⟨ d − ∣ E ℓ , m ⟩ ⟨ E ℓ , m ∣ x 0 ⟩ 1 − z e − i λ ℓ z e − i λ ℓ (96)
In all of these expressions, the summation in a given sector runs over the distinct quasienergies λ ℓ = E ℓ τ m o d 2 π \lambda_{\ell}=E_{\ell}\tau\ \mathrm{mod}\ 2\pi λ ℓ = E ℓ τ mod 2 π . As noted previously, for most values of τ \tau τ , the number of distinct λ ℓ \lambda_{\ell} λ ℓ is same as the number of E ℓ E_{\ell} E ℓ , leading to Eq. (61) . For certain exceptional measurement times, however, two or more quasienergies from the same parity-sector can become equal, as shown in Figure 2 (b). Then
λ ℓ = λ ℓ ′ ⟹ e − i E ℓ τ = e − i E ℓ ′ τ (97) \lambda_{\ell}=\lambda_{\ell^{\prime}}\ \Longrightarrow\ e^{-iE_{\ell}\tau}=e^{-iE_{\ell^{\prime}}\tau} \tag*{(97)} λ ℓ = λ ℓ ′ ⟹ e − i E ℓ τ = e − i E ℓ ′ τ (97)
where ℓ ≠ ℓ ′ \ell\ne\ell^{\prime} ℓ = ℓ ′ and ℓ , ℓ ′ \ell,\ell^{\prime} ℓ , ℓ ′ are both odd or both even, so that the associated eigenstates, ∣ E ℓ ⟩ \lvert E_{\ell}\rangle ∣ E ℓ ⟩ and ∣ E ℓ ′ ⟩ \lvert E_{\ell^{\prime}}\rangle ∣ E ℓ ′ ⟩ , have the same parity under reflection. At such exceptional sampling times, the sums appearing in ρ ± ( x 0 ) \rho_{\pm}(x_{0}) ρ ± ( x 0 ) and χ ~ ± ( z ) \widetilde{\chi}_{\pm}(z) χ ± ( z ) in Eqs. (93) –(96) are reorganized, because several previously separate terms merge into a single degenerate contribution. This reorganization produces discontinuous changes in the auxiliary quantities, which in turn appear as sharp jumps in the splitting probabilities P L ( x 0 ) P_{L}(x_{0}) P L ( x 0 ) and P R ( x 0 ) P_{R}(x_{0}) P R ( x 0 ) . Thus, Eq. (97) gives the resonance condition, equivalent to Eq. (10) noted before.
By contrast, quasienergy coincidences between different parity sectors (i.e., odd ℓ \ell ℓ , even ℓ ′ \ell^{\prime} ℓ ′ or vice versa) do not alter the structure of the summation, and therefore do not generate such resonances. We next turn to a more physical picture that illustrates that the appearance of the resonance is related to the formation of dark states that are orthogonal to both targets, ∣ D 1 ⟩ = ∣ 1 ⟩ \lvert D_{1}\rangle=\lvert1\rangle ∣ D 1 ⟩ = ∣ 1 ⟩ and ∣ D 2 ⟩ = ∣ N ⟩ \lvert D_{2}\rangle=\lvert N\rangle ∣ D 2 ⟩ = ∣ N ⟩ .
Contrast the phase-matching condition associated with the resonances in Eq. (97) with that in Eq. (82) . Although both conditions correspond to matching of the dynamical phases, they involve different parity sectors and lead to different physical effects. The resonance condition in Eq. (97) involves states of the same parity and is associated to the formation of dark states (as shown next). In contrast, Eq. (82) involves states of opposite parity and governs the interference between the two parity sectors, which gives rise to the spatial oscillations of the splitting probabilities.
A physical picture based on dark-states
For simplicity, consider that ℓ \ell ℓ and ℓ ′ \ell^{\prime} ℓ ′ are both odd (ℓ ≠ ℓ ′ \ell\ne\ell^{\prime} ℓ = ℓ ′ ), and we have the phase-matching condition in Eq. (97) . In this case, the following state is orthogonal with respect to the both target states ∣ d ± ⟩ \lvert d_{\pm}\rangle ∣ d ± ⟩ in the auxiliary single-target detection problems
∣ D ℓ ℓ ′ ( + ) ⟩ ∝ ( ⟨ d + ∣ E ℓ ′ ⟩ ∣ E ℓ ⟩ − ⟨ d + ∣ E ℓ ⟩ ∣ E ℓ ′ ⟩ ) (98) \lvert D_{\ell\ell^{\prime}}^{(+)}\rangle\propto\left(\langle d_{+}\mid E_{\ell^{\prime}}\rangle\lvert E_{\ell}\rangle-\langle d_{+}\mid E_{\ell}\rangle\lvert E_{\ell^{\prime}}\rangle\right) \tag*{(98)} ∣ D ℓ ℓ ′ ( + ) ⟩ ∝ ( ⟨ d + ∣ E ℓ ′ ⟩ ∣ E ℓ ⟩ − ⟨ d + ∣ E ℓ ⟩ ∣ E ℓ ′ ⟩ ) (98)
∝ [ sin ( π ℓ ′ N + 1 ) ∣ E ℓ ⟩ − sin ( π ℓ N + 1 ) ∣ E ℓ ′ ⟩ ] . (99) \propto\left[\sin\left(\frac{\pi\ell^{\prime}}{N+1}\right)\lvert E_{\ell}\rangle-\sin\left(\frac{\pi\ell}{N+1}\right)\lvert E_{\ell^{\prime}}\rangle\right]. \tag*{(99)} ∝ [ sin ( N + 1 π ℓ ′ ) ∣ E ℓ ⟩ − sin ( N + 1 π ℓ ) ∣ E ℓ ′ ⟩ ] . (99)
Remember that ⟨ d − ∣ E ℓ ⟩ = ⟨ d − ∣ E ℓ ′ ⟩ = 0 \langle d_{-}\mid E_{\ell}\rangle=\langle d_{-}\mid E_{\ell^{\prime}}\rangle=0 ⟨ d − ∣ E ℓ ⟩ = ⟨ d − ∣ E ℓ ′ ⟩ = 0 when both ℓ \ell ℓ and ℓ ′ \ell^{\prime} ℓ ′ are odd [see Eq. (58) ]. This leads to ⟨ d − ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 \langle d_{-}\mid D_{\ell\ell^{\prime}}^{(+)}\rangle=0 ⟨ d − ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 . We also have ⟨ d + ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 \langle d_{+}\mid D_{\ell\ell^{\prime}}^{(+)}\rangle=0 ⟨ d + ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 . Together, these conditions establish that ∣ D ℓ ℓ ′ ( + ) ⟩ \lvert D_{\ell\ell^{\prime}}^{(+)}\rangle ∣ D ℓ ℓ ′ ( + ) ⟩ is orthogonal with respect to the target states, ∣ D 1 ⟩ \lvert D_{1}\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ \lvert D_{2}\rangle ∣ D 2 ⟩ , in the original splitting problem
⟨ D 1 ∣ D ℓ ℓ ′ ( + ) ⟩ = ⟨ D 2 ∣ D ℓ ℓ ′ ( + ) ⟩ = 0. (100) \langle D_{1}\mid D_{\ell\ell^{\prime}}^{(+)}\rangle=\langle D_{2}\mid D_{\ell\ell^{\prime}}^{(+)}\rangle=0. \tag*{(100)} ⟨ D 1 ∣ D ℓ ℓ ′ ( + ) ⟩ = ⟨ D 2 ∣ D ℓ ℓ ′ ( + ) ⟩ = 0. (100)
The phase-matching condition implies that ∣ D ℓ ℓ ′ ( + ) ⟩ |D_{\ell\ell'}^{(+)}\rangle ∣ D ℓ ℓ ′ ( + ) ⟩ is an eigenstate of the unitary operator precisely at the measurement time: U ( τ ) ∣ D ℓ ℓ ′ ( + ) ⟩ = e − i E ℓ τ ∣ D ℓ ℓ ′ ( + ) ⟩ U(\tau)|D_{\ell\ell'}^{(+)}\rangle= e^{-iE_\ell\tau}|D_{\ell\ell'}^{(+)}\rangle U ( τ ) ∣ D ℓ ℓ ′ ( + ) ⟩ = e − i E ℓ τ ∣ D ℓ ℓ ′ ( + ) ⟩ . This result, along with Eq. (100) , implies that ∣ D ℓ ℓ ′ ( + ) ⟩ |D_{\ell\ell'}^{(+)}\rangle ∣ D ℓ ℓ ′ ( + ) ⟩ is also an eigenstate of the survival operator S = ( 1 − D ^ ) U ( τ ) S = (1-\hat{D})U(\tau) S = ( 1 − D ^ ) U ( τ ) appearing in Eq. (27) .
S ∣ D ℓ ℓ ′ ( + ) ⟩ = e − i E ℓ τ ∣ D ℓ ℓ ′ ( + ) ⟩ , (101) S|D_{\ell\ell'}^{(+)}\rangle= e^{-iE_\ell\tau}|D_{\ell\ell'}^{(+)}\rangle, \tag*{(101)} S ∣ D ℓ ℓ ′ ( + ) ⟩ = e − i E ℓ τ ∣ D ℓ ℓ ′ ( + ) ⟩ , (101)
where D ^ \hat{D} D ^ , defined in Eq. (21) , is the projection operator onto the detector subspace. This means that ⟨ D 1 ∣ U ( τ ) S n − 1 ∣ D ℓ ℓ ′ ( + ) ⟩ = ⟨ D 2 ∣ U ( τ ) S n − 1 ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 \langle D_1|U(\tau)S^{n-1}|D_{\ell\ell'}^{(+)}\rangle= \langle D_2|U(\tau)S^{n-1}|D_{\ell\ell'}^{(+)}\rangle= 0 ⟨ D 1 ∣ U ( τ ) S n − 1 ∣ D ℓ ℓ ′ ( + ) ⟩ = ⟨ D 2 ∣ U ( τ ) S n − 1 ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 , implying that ∣ D ℓ ℓ ′ ( + ) ⟩ |D_{\ell\ell'}^{(+)}\rangle ∣ D ℓ ℓ ′ ( + ) ⟩ is a dark state with respect to both ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ at the measurement times τ \tau τ , 2 τ 2\tau 2 τ , 3 τ 3\tau 3 τ and so on. This state is never detected by the target detectors at the moments of measurement, and the corresponding splitting probabilities vanish.
If the initial state ∣ x 0 ⟩ |x_0\rangle ∣ x 0 ⟩ of the quantum walker has a finite overlap with this dark state, i.e., ⟨ x 0 ∣ D ℓ ℓ ′ ( + ) ⟩ ≠ 0 \langle x_0|D_{\ell\ell'}^{(+)}\rangle\ne0 ⟨ x 0 ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 , then the phase-matching condition implies that this component is completely decoupled from the two detectors ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩ , and a dark state is formed during the measurement attempts. As a result, we get a sudden reduction in the splitting probabilities. This will, however, not happen if ∣ x 0 ⟩ |x_0\rangle ∣ x 0 ⟩ is orthogonal to ∣ D ℓ ℓ ′ ( + ) ⟩ |D_{\ell\ell'}^{(+)}\rangle ∣ D ℓ ℓ ′ ( + ) ⟩ .
To summarize, the conditions for observing a resonance in P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) arising from the symmetric parity-sector (i.e., both ℓ \ell ℓ and ℓ ′ \ell' ℓ ′ are odd) are
e − i E ℓ τ = e − i E ℓ ′ τ , and ⟨ x 0 ∣ D ℓ ℓ ′ ( + ) ⟩ ≠ 0 , (102) e^{-iE_\ell\tau} = e^{-iE_{\ell'}\tau}, \quad\text{and} \quad\langle x_0|D_{\ell\ell'}^{(+)}\rangle\ne0, \tag*{(102)} e − i E ℓ τ = e − i E ℓ ′ τ , and ⟨ x 0 ∣ D ℓ ℓ ′ ( + ) ⟩ = 0 , (102)
and it related to the formation of a dark state right at the stroboscopic measurement times τ \tau τ , 2 τ 2\tau 2 τ , 3 τ 3\tau 3 τ and so on.
Similarly, when ℓ \ell ℓ and ℓ ′ \ell' ℓ ′ are both even, the resonance conditions are
e − i E ℓ τ = e − i E ℓ ′ τ , and ⟨ x 0 ∣ D ℓ ℓ ′ ( − ) ⟩ ≠ 0 , (103) e^{-iE_\ell\tau} = e^{-iE_{\ell'}\tau}, \quad\text{and} \quad\langle x_0|D_{\ell\ell'}^{(-)}\rangle\ne0, \tag*{(103)} e − i E ℓ τ = e − i E ℓ ′ τ , and ⟨ x 0 ∣ D ℓ ℓ ′ ( − ) ⟩ = 0 , (103)
where
∣ D ℓ ℓ ′ ( − ) ⟩ = N ( ⟨ d − ∣ E ℓ ⟩ ∣ E ℓ ⟩ − ⟨ d − ∣ E ℓ ′ ⟩ ∣ E ℓ ′ ⟩ ) , (104) |D_{\ell\ell'}^{(-)}\rangle= \mathcal{N}\left(\langle d_-|E_\ell\rangle|E_\ell\rangle- \langle d_-|E_{\ell'}\rangle|E_{\ell'}\rangle\right), \tag*{(104)} ∣ D ℓ ℓ ′ ( − ) ⟩ = N ( ⟨ d − ∣ E ℓ ⟩ ∣ E ℓ ⟩ − ⟨ d − ∣ E ℓ ′ ⟩ ∣ E ℓ ′ ⟩ ) , (104)
= N [ sin ( π ℓ ′ N + 1 ) ∣ E ℓ ⟩ − sin ( π ℓ N + 1 ) ∣ E ℓ ′ ⟩ ] , (105) = \mathcal{N}\left[\sin\left(\frac{\pi\ell'}{N+1}\right)|E_\ell\rangle- \sin\left(\frac{\pi\ell}{N+1}\right)|E_{\ell'}\rangle\right], \tag*{(105)} = N [ sin ( N + 1 π ℓ ′ ) ∣ E ℓ ⟩ − sin ( N + 1 π ℓ ) ∣ E ℓ ′ ⟩ ] , (105)
is a dark-state with respect to ∣ D 1 ⟩ |D_1\rangle ∣ D 1 ⟩ and ∣ D 2 ⟩ |D_2\rangle ∣ D 2 ⟩
⟨ D 1 ∣ D ℓ ℓ ′ ( − ) ⟩ = ⟨ D 2 ∣ D ℓ ℓ ′ ( − ) ⟩ = 0. (106) \langle D_1|D_{\ell\ell'}^{(-)}\rangle= \langle D_2|D_{\ell\ell'}^{(-)}\rangle= 0. \tag*{(106)} ⟨ D 1 ∣ D ℓ ℓ ′ ( − ) ⟩ = ⟨ D 2 ∣ D ℓ ℓ ′ ( − ) ⟩ = 0. (106)
Eqs. (102) and (103) thus specify the special τ \tau τ values, characterized by the quasienergy phase matching together with a nonvanishing overlap of the initial state with the corresponding dark subspace. They were stated before in Eq. (10) . Furthermore, Eqs. (100) and (106) show that no resonance arises when the initial position coincides with one of the targets, i.e., ∣ x 0 ⟩ = ∣ D 1 ⟩ |x_0\rangle= |D_1\rangle ∣ x 0 ⟩ = ∣ D 1 ⟩ or ∣ x 0 ⟩ = ∣ D 2 ⟩ |x_0\rangle= |D_2\rangle ∣ x 0 ⟩ = ∣ D 2 ⟩ .
A CASE STUDY: N = 5 N = 5 N = 5
As a concrete example, we apply our framework to the nearest-neighbor tight-binding Hamiltonian in Eq. (2) for N = 5 N = 5 N = 5 , and compute the splitting probabilities, and the resonance points. Following Eq. (55) , the eigenvalues of H H H in Eq. (2) are { − 3 γ , − γ , 0 , γ , 3 γ } \{-\sqrt{3}\gamma,-\gamma,0,\gamma,\sqrt{3}\gamma\} { − 3 γ , − γ , 0 , γ , 3 γ } , and its eigenstates are ∣ E ℓ ⟩ = ( sin k ℓ , sin 2 k ℓ , sin 3 k ℓ , sin 4 k ℓ , sin 5 k ℓ ) T / 3 |E_\ell\rangle= (\sin k_\ell,\sin2k_\ell,\sin3k_\ell,\sin4k_\ell,\sin5k_\ell)^T/\sqrt{3} ∣ E ℓ ⟩ = ( sin k ℓ , sin 2 k ℓ , sin 3 k ℓ , sin 4 k ℓ , sin 5 k ℓ ) T / 3 with k ℓ = π ℓ / 6 k_\ell= \pi\ell/6 k ℓ = π ℓ /6 (T T T is the transpose). Observe that all eigenvalues are distinct. Furthermore, recall that the measurements are made at the boundary sites, ∣ D 1 ⟩ = ∣ 1 ⟩ |D_1\rangle= |1\rangle ∣ D 1 ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ 5 ⟩ |D_2\rangle= |5\rangle ∣ D 2 ⟩ = ∣5 ⟩ . Using Eqs. [57 ] –[58 ] , we have
∣ d + ⟩ = 1 6 ∣ E 1 ⟩ + 2 3 ∣ E 3 ⟩ + 1 6 ∣ E 5 ⟩ , (107) |d_+\rangle= \frac{1}{\sqrt{6}}|E_1\rangle+ \sqrt{\frac{2}{3}}|E_3\rangle+ \frac{1}{\sqrt{6}}|E_5\rangle, \tag*{(107)} ∣ d + ⟩ = 6 1 ∣ E 1 ⟩ + 3 2 ∣ E 3 ⟩ + 6 1 ∣ E 5 ⟩ , (107)
∣ d − ⟩ = 1 2 ∣ E 2 ⟩ + 1 2 ∣ E 4 ⟩ . (108) |d_-\rangle= \frac{1}{\sqrt{2}}|E_2\rangle+ \frac{1}{\sqrt{2}}|E_4\rangle. \tag*{(108)} ∣ d − ⟩ = 2 1 ∣ E 2 ⟩ + 2 1 ∣ E 4 ⟩ . (108)
Below, we compute the splitting probabilities for three different localized initial conditions, ∣ x 0 ⟩ = ∣ 3 ⟩ , ∣ 4 ⟩ , ∣ 5 ⟩ |x_0\rangle= |3\rangle, |4\rangle, |5\rangle ∣ x 0 ⟩ = ∣3 ⟩ , ∣4 ⟩ , ∣5 ⟩ . The other two localized initial conditions, ∣ x 0 ⟩ = ∣ 1 ⟩ , ∣ 2 ⟩ |x_0\rangle= |1\rangle, |2\rangle ∣ x 0 ⟩ = ∣1 ⟩ , ∣2 ⟩ , can be treated by exploiting the symmetry
{ P R ( 1 ) = P L ( 5 ) , P L ( 1 ) = P R ( 5 ) , P R ( 2 ) = P L ( 4 ) , P L ( 2 ) = P R ( 4 ) . (109) \begin{cases}
P_R(1) = P_L(5), \quad P_L(1) = P_R(5), \\
P_R(2) = P_L(4), \quad P_L(2) = P_R(4). \tag*{(109)}
\end{cases} { P R ( 1 ) = P L ( 5 ) , P L ( 1 ) = P R ( 5 ) , P R ( 2 ) = P L ( 4 ) , P L ( 2 ) = P R ( 4 ) . (109)
Example I: ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩
We choose to first consider the initial condition ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩ , located at the center of the chain and therefore symmetric relative to the boundaries. Eq. (56) gives
∣ 3 ⟩ = 1 3 ( ∣ E 1 ⟩ − ∣ E 3 ⟩ + ∣ E 5 ⟩ ) . (110) |3\rangle=\frac{1}{\sqrt{3}}\left(|E_1\rangle-|E_3\rangle+|E_5\rangle\right). \tag*{(110)} ∣3 ⟩ = 3 1 ( ∣ E 1 ⟩ − ∣ E 3 ⟩ + ∣ E 5 ⟩ ) . (110)
Turning to the splitting probability in Eq. (61) , we first consider the non-special τ \tau τ , for which all quasienergies are distinct. To evaluate the interference term in this expression, we use the auxiliary χ ~ − ( z ) \widetilde{\chi}_{-}(z) χ − ( z ) from Eq. (96) with g ℓ = 1 g_\ell=1 g ℓ = 1 . Since ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩ involves only odd values of ℓ \ell ℓ in Eq. (110) , we find χ ~ − ( z ) = 0 \widetilde{\chi}_{-}(z)=0 χ − ( z ) = 0 . Hence, the interference term in Eq. (61) vanishes, and we get
P L ( 3 ) = P R ( 3 ) = 1 / 2. (111) P_L(3)=P_R(3)=1/2. \tag*{(111)} P L ( 3 ) = P R ( 3 ) = 1/2. (111)
Given that ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩ is exactly in the middle of the chain, equidistant from both boundaries, the two splitting probabilities are equal to each other, as expected from the symmetry.
We next turn to special values of τ \tau τ , where two or more quasienergies become degenerate. The first case is τ = 0 \tau=0 τ = 0 , in which all quasienergies collapse to zero. As discussed in Eq. (70) , the walker is completely restricted to the interior of the chain, and is never detected at the boundaries, i.e., P L ( 3 ) = P R ( 3 ) = 0 P_L(3)=P_R(3)=0 P L ( 3 ) = P R ( 3 ) = 0 .
The second case corresponds to the special τ \tau τ where the phases coincide e − i E 1 τ = e − i E 3 τ e^{-iE_1\tau}=e^{-iE_3\tau} e − i E 1 τ = e − i E 3 τ . Since E 3 = 0 E_3=0 E 3 = 0 and E 5 = − E 1 = 3 γ E_5=-E_1=\sqrt{3}\gamma E 5 = − E 1 = 3 γ , this also implies e − i E 5 τ = e − i E 3 τ e^{-iE_5\tau}=e^{-iE_3\tau} e − i E 5 τ = e − i E 3 τ ; see Figure 2 (b). The corresponding time interval is 3 γ τ = 2 m π \sqrt{3}\gamma\tau=2m\pi 3 γ τ = 2 mπ with m = 1 , 2 , … m=1,2,\ldots m = 1 , 2 , … . At these times, the initial state ∣ x 0 ⟩ = ∣ 3 ⟩ |x_0\rangle=|3\rangle ∣ x 0 ⟩ = ∣3 ⟩ is fully revived [see Eq. (110) ], so the walker is never detected at the boundaries. Consequently, the splitting probabilities vanish, P L ( 3 ) = P R ( 3 ) = 0 P_L(3)=P_R(3)=0 P L ( 3 ) = P R ( 3 ) = 0 .
The third case arises when e − i E 1 τ = e − i E 5 τ e^{-iE_1\tau}=e^{-iE_5\tau} e − i E 1 τ = e − i E 5 τ but they are not equal to unity. In this situation, the dark state from Eq. (99) takes the form ∣ D 15 ( + ) ⟩ ∼ ( ∣ E 1 ⟩ − ∣ E 5 ⟩ ) |D_{15}^{(+)}\rangle\sim(|E_1\rangle-|E_5\rangle) ∣ D 15 ( + ) ⟩ ∼ ( ∣ E 1 ⟩ − ∣ E 5 ⟩) , and Eq. (110) gives ⟨ 3 ∣ D 15 ( + ) ⟩ = 0 \langle3|D_{15}^{(+)}\rangle=0 ⟨ 3∣ D 15 ( + ) ⟩ = 0 . Thus, the overlap condition in Eq. (102) is not satisfied, and no resonance occurs. For the same reason, the case e − i E 2 τ = e − i E 4 τ e^{-iE_2\tau}=e^{-iE_4\tau} e − i E 2 τ = e − i E 4 τ also does not lead to a resonance.
Collecting all cases together, these special points and the corresponding values of the probabilities are summarized in Table I. In Figure 9 (a) and Figure 9 (b), the splitting probabilities are shown as functions of τ \tau τ . The red and blue symbols
caption: Splitting detection probabilities P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) at the special values of τ \tau τ with m = 1 , 2 , … m = 1, 2, \ldots m = 1 , 2 , … for a chain of size N = 5 N = 5 N = 5 , and measurements performed at ∣ D 1 ⟩ = ∣ 1 ⟩ \lvert D_1\rangle= \lvert1\rangle ∣ D 1 ⟩ = ∣ 1 ⟩ and ∣ D 2 ⟩ = ∣ 5 ⟩ \lvert D_2\rangle= \lvert5\rangle ∣ D 2 ⟩ = ∣ 5 ⟩ . These resonant points are determined by the phase-matching and overlap conditions given in Eqs. (102) and (103) . We have shown results for the initial conditions ∣ x 0 ⟩ = ∣ 3 ⟩ \lvert x_0\rangle= \lvert3\rangle ∣ x 0 ⟩ = ∣ 3 ⟩ , ∣ 4 ⟩ \lvert4\rangle ∣ 4 ⟩ , ∣ 5 ⟩ \lvert5\rangle ∣ 5 ⟩ . The other two initial conditions, ∣ x 0 ⟩ = ∣ 1 ⟩ \lvert x_0\rangle= \lvert1\rangle ∣ x 0 ⟩ = ∣ 1 ⟩ , ∣ 2 ⟩ \lvert2\rangle ∣ 2 ⟩ , can be analyzed using the symmetry in Eq. (109) .
:::
correspond to the numerical simulations, with the blue symbols marking the resonances that occur at special values of τ \tau τ . The details of the simulations are given in Appendix B. Except at these special values, both P L ( 3 ) P_L(3) P L ( 3 ) and P R ( 3 ) P_R(3) P R ( 3 ) remain equal to 1 / 2 1/2 1/2 irrespective of τ \tau τ , confirming our theoretical result in Eq. (111) .
Example II: ∣ x 0 ⟩ = ∣ 4 ⟩ \lvert x_0\rangle= \lvert4\rangle ∣ x 0 ⟩ = ∣ 4 ⟩
We next consider the initial condition ∣ x 0 ⟩ = ∣ 4 ⟩ \lvert x_0\rangle= \lvert4\rangle ∣ x 0 ⟩ = ∣ 4 ⟩ . Unlike the previous example, this initial condition is no longer symmetric with respect to the two detectors, ∣ D 1 ⟩ = ∣ 1 ⟩ \lvert D_1\rangle= \lvert1\rangle ∣ D 1 ⟩ = ∣ 1 ⟩ and ∣ D 2 ⟩ = ∣ 5 ⟩ \lvert D_2\rangle= \lvert5\rangle ∣ D 2 ⟩ = ∣ 5 ⟩ . The initial position, being closer to the right boundary, would, in a classical random walk, naturally bias the splitting probability toward the right. This is the proximity effect that we explained in the Introduction. Here, we illustrate the breakdown of this effect for the quantum case.
For x 0 = 4 x_0 = 4 x 0 = 4 , χ ~ ± ( z ) \widetilde{\chi}_{\pm}(z) χ ± ( z ) in Eqs. (95) and (96) take the form
χ ~ + ( z ) = − i 3 2 ( 1 − z ) sin ( 3 γ τ ) z ( 1 − 2 z ) + z cos ( 3 γ τ ) ( 5 − z ) − 3 (112) \widetilde{\chi}_{+}(z)=-i\sqrt{\frac{3}{2}}\frac{(1-z)\sin(\sqrt{3}\gamma\tau)}{z(1-2z)+z\cos(\sqrt{3}\gamma\tau)(5-z)-3} \tag*{(112)} χ + ( z ) = − i 2 3 z ( 1 − 2 z ) + z cos ( 3 γ τ ) ( 5 − z ) − 3 ( 1 − z ) sin ( 3 γ τ ) (112)
χ ~ − ( z ) = − i sin ( γ τ ) 2 [ 1 − z cos ( γ τ ) ] (113) \widetilde{\chi}_{-}(z)=-\frac{i\sin(\gamma\tau)}{\sqrt{2}[1-z\cos(\gamma\tau)]} \tag*{(113)} χ − ( z ) = − 2 [ 1 − z cos ( γ τ )] i sin ( γ τ ) (113)
Inserting them in Eq. (53) , and carrying out the integration over ω \omega ω , we obtain
ζ + − ( 4 ) = − 3 sin 2 ( γ τ 2 ) sin ( γ τ ) sin ( 3 γ τ ) 4 + cos ( 2 γ τ ) + cos ( γ τ ) [ − 1 + cos ( 3 γ τ ) { − 5 + cos ( γ τ ) } ] (114) \zeta_{+-}(4)=-\frac{\sqrt{3}\sin^2\left(\frac{\gamma\tau}{2}\right)\sin(\gamma\tau)\sin(\sqrt{3}\gamma\tau)}{4+\cos(2\gamma\tau)+\cos(\gamma\tau)\left[-1+\cos(\sqrt{3}\gamma\tau)\{-5+\cos(\gamma\tau)\}\right]} \tag*{(114)} ζ +− ( 4 ) = − 4 + cos ( 2 γ τ ) + cos ( γ τ ) [ − 1 + cos ( 3 γ τ ) { − 5 + cos ( γ τ )} ] 3 sin 2 ( 2 γ τ ) sin ( γ τ ) sin ( 3 γ τ ) (114)
Finally, the splitting probabilities in Eq. (52) are
P L ( 4 ) = 1 2 − 3 sin 2 ( γ τ 2 ) sin ( γ τ ) sin ( 3 γ τ ) 4 + cos ( 2 γ τ ) + cos ( γ τ ) [ − 1 + cos ( 3 γ τ ) { − 5 + cos ( γ τ ) } ] , P R ( 4 ) = 1 − P L ( 4 ) . (115) \begin{aligned}
P_L(4)&=\frac{1}{2}-\frac{\sqrt{3}\sin^2\left(\frac{\gamma\tau}{2}\right)\sin(\gamma\tau)\sin(\sqrt{3}\gamma\tau)}{4+\cos(2\gamma\tau)+\cos(\gamma\tau)\left[-1+\cos(\sqrt{3}\gamma\tau)\{-5+\cos(\gamma\tau)\}\right]},\\
P_R(4)&=1-P_L(4). \tag*{(115)}
\end{aligned} P L ( 4 ) P R ( 4 ) = 2 1 − 4 + cos ( 2 γ τ ) + cos ( γ τ ) [ − 1 + cos ( 3 γ τ ) { − 5 + cos ( γ τ )} ] 3 sin 2 ( 2 γ τ ) sin ( γ τ ) sin ( 3 γ τ ) , = 1 − P L ( 4 ) . (115)
Both P L ( 4 ) P_L(4) P L ( 4 ) and P R ( 4 ) P_R(4) P R ( 4 ) show a clear dependence on the measurement time τ \tau τ . Nevertheless, they continue to satisfy the normalisation condition, P L ( 4 ) + P R ( 4 ) = 1 P_L(4)+P_R(4)=1 P L ( 4 ) + P R ( 4 ) = 1 , for all values of τ \tau τ , except at the special cases.
These special times, corresponding to phase matching and overlapping conditions in Eqs. (102) and (103) , are summarized in Table I. At these values, the splitting probabilities change abruptly to P L ( 4 ) = P R ( 4 ) = 1 / 4 P_L(4)=P_R(4)=1/4 P L ( 4 ) = P R ( 4 ) = 1/4 . Since P L ( 4 ) + P R ( 4 ) = 1 / 2 P_L(4)+P_R(4)=1/2 P L ( 4 ) + P R ( 4 ) = 1/2 in this case, the total detection probability is less than unity. As explained before, this occurs because a part of the wavefunction behaves as a dark state that remains decoupled from the two detectors at the moments of measurement, and therefore does not get detected. This reduces the overall detection probability.
Figure 9 (c) and Figure 9 (d) demonstrate a comparison of our theoretical results with the numerical simulations. We see excellent agreement for both P R ( 4 ) P_R(4) P R ( 4 ) and P L ( 4 ) P_L(4) P L ( 4 ) , including at the resonance points (shown by the blue symbols). From the figure, we also notice that, unlike in our previous symmetric example with ∣ x 0 ⟩ = ∣ 3 ⟩ \lvert x_0\rangle=\lvert3\rangle ∣ x 0 ⟩ = ∣ 3 ⟩ , the two splitting probabilities are generally unequal [ P L ( 4 ) ≠ P R ( 4 ) ] [P_L(4)\ne P_R(4)] [ P L ( 4 ) = P R ( 4 )] , owing to the initial state, ∣ x 0 ⟩ = ∣ 4 ⟩ \lvert x_0\rangle=\lvert4\rangle ∣ x 0 ⟩ = ∣ 4 ⟩ , being not symmetric with two detectors.
Let us now examine the proximity effect. For a classical random walk, this can be seen from Eq. (1) , which satisfies P R ( c l ) ( 4 ) > P L ( c l ) ( 4 ) P_{R}^{(\mathrm{cl})}(4) > P_{L}^{(\mathrm{cl})}(4) P R ( cl ) ( 4 ) > P L ( cl ) ( 4 ) . However, for the quantum walk, Figure 9 (c) and Figure 9 (d) reveal that this classical expectation is generally not true, and the opposite behavior P R ( 4 ) < P L ( 4 ) P_{R}(4) < P_{L}(4) P R ( 4 ) < P L ( 4 ) is observed for certain values of τ \tau τ . For instance, when τ = 2 \tau= 2 τ = 2 , we obtain P R ( 4 ) ≈ 0.28 P_{R}(4) \approx0.28 P R ( 4 ) ≈ 0.28 and P L ( 4 ) ≈ 0.72 P_{L}(4) \approx0.72 P L ( 4 ) ≈ 0.72 , clearly departing from the classical proximity effect. This departure arises because, for τ = 2 \tau= 2 τ = 2 , some components of the wave function interfere constructively at the left boundary, while interfering destructively at the right boundary. As a result, the probability of detecting the walker on the right boundary becomes lower than the left boundary, leading to the observed reversal of the classical expectation. This demonstrates that, in the quantum case, proximity to a detector does not necessarily guarantee detection at that side, and interference effects can even enhance the probability of detection at the opposite detector.
Finally, in the Zeno limit, where 0 < γ τ ≪ 1 0 < \gamma\tau\ll1 0 < γ τ ≪ 1 , one can observe in Figure 9 (c) and Figure 9 (d) that both P L ( 4 ) P_{L}(4) P L ( 4 ) and P R ( 4 ) P_{R}(4) P R ( 4 ) are close to 1 / 2 1/2 1/2 . This can also be validated by expanding Eqs. (115) and (116) for γ τ ≪ 1 \gamma\tau\ll1 γ τ ≪ 1
P R ( 4 ) ≃ 1 2 + γ 2 τ 2 8 , P L ( 4 ) ≃ 1 2 − γ 2 τ 2 8 . (116) P_{R}(4) \simeq\frac{1}{2}+\frac{\gamma^{2}\tau^{2}}{8}, \qquad P_{L}(4) \simeq\frac{1}{2}-\frac{\gamma^{2}\tau^{2}}{8}. \tag*{(116)} P R ( 4 ) ≃ 2 1 + 8 γ 2 τ 2 , P L ( 4 ) ≃ 2 1 − 8 γ 2 τ 2 . (116)
The leading order term is 1 / 2 1/2 1/2 , which is consistent with our previous analysis of the Zeno limit for general N N N [Eq. (68) ].
Loss of the flat-to-fluctuating transition for small N N N
Beyond the Zeno-limit, we observe in Figure 9 (c) and Figure 9 (d) that both P L ( 4 ) P_{L}(4) P L ( 4 ) and P R ( 4 ) P_{R}(4) P R ( 4 ) begin to deviate from 1 / 2 1/2 1/2 as the sampling time increases. In particular, we do not observe the flat-to-fluctuating transition discussed earlier in § 7 or Figure 6 . This transition, as noted before, occurs only in the limit of large system size. For small system sizes, such as N = 5 N = 5 N = 5 , this transition is lost.
Example III: ∣ x 0 ⟩ = ∣ 5 ⟩ |x_{0}\rangle= |5\rangle ∣ x 0 ⟩ = ∣5 ⟩
Here, we consider the situation when the initial position, x 0 = 5 x_{0} = 5 x 0 = 5 , coincides with the right boundary
∣ 5 ⟩ = 1 3 [ 1 2 ∣ E 1 ⟩ − 3 2 ∣ E 2 ⟩ + ∣ E 3 ⟩ − 3 2 ∣ E 4 ⟩ + 1 2 ∣ E 5 ⟩ ] . (117) |5\rangle= \frac{1}{\sqrt{3}}\left[\frac{1}{2}|E_{1}\rangle- \frac{\sqrt{3}}{2}|E_{2}\rangle+ |E_{3}\rangle- \frac{\sqrt{3}}{2}|E_{4}\rangle+ \frac{1}{2}|E_{5}\rangle\right]. \tag*{(117)} ∣5 ⟩ = 3 1 [ 2 1 ∣ E 1 ⟩ − 2 3 ∣ E 2 ⟩ + ∣ E 3 ⟩ − 2 3 ∣ E 4 ⟩ + 2 1 ∣ E 5 ⟩ ] . (117)
For this case, the overlap conditions in Eqs. (102) and (103) are never satisfied (also see Eqs. (100) and (106) ). Consequently, we do not get any resonant points for this case.
For generic τ \tau τ , treating this initial condition in the same way as the previous two, we find that the splitting probabilities take the form
P R ( 5 ) = 1 2 + A ( γ τ ) B ( γ τ ) , P L ( 5 ) = 1 − P R ( 5 ) , (118) P_{R}(5) = \frac{1}{2} + \frac{A(\gamma\tau)}{B(\gamma\tau)}, \qquad P_{L}(5) = 1 - P_{R}(5), \tag*{(118)} P R ( 5 ) = 2 1 + B ( γ τ ) A ( γ τ ) , P L ( 5 ) = 1 − P R ( 5 ) , (118)
with A ( y ) A(y) A ( y ) and B ( y ) B(y) B ( y ) defined as
A ( y ) = 33 cos ( y ) − 6 ( 2 + cos ( 2 y ) ) + 9 cos ( 3 y ) + 2 cos ( 2 3 y ) ( cos ( y ) − 5 ) A(y) = 33\cos(y) - 6\left(2+\cos(2y)\right) + 9\cos(3y) + 2\cos(2\sqrt{3}y)\left(\cos(y)-5\right) A ( y ) = 33 cos ( y ) − 6 ( 2 + cos ( 2 y ) ) + 9 cos ( 3 y ) + 2 cos ( 2 3 y ) ( cos ( y ) − 5 )
− 2 cos ( 3 y ) { 7 − 14 cos ( y ) + 15 cos ( 2 y ) } , (119) {} \qquad- 2\cos(\sqrt{3}y)\left\{7-14\cos(y)+15\cos(2y)\right\}, \tag*{(119)} − 2 cos ( 3 y ) { 7 − 14 cos ( y ) + 15 cos ( 2 y ) } , (119)
B ( y ) = 24 [ 4 + cos ( 2 y ) + cos ( y ) { cos ( 3 y ) ( cos ( y ) − 5 ) − 1 } ] . (120) B(y) = 24\left[4+\cos(2y)+\cos(y)\left\{\cos(\sqrt{3}y)\left(\cos(y)-5\right)-1\right\}\right]. \tag*{(120)} B ( y ) = 24 [ 4 + cos ( 2 y ) + cos ( y ) { cos ( 3 y ) ( cos ( y ) − 5 ) − 1 } ] . (120)
Expansion for small y y y gives A ( y ) ≃ 72 y 2 − 108 y 4 A(y) \simeq72y^{2}-108y^{4} A ( y ) ≃ 72 y 2 − 108 y 4 and B ( y ) ≃ 144 y 2 − 72 y 4 B(y) \simeq144y^{2}-72y^{4} B ( y ) ≃ 144 y 2 − 72 y 4 . From these, the splitting probabilities simplify to
P R ( 5 ) ≃ 1 − γ 2 τ 2 2 , P L ( 5 ) ≃ γ 2 τ 2 2 , (121) P_{R}(5) \simeq1-\frac{\gamma^{2}\tau^{2}}{2}, \qquad P_{L}(5) \simeq\frac{\gamma^{2}\tau^{2}}{2}, \tag*{(121)} P R ( 5 ) ≃ 1 − 2 γ 2 τ 2 , P L ( 5 ) ≃ 2 γ 2 τ 2 , (121)
in the Zeno limit of γ τ ≪ 1 \gamma\tau\ll1 γ τ ≪ 1 . To the leading order, the walker, as expected, is detected with unit probability at the right detector, since the initial position coincides with this detector. This result is again consistent with our general Zeno-analysis in Eq. (68) .
In Figure 9 (e) and Figure 9 (f), we use Eq. (118) to plot P R ( 5 ) P_R(5) P R ( 5 ) and P L ( 5 ) P_L(5) P L ( 5 ) as functions of τ \tau τ , together with the numerical simulations, see Appendix § 15 for the simulation details. Once again, our theoretical predictions show excellent agreement with the simulation data, with no resonances. As seen in the figure, these probabilities display pronounced oscillations, and at certain values of τ \tau τ , the splitting probability at the right boundary can even fall below the left one. For example, at γ τ = π \gamma\tau=\pi γ τ = π , we find P R ( 5 ) ≈ 0.055 P_R(5)\approx0.055 P R ( 5 ) ≈ 0.055 while P L ( 5 ) ≈ 0.945 P_L(5)\approx0.945 P L ( 5 ) ≈ 0.945 . The behavior P L ( 5 ) > P R ( 5 ) P_L(5)>P_R(5) P L ( 5 ) > P R ( 5 ) again reflects the loss of the classical proximity effect. It highlights that the quantum interference effect can modify the splitting probabilities in ways that have no classical analog.
Coherent Initial States
Having analyzed in detail the case of localized initial states in the tight-binding chain with Hamiltonian in Eq. (54) , we now investigate more general ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ in which the walker initially can be in a coherent superposition over several sites. For localized initial conditions, we saw in Figure 6 that the splitting probabilities exhibit flat-to-fluctuating transition. We now investigate it for general ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ .
Using Eqs. (61) –(63) , we first note that if ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ has odd parity, then ⟨ E ℓ ∣ ψ 0 ⟩ = 0 \langle E_\ell\vert\psi_0\rangle=0 ⟨ E ℓ ∣ ψ 0 ⟩ = 0 for odd ℓ \ell ℓ , since ∣ E ℓ ⟩ \lvert E_\ell\rangle ∣ E ℓ ⟩ then has even parity [see Eq. (55) ]. Substituting this into Eq. (63) , we find I ( ψ 0 , z ) = 0 I(\psi_0,z)=0 I ( ψ 0 , z ) = 0 , which according to Eq. (62) leads to ζ + − ( ψ 0 ) = 0 \zeta_{+-}(\psi_0)=0 ζ +− ( ψ 0 ) = 0 . Similarly, for ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ with even parity, one has ζ + − ( ψ 0 ) = 0 \zeta_{+-}(\psi_0)=0 ζ +− ( ψ 0 ) = 0 . As a result, Eq. (61) reduces to
P L ( ψ 0 ) = P R ( ψ 0 ) = 1 / 2 , for ∣ ψ 0 ⟩ of definite parity . (122) P_L(\psi_0)=P_R(\psi_0)=1/2,\qquad\text{for }\lvert\psi_0\rangle\text{ of definite parity}. \tag*{(122)} P L ( ψ 0 ) = P R ( ψ 0 ) = 1/2 , for ∣ ψ 0 ⟩ of definite parity . (122)
For these initial states, the splitting probabilities therefore are independent of the sampling time τ \tau τ , and remain equal to 1 / 2 1/2 1/2 for all non-exceptional sampling times. Thus, the flat behavior observed below the critical sampling time for localized initial conditions in Figure 6 extends to the entire non-exceptional values of τ \tau τ for ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ of definite parity.
We also noted in Eq. (68) that, in the Zeno limit, the splitting probabilities are equal to 1 / 2 1/2 1/2 for all initial conditions localized in the interior of the chain. More generally, as indicated in Eq. (11) , the Zeno-limit splitting probabilities depend only on the projections of ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ onto the two boundary sites, and are independent of its detailed structure in the interior of the chain. Therefore, whenever ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ satisfies ∣ ⟨ 1 ∣ ψ 0 ⟩ ∣ 2 = ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 \lvert\langle1\vert\psi_0\rangle\rvert^2=\lvert\langle N\vert\psi_0\rangle\rvert^2 ∣⟨ 1∣ ψ 0 ⟩ ∣ 2 = ∣⟨ N ∣ ψ 0 ⟩ ∣ 2 , Eq. (11) implies that both splitting probabilities are equal to 1 / 2 1/2 1/2 in the Zeno limit. Thus, the Zeno-limit result, P L ( ψ 0 ) ≃ P R ( ψ 0 ) ≃ 1 / 2 P_L(\psi_0)\simeq P_R(\psi_0)\simeq1/2 P L ( ψ 0 ) ≃ P R ( ψ 0 ) ≃ 1/2 , applies to a broad class of initial states beyond the localized ones.
We next turn to the role of coherence in the initial state in shaping the splitting probabilities. As seen above, initial states of definite parity lead to a splitting probability of 1 / 2 1/2 1/2 for either target for generic sampling times. A general coherent superposition, however, need not possess definite parity, and the relative phase between its components can therefore influence the splitting probabilities. To examine this effect, we consider the initial state
∣ ψ 0 ⟩ = ∣ x 0 ⟩ + e i Φ ∣ x 0 + d 0 ⟩ 2 , (123) \lvert\psi_0\rangle=\frac{\lvert x_0\rangle+e^{i\Phi}\lvert x_0+d_0\rangle}{\sqrt{2}}, \tag*{(123)} ∣ ψ 0 ⟩ = 2 ∣ x 0 ⟩ + e i Φ ∣ x 0 + d 0 ⟩ , (123)
where Φ ∈ [ 0 , 2 π ] \Phi\in[0,2\pi] Φ ∈ [ 0 , 2 π ] is the relative phase between the two components, while recall 1 ≤ x 0 ≤ N 1\leq x_0\leq N 1 ≤ x 0 ≤ N and 1 ≤ d 0 ≤ ( N − x 0 ) 1\leq d_0\leq(N-x_0) 1 ≤ d 0 ≤ ( N − x 0 ) . Using Eqs. (61) –(63) , we calculate P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) for this initial condition and plot its dependence on the phase Φ \Phi Φ in Figure 10 (a). We have used N = 50 N=50 N = 50 , γ = 1 \gamma=1 γ = 1 , τ = 1 \tau=1 τ = 1 , x 0 = 10 x_0=10 x 0 = 10 , and chosen three separations between the two components, namely d 0 = 1 d_0=1 d 0 = 1 , 2 2 2 , and 3 3 3 . As shown in Figure 10 (a), the dependence of P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) on Φ \Phi Φ is strongest for d 0 = 1 d_0=1 d 0 = 1 and becomes considerably weaker for d 0 = 3 d_0=3 d 0 = 3 . For the even separation d 0 = 2 d_0=2 d 0 = 2 , the splitting probability is essentially independent of Φ \Phi Φ , remaining 1 / 2 1/2 1/2 over the entire range of Φ \Phi Φ . This behavior qualitatively persists for other choices of the parameters also. For example, Figure 10 (b) shows the corresponding results for τ = 1.2 \tau=1.2 τ = 1.2 and x 0 = 17 x_0=17 x 0 = 17 , again exhibiting an essentially Φ \Phi Φ -independent splitting probability for even d 0 d_0 d 0 , while for odd d 0 d_0 d 0 the phase dependence of P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) becomes markedly weaker as d 0 d_0 d 0 increases.
To understand this behavior, let us take the initial state in Eq. (123) , and plug it in Eq. (5) with ∣ D 1 ⟩ = ∣ 1 ⟩ \lvert D_1\rangle=\lvert1\rangle ∣ D 1 ⟩ = ∣ 1 ⟩ and ∣ D 2 ⟩ = ∣ N ⟩ \lvert D_2\rangle=\lvert N\rangle ∣ D 2 ⟩ = ∣ N ⟩ . The resulting first-detection amplitude can be expressed as
ϕ n ( L ) ( ψ 0 ) = 1 2 [ ϕ n ( L ) ( x 0 ) + e i Φ ϕ n ( L ) ( x 0 + d 0 ) ] . (124) \phi_n^{(L)}(\psi_0)=\frac{1}{\sqrt{2}}\left[\phi_n^{(L)}(x_0)+e^{i\Phi}\phi_n^{(L)}(x_0+d_0)\right]. \tag*{(124)} ϕ n ( L ) ( ψ 0 ) = 2 1 [ ϕ n ( L ) ( x 0 ) + e i Φ ϕ n ( L ) ( x 0 + d 0 ) ] . (124)
The analysis for the right boundary proceeds similarly. From Eq. (124) , the corresponding splitting probability is
P L ( ψ 0 ) = 1 2 [ P L ( x 0 ) + P L ( x 0 + d 0 ) ] + Re [ ∑ n = 1 ∞ ( ϕ n ( L ) ( x 0 ) ) ∗ ϕ n ( L ) ( x 0 + d 0 ) e i Φ ] . (125) P_L(\psi_0)=\frac{1}{2}\left[P_L(x_0)+P_L(x_0+d_0)\right]+\operatorname{Re}\left[\sum_{n=1}^{\infty}\left(\phi_n^{(L)}(x_0)\right)^*\phi_n^{(L)}(x_0+d_0)e^{i\Phi}\right]. \tag*{(125)} P L ( ψ 0 ) = 2 1 [ P L ( x 0 ) + P L ( x 0 + d 0 ) ] + Re [ n = 1 ∑ ∞ ( ϕ n ( L ) ( x 0 ) ) ∗ ϕ n ( L ) ( x 0 + d 0 ) e i Φ ] . (125)
where recall P L ( x 0 ) P_L(x_0) P L ( x 0 ) denotes the splitting probability for a walker initially localized at site x 0 x_0 x 0 , and similarly for P L ( x 0 + d 0 ) P_L(x_0+d_0) P L ( x 0 + d 0 ) . As discussed before in Figure 3 , for large system sizes and sampling times τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , where τ c \tau_c τ c is given in Eq. (12) , the splitting probabilities for localized initial states are equal to 1 / 2 1/2 1/2 . Hence, we rewrite Eq. (125) as
P L ( ψ 0 ) = 1 2 + Re [ ∑ n = 1 ∞ ( ϕ n ( L ) ( x 0 ) ) ∗ ϕ n ( L ) ( x 0 + d 0 ) e i Φ ] . (126) P_L(\psi_0)=\frac{1}{2}+\operatorname{Re}\left[\sum_{n=1}^{\infty}\left(\phi_n^{(L)}(x_0)\right)^*\phi_n^{(L)}(x_0+d_0)e^{i\Phi}\right]. \tag*{(126)} P L ( ψ 0 ) = 2 1 + Re [ n = 1 ∑ ∞ ( ϕ n ( L ) ( x 0 ) ) ∗ ϕ n ( L ) ( x 0 + d 0 ) e i Φ ] . (126)
where we are looking at τ ≤ τ c \tau\leq\tau_c τ ≤ τ c and N ≫ 1 N\gg1 N ≫ 1 . For the nearest-neighbor hopping Hamiltonian, we show in Appendix E that
[ ϕ n ( L ) ( x 0 ) ] ∗ = ( − 1 ) 1 + x 0 ϕ n ( L ) ( x 0 ) . (127) \left[\phi_n^{(L)}(x_0)\right]^*=(-1)^{1+x_0}\phi_n^{(L)}(x_0). \tag*{(127)} [ ϕ n ( L ) ( x 0 ) ] ∗ = ( − 1 ) 1 + x 0 ϕ n ( L ) ( x 0 ) . (127)
This means ϕ n ( L ) ( x 0 ) \phi_n^{(L)}(x_0) ϕ n ( L ) ( x 0 ) is real for odd x 0 x_0 x 0 , and purely imaginary for even x 0 x_0 x 0 . Using this result in Eq. (126) , we find ( ϕ n ( L ) ( x 0 ) ) ∗ ϕ n ( L ) ( x 0 + d 0 ) \left(\phi_n^{(L)}(x_0)\right)^*\phi_n^{(L)}(x_0+d_0) ( ϕ n ( L ) ( x 0 ) ) ∗ ϕ n ( L ) ( x 0 + d 0 ) is real for even d 0 d_0 d 0 , and purely imaginary for odd d 0 d_0 d 0 , independently of $x_0
P L ( ψ 0 ) = { 1 / 2 + K d 0 ( x 0 , τ ) sin Φ , d 0 odd , 1 / 2 + K d 0 ( x 0 , τ ) cos Φ , d 0 even . (128) P_L(\psi_0)=
\begin{cases}
1/2+\mathcal{K}_{d_0}(x_0,\tau)\sin\Phi, & d_0\ \text{odd},\\
1/2+\mathcal{K}_{d_0}(x_0,\tau)\cos\Phi, & d_0\ \text{even}.
\end{cases} \tag*{(128)} P L ( ψ 0 ) = { 1/2 + K d 0 ( x 0 , τ ) sin Φ , 1/2 + K d 0 ( x 0 , τ ) cos Φ , d 0 odd , d 0 even . (128)
where
K d 0 ( x 0 , τ ) = { i ( − 1 ) 1 + x 0 ∑ n = 1 ∞ ϕ n ( L ) ( x 0 ) ϕ n ( L ) ( x 0 + d 0 ) , d 0 odd , ( − 1 ) 1 + x 0 ∑ n = 1 ∞ ϕ n ( L ) ( x 0 ) ϕ n ( L ) ( x 0 + d 0 ) , d 0 even . (129) \mathcal{K}_{d_0}(x_0,\tau)=
\begin{cases}
i(-1)^{1+x_0}\sum_{n=1}^{\infty}\phi_n^{(L)}(x_0)\phi_n^{(L)}(x_0+d_0), & d_0\ \text{odd},\\
(-1)^{1+x_0}\sum_{n=1}^{\infty}\phi_n^{(L)}(x_0)\phi_n^{(L)}(x_0+d_0), & d_0\ \text{even}.
\end{cases} \tag*{(129)} K d 0 ( x 0 , τ ) = { i ( − 1 ) 1 + x 0 ∑ n = 1 ∞ ϕ n ( L ) ( x 0 ) ϕ n ( L ) ( x 0 + d 0 ) , ( − 1 ) 1 + x 0 ∑ n = 1 ∞ ϕ n ( L ) ( x 0 ) ϕ n ( L ) ( x 0 + d 0 ) , d 0 odd , d 0 even . (129)
Eq. (128) was announced before in Eq. (14) . Thus, for odd separations d 0 d_0 d 0 , the splitting probability in Eq. (128) exhibits a sin Φ \sin\Phi sin Φ dependence on the relative phase Φ \Phi Φ in the initial state ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ in Eq. (123) . This is in agreement with the behavior observed in Figure 10 (a) and (Figure 10 b) for d 0 = 1 d_0=1 d 0 = 1 and d 0 = 3 d_0=3 d 0 = 3 . However, for d 0 = 2 d_0=2 d 0 = 2 , P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) in these figures is essentially independent of Φ \Phi Φ , even though Eq. (128) predicts a cos Φ \cos\Phi cos Φ dependence for even d 0 d_0 d 0 . This apparent discrepancy arises because the prefactor K d 0 ( x 0 , τ ) \mathcal{K}_{d_0}(x_0,\tau) K d 0 ( x 0 , τ ) multiplying the cos Φ \cos\Phi cos Φ term in Eq. (128) is extremely small for even d 0 d_0 d 0 . For example, for the parameters used in Figure 10 (a), namely N = 50 N=50 N = 50 , γ = 1 \gamma=1 γ = 1 , τ = 1 \tau=1 τ = 1 , and x 0 = 10 x_0=10 x 0 = 10 , we find K d 0 ( x 0 , τ ) ≈ − 3 × 10 − 4 \mathcal{K}_{d_0}(x_0,\tau)\approx-3\times10^{-4} K d 0 ( x 0 , τ ) ≈ − 3 × 1 0 − 4 for d 0 = 2 d_0=2 d 0 = 2 . Similarly, for the parameters of Figure 10 (b), namely N = 50 N=50 N = 50 , γ = 1 \gamma=1 γ = 1 , τ = 1.2 \tau=1.2 τ = 1.2 , and x 0 = 17 x_0=17 x 0 = 17 , we obtain K d 0 ( x 0 , τ ) ≈ − 1.2 × 10 − 4 \mathcal{K}_{d_0}(x_0,\tau)\approx-1.2\times10^{-4} K d 0 ( x 0 , τ ) ≈ − 1.2 × 1 0 − 4 for d 0 = 2 d_0=2 d 0 = 2 . Inserting these values into Eq. (128) , we find P L ( ψ 0 ) ≈ 1 / 2 P_L(\psi_0)\approx1/2 P L ( ψ 0 ) ≈ 1/2 for both cases. Hence for ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ with even d 0 d_0 d 0 , the splitting probability is 1 / 2 1/2 1/2 in the τ ≤ τ c \tau\leq\tau_c τ ≤ τ c regime. We also verify this behavior in Figure 10 (c), where we plot P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) as a function of γ τ \gamma\tau γ τ for Φ = 1 \Phi=1 Φ = 1 and even separations d 0 = 2 , 4 d_0=2,4 d 0 = 2 , 4 , and 6 6 6 . In all three cases, P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) remains essentially equal to 1 / 2 1/2 1/2 for τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , while for τ > τ c \tau>\tau_c τ > τ c it develops pronounced fluctuations. Thus, the flat-to-fluctuating transition at τ c \tau_c τ c , found previously for localized initial conditions in Figure 6 , persists also for ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ with even separation d 0 d_0 d 0 .
By contrast, K d 0 ( x 0 , τ ) \mathcal{K}_{d_0}(x_0,\tau) K d 0 ( x 0 , τ ) can be significant for odd values of d 0 d_0 d 0 , especially for smaller values of d 0 d_0 d 0 . For example, for the parameter set in Figure 10 (a), we obtain K d 0 ( x 0 , τ ) ≈ − 0.215 \mathcal{K}_{d_0}(x_0,\tau)\approx-0.215 K d 0 ( x 0 , τ ) ≈ − 0.215 for d 0 = 1 d_0=1 d 0 = 1 and K d 0 ( x 0 , τ ) ≈ − 0.017 \mathcal{K}_{d_0}(x_0,\tau)\approx-0.017 K d 0 ( x 0 , τ ) ≈ − 0.017 for d 0 = 3 d_0=3 d 0 = 3 , while for Figure 10 (b), we have K d 0 ( x 0 , τ ) ≈ − 0.24 \mathcal{K}_{d_0}(x_0,\tau)\approx-0.24 K d 0 ( x 0 , τ ) ≈ − 0.24 for d 0 = 1 d_0=1 d 0 = 1 and K d 0 ( x 0 , τ ) ≈ − 0.02 \mathcal{K}_{d_0}(x_0,\tau)\approx-0.02 K d 0 ( x 0 , τ ) ≈ − 0.02 for d 0 = 3 d_0=3 d 0 = 3 . From Eq. (128) , it then follows that P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) can deviate significantly from 1 / 2 1/2 1/2 , particularly for d 0 = 1 d_0=1 d 0 = 1 . We also observed this behavior before in Figure 10 (a) and (Figure 10 b). This behavior is further illustrated in Figure 10 (d), where P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) is plotted as a function of γ τ \gamma\tau γ τ . For d 0 = 1 d_0=1 d 0 = 1 , the splitting probability already exhibits a clear dependence on τ \tau τ in the regime τ ≤ τ c \tau\leq\tau_c τ ≤ τ c , in contrast to the flat P L = 1 / 2 P_L=1/2 P L = 1/2 behavior found above for localized initial conditions. As d 0 d_0 d 0 increases, this dependence is strongly suppressed. For example, for d 0 = 5 d_0=5 d 0 = 5 , P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) remains essentially equal to 1 / 2 1/2 1/2 below τ c \tau_c τ c , and the flat-to-fluctuating transition at τ c \tau_c τ c is recovered.
Finite Number of Measurements
Thus far, we have assumed that it is possible to take an infinite number of measurements. Accordingly, the summations over n n n appearing in the expressions for the splitting probabilities in Eq. (4) extend to infinity. Yet, as mentioned earlier, only a finite number of measurements can be performed in real experiments [72 , 73 , 74 ] . For instance, starting from ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ , a typical measurement record begins with a sequence of null outcomes, represented as no, no, . . . . This sequence is truncated after a finite number of measurements, denoted by T max T_{\max} T m a x (“max” for maximum number of measurements that can be performed). If a “yes” outcome is obtained at either detector on or before the T max T_{\max} T m a x -th measurement, detection is deemed successful. Otherwise, the record consists entirely of “no” outcomes, indicating that no detection has occurred within T max T_{\max} T m a x number of measurements. The splitting probabilities are obtained by summing the first-detection probabilities, F n ( L ) ( ψ 0 ) F_n^{(L)}(\psi_0) F n ( L ) ( ψ 0 ) and F n ( R ) ( ψ 0 ) F_n^{(R)}(\psi_0) F n ( R ) ( ψ 0 ) , till the T max T_{\max} T m a x -th measurement,
P L ( ψ 0 ) = ∑ n = 1 T max F n ( L ) ( ψ 0 ) , P R ( ψ 0 ) = ∑ n = 1 T max F n ( R ) ( ψ 0 ) . (130) P_L(\psi_0)=\sum_{n=1}^{T_{\max}}F_n^{(L)}(\psi_0),\qquad
P_R(\psi_0)=\sum_{n=1}^{T_{\max}}F_n^{(R)}(\psi_0). \tag*{(130)} P L ( ψ 0 ) = n = 1 ∑ T m a x F n ( L ) ( ψ 0 ) , P R ( ψ 0 ) = n = 1 ∑ T m a x F n ( R ) ( ψ 0 ) . (130)
Our analysis in previous sections focused on T max → ∞ T_{\max}\to\infty T m a x → ∞ . It is natural to examine how a finite T max T_{\max} T m a x affects the flat-to-fluctuating transition of the splitting probability discussed in § 7 .
In Figure 11 (a)-Figure 11 (c), we have used numerical simulations to obtain P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) for the nearest-neighbor hopping Hamiltonian in Eq. (54) and plot it as a function of γ τ \gamma\tau γ τ plot for T max = 500 , 1000 , 1500 T_{\max}=500,1000,1500 T m a x = 500 , 1000 , 1500 . We choose N = 15 N=15 N = 15 and ∣ ψ 0 ⟩ = ∣ 11 ⟩ |\psi_0\rangle=|11\rangle ∣ ψ 0 ⟩ = ∣11 ⟩ . For comparison, we have also shown in black the ideal (T max → ∞ T_{\max}\to\infty T m a x → ∞ ) expression of P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) in Eq. (61) . For T max = 500 , T_{\max}=500, T m a x = 500 , shown in panel (a), noticeable deviations from the ideal result are observed, especially at smaller values of the sampling time. However, as T max T_{\max} T m a x increases, the simulation data progressively approach the black curve. Close to τ = 0 \tau=0 τ = 0 , the quantum walker requires a large number of measurement attempts before detection occurs. Imposing a finite T max T_{\max} T m a x excludes many long detection trajectories, causing P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) to deviate from the ideal flat value of 1 / 2 1/2 1/2 in Eq. (68) close to τ = 0 \tau=0 τ = 0 . This deviation is gradually reduced as T max T_{\max} T m a x is increased.
Figure 11. Effect of a finite number of measurements, T max T_{\max} T m a x , on the splitting probability. Panels (a-c) show P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) as a function of γ τ \gamma\tau γ τ for T max = 500 T_{\max}=500 T m a x = 500 , 1000, and 1500, respectively. The red dots are numerical simulations based on the nearest-neighbor hopping Hamiltonian in Eq. (54) , with N = 15 N=15 N = 15 and ∣ ψ 0 ⟩ = ∣ 11 ⟩ |\psi_0\rangle=|11\rangle ∣ ψ 0 ⟩ = ∣11 ⟩ . The black curve is the ideal theory (T max → ∞ T_{\max}\to\infty T m a x → ∞ ) for P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) given by Eq. (61) . Finite-T max T_{\max} T m a x effects are most pronounced near resonances such as τ = 0 \tau=0 τ = 0 , where detection typically requires many measurement attempts. Panels (d-f) show the corresponding normalization, P L ( ψ 0 ) + P R ( ψ 0 ) P_L(\psi_0)+P_R(\psi_0) P L ( ψ 0 ) + P R ( ψ 0 ) , for the same three values of T max T_{\max} T m a x . The numerical simulations are compared with the finite-T max T_{\max} T m a x theory in Eq. (148) . Recall that in the ideal limit, T max → ∞ T_{\max}\to\infty T m a x → ∞ , the normalization is mostly equal to unity with pointwise discontinuous jumps at the special sampling times in Eq. (10) , where P L ( ψ 0 ) + P R ( ψ 0 ) < 1 P_L(\psi_0)+P_R(\psi_0)<1 P L ( ψ 0 ) + P R ( ψ 0 ) < 1 . For finite T max T_{\max} T m a x , we see in panels (d-f) that these discontinuities are broadened into dips of finite width. Except around Zeno limit, our Eq. (148) agrees very well with the simulation results, and the dips gradually become narrower as T max T_{\max} T m a x increases. Deviations from the finite-T max T_{\max} T m a x theory appear close to the Zeno limit, where many quasienergies can merge simultaneously. As explained in the paper, Eq. (148) is no longer valid in this regime.
Another consequence of a finite T max T_{\max} T m a x is seen in the normalization, P L ( ψ 0 ) + P R ( ψ 0 ) P_L(\psi_0)+P_R(\psi_0) P L ( ψ 0 ) + P R ( ψ 0 ) . We noted before that in the ideal T max → ∞ T_{\max}\to\infty T m a x → ∞ limit, the normalization is mostly equal to unity as a function of the sampling time, with pointwise discontinuous jumps at the special sampling times given in Eq. (10) . We explained that these jumps are related to the formation of the dark states in the system. We now ask how these pointwise discontinuous jumps are modified for a finite T max T_{\max} T m a x . In Figure 11 (d)-Figure 11 (f), we use numerical simulations to obtain P L ( ψ 0 ) + P R ( ψ 0 ) P_L(\psi_0)+P_R(\psi_0) P L ( ψ 0 ) + P R ( ψ 0 ) vs γ τ \gamma\tau γ τ for T max = 500 T_{\max}=500 T m a x = 500 , 1000, 1500 and N = 15 N=15 N = 15 , ∣ ψ 0 ⟩ = ∣ 11 ⟩ |\psi_0\rangle=|11\rangle ∣ ψ 0 ⟩ = ∣11 ⟩ . Interestingly, for finite T max T_{\max} T m a x , the pointwise discontinuities are broadened into dips of finite width. This broadening is especially pronounced for smaller T max T_{\max} T m a x , while increasing T max T_{\max} T m a x makes the dips narrower and drives the normalization toward the ideal result. Broadening effect associated with a finite number of measurements has also been observed in single-target detection problem [73 , 74 ] . We now proceed to derive P L ( ψ 0 ) + P R ( ψ 0 ) P_L(\psi_0)+P_R(\psi_0) P L ( ψ 0 ) + P R ( ψ 0 ) for a finite T max T_{\max} T m a x .
Normalization for finite T max T_{\max} T m a x . Let us consider the one-dimensional lattice chain, with parity symmetric Hamiltonian H H H and target states at the boundary sites, ∣ D 1 ⟩ = ∣ 1 ⟩ |D_1\rangle=|1\rangle ∣ D 1 ⟩ = ∣1 ⟩ and ∣ D 2 ⟩ = ∣ N ⟩ |D_2\rangle=|N\rangle ∣ D 2 ⟩ = ∣ N ⟩ . The survival probability that the quantum walker has not been detected at either target after n n n measurements is given by [50 , 53 ]
q n ( ψ 0 ) = ⟨ ψ 0 ∣ ( S † ) n S n ∣ ψ 0 ⟩ , (131) q_n(\psi_0)=\langle\psi_0|(\mathcal{S}^{\dagger})^n\mathcal{S}^n|\psi_0\rangle, \tag*{(131)} q n ( ψ 0 ) = ⟨ ψ 0 ∣ ( S † ) n S n ∣ ψ 0 ⟩ , (131)
where remember S = ( 1 − D ^ ) U ( τ ) \mathcal{S}=(1-\hat{D})U(\tau) S = ( 1 − D ^ ) U ( τ ) is the survival operator with D ^ \hat{D} D ^ given in Eq. (21) and the unitary operator U ( τ ) U(\tau) U ( τ ) in Eq. (3) . Using Eq. (131) , we get
q n − 1 ( ψ 0 ) = ⟨ ψ 0 ∣ ( S † ) n − 1 U † D ^ U S n − 1 ∣ ψ 0 ⟩ + ⟨ ψ 0 ∣ ( S † ) n − 1 U † ( 1 − D ^ ) U S n − 1 ∣ ψ 0 ⟩ . (132) q_{n-1}(\psi_0)=\langle\psi_0|(\mathcal{S}^{\dagger})^{n-1}U^{\dagger}\hat{D}U\mathcal{S}^{n-1}|\psi_0\rangle+\langle\psi_0|(\mathcal{S}^{\dagger})^{n-1}U^{\dagger}(1-\hat{D})U\mathcal{S}^{n-1}|\psi_0\rangle. \tag*{(132)} q n − 1 ( ψ 0 ) = ⟨ ψ 0 ∣ ( S † ) n − 1 U † D ^ U S n − 1 ∣ ψ 0 ⟩ + ⟨ ψ 0 ∣ ( S † ) n − 1 U † ( 1 − D ^ ) U S n − 1 ∣ ψ 0 ⟩ . (132)
Following the definition of first-detection probabilities from Eq. (27) , one can recast Eq. (132) as
F n ( x L ) ( ψ 0 ) + F n ( x R ) ( ψ 0 ) = q n − 1 ( ψ 0 ) − q n ( ψ 0 ) . (133) F_n^{(x_L)}(\psi_0) + F_n^{(x_R)}(\psi_0) = q_{n-1}(\psi_0) - q_n(\psi_0). \tag*{(133)} F n ( x L ) ( ψ 0 ) + F n ( x R ) ( ψ 0 ) = q n − 1 ( ψ 0 ) − q n ( ψ 0 ) . (133)
Summing this equation from n = 1 n=1 n = 1 to n = T max n=T_{\max} n = T m a x and using Eq. (130) , we get
P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − q T max ( ψ 0 ) . (134) P_L(\psi_0) + P_R(\psi_0) = 1 - q_{T_{\max}}(\psi_0). \tag*{(134)} P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − q T m a x ( ψ 0 ) . (134)
From Eq. (131) , it then follows
P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − ⟨ ψ 0 ∣ ( S † ) T max S T max ∣ ψ 0 ⟩ . (135) P_L(\psi_0) + P_R(\psi_0) = 1 - \langle\psi_0|(S^\dagger)^{T_{\max}}S^{T_{\max}}|\psi_0\rangle. \tag*{(135)} P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − ⟨ ψ 0 ∣ ( S † ) T m a x S T m a x ∣ ψ 0 ⟩ . (135)
We saw before that for the parity-symmetric Hamiltonian, the dual-target splitting problem decomposes into two parity-specific auxiliary single-target detection problems with target states ∣ d ± ⟩ = ( ∣ 1 ⟩ ± ∣ N ⟩ ) / 2 |d_{\pm}\rangle=(|1\rangle\pm|N\rangle)/\sqrt{2} ∣ d ± ⟩ = ( ∣1 ⟩ ± ∣ N ⟩) / 2 . Eq. (135) can be expressed as
P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − ⟨ ψ 0 ( + ) ∣ ( S + † ) T max S + T max ∣ ψ 0 ( + ) ⟩ − ⟨ ψ 0 ( − ) ∣ ( S − † ) T max S − T max ∣ ψ 0 ( − ) ⟩ . (136) P_L(\psi_0) + P_R(\psi_0) = 1 - \langle\psi_0^{(+)}|(S_+^\dagger)^{T_{\max}}S_+^{T_{\max}}|\psi_0^{(+)}\rangle- \langle\psi_0^{(-)}|(S_-^\dagger)^{T_{\max}}S_-^{T_{\max}}|\psi_0^{(-)}\rangle. \tag*{(136)} P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − ⟨ ψ 0 ( + ) ∣ ( S + † ) T m a x S + T m a x ∣ ψ 0 ( + ) ⟩ − ⟨ ψ 0 ( − ) ∣ ( S − † ) T m a x S − T m a x ∣ ψ 0 ( − ) ⟩ . (136)
Here S ± = ( 1 − ∣ d ± ⟩ ⟨ d ± ∣ ) U ( τ ) S_{\pm}=(1-|d_{\pm}\rangle\langle d_{\pm}|)U(\tau) S ± = ( 1 − ∣ d ± ⟩ ⟨ d ± ∣ ) U ( τ ) are the single-target survival operators and ∣ ψ 0 ( ± ) ⟩ |\psi_0^{(\pm)}\rangle ∣ ψ 0 ( ± ) ⟩ are the even and odd components of the initial state ∣ ψ 0 ⟩ |\psi_0\rangle ∣ ψ 0 ⟩ , defined as
∣ ψ 0 ( ± ) ⟩ = ( 1 ± P 2 ) ∣ ψ 0 ⟩ , (137) |\psi_0^{(\pm)}\rangle=\left(\frac{1\pm\mathcal{P}}{2}\right)|\psi_0\rangle, \tag*{(137)} ∣ ψ 0 ( ± ) ⟩ = ( 2 1 ± P ) ∣ ψ 0 ⟩ , (137)
with the parity operator P \mathcal{P} P in Eq. (36) . Introducing the right and the left eigenvectors of S ± S_{\pm} S ± and its eigenvalues
S ± ∣ Λ ± ( r i g h t ) ⟩ = Λ ± ∣ Λ ± ( r i g h t ) ⟩ , ⟨ Λ ± ( l e f t ) ∣ S ± = Λ ± ⟨ Λ ± ( l e f t ) ∣ , (138) S_{\pm}|\Lambda_{\pm}^{(\mathrm{right})}\rangle=\Lambda_{\pm}|\Lambda_{\pm}^{(\mathrm{right})}\rangle,\qquad
\langle\Lambda_{\pm}^{(\mathrm{left})}|S_{\pm}=\Lambda_{\pm}\langle\Lambda_{\pm}^{(\mathrm{left})}|, \tag*{(138)} S ± ∣ Λ ± ( right ) ⟩ = Λ ± ∣ Λ ± ( right ) ⟩ , ⟨ Λ ± ( left ) ∣ S ± = Λ ± ⟨ Λ ± ( left ) ∣ , (138)
the survival operators can be spectrally decomposed as
S ± = ∑ { Λ ± } Λ ± ∣ Λ ± ( r i g h t ) ⟩ ⟨ Λ ± ( l e f t ) ∣ , S ± † = ∑ { Λ ± } Λ ± ∗ ∣ Λ ± ( l e f t ) ⟩ ⟨ Λ ± ( r i g h t ) ∣ . (139) S_{\pm}=\sum_{\{\Lambda_{\pm}\}}\Lambda_{\pm}|\Lambda_{\pm}^{(\mathrm{right})}\rangle\langle\Lambda_{\pm}^{(\mathrm{left})}|,\qquad
S_{\pm}^{\dagger}=\sum_{\{\Lambda_{\pm}\}}\Lambda_{\pm}^{*}|\Lambda_{\pm}^{(\mathrm{left})}\rangle\langle\Lambda_{\pm}^{(\mathrm{right})}|. \tag*{(139)} S ± = { Λ ± } ∑ Λ ± ∣ Λ ± ( right ) ⟩ ⟨ Λ ± ( left ) ∣ , S ± † = { Λ ± } ∑ Λ ± ∗ ∣ Λ ± ( left ) ⟩ ⟨ Λ ± ( right ) ∣. (139)
Eq. (136) then becomes
P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − ∑ { Λ + } ∑ { Λ ~ + } ( Λ + ) T max ( Λ ~ + ∗ ) T max ⟨ ψ 0 ( + ) ∣ Λ ~ + ( l e f t ) ⟩ ⟨ Λ ~ + ( r i g h t ) ∣ Λ + ( r i g h t ) ⟩ ⟨ Λ + ( l e f t ) ∣ ψ 0 ( + ) ⟩ − ∑ { Λ − } ∑ { Λ ~ − } ( Λ − ) T max ( Λ ~ − ∗ ) T max ⟨ ψ 0 ( − ) ∣ Λ ~ − ( l e f t ) ⟩ ⟨ Λ ~ − ( r i g h t ) ∣ Λ − ( r i g h t ) ⟩ ⟨ Λ − ( l e f t ) ∣ ψ 0 ( − ) ⟩ . (140) \begin{aligned}
P_L(\psi_0)+P_R(\psi_0)
={}&1-\sum_{\{\Lambda_+\}}\sum_{\{\widetilde{\Lambda}_+\}}(\Lambda_+)^{T_{\max}}(\widetilde{\Lambda}_+^{*})^{T_{\max}}
\langle\psi_0^{(+)}|\widetilde{\Lambda}_+^{(\mathrm{left})}\rangle
\langle\widetilde{\Lambda}_+^{(\mathrm{right})}|\Lambda_+^{(\mathrm{right})}\rangle
\langle\Lambda_+^{(\mathrm{left})}|\psi_0^{(+)}\rangle\\
&-\sum_{\{\Lambda_-\}}\sum_{\{\widetilde{\Lambda}_-\}}(\Lambda_-)^{T_{\max}}(\widetilde{\Lambda}_-^{*})^{T_{\max}}
\langle\psi_0^{(-)}|\widetilde{\Lambda}_-^{(\mathrm{left})}\rangle
\langle\widetilde{\Lambda}_-^{(\mathrm{right})}|\Lambda_-^{(\mathrm{right})}\rangle
\langle\Lambda_-^{(\mathrm{left})}|\psi_0^{(-)}\rangle. \tag*{(140)}
\end{aligned} P L ( ψ 0 ) + P R ( ψ 0 ) = 1 − { Λ + } ∑ { Λ + } ∑ ( Λ + ) T m a x ( Λ + ∗ ) T m a x ⟨ ψ 0 ( + ) ∣ Λ + ( left ) ⟩ ⟨ Λ + ( right ) ∣ Λ + ( right ) ⟩ ⟨ Λ + ( left ) ∣ ψ 0 ( + ) ⟩ − { Λ − } ∑ { Λ − } ∑ ( Λ − ) T m a x ( Λ − ∗ ) T m a x ⟨ ψ 0 ( − ) ∣ Λ − ( left ) ⟩ ⟨ Λ − ( right ) ∣ Λ − ( right ) ⟩ ⟨ Λ − ( left ) ∣ ψ 0 ( − ) ⟩ . (140)
This is an exact result for any parity symmetric Hamiltonian. To gain further insight, we now focus on the vicinity of resonances, where we saw in Figure 11 (d)-Figure 11 (f) that the total detection probability is less than unity and has a broadening effect.
Near resonance: The eigenvalues satisfy ∣ Λ ± ∣ < 1 |\Lambda_{\pm}|<1 ∣ Λ ± ∣ < 1 in the complex plane and their values approach the unit circle, ∣ Λ ± ∣ → 1 − |\Lambda_{\pm}|\to1^{-} ∣ Λ ± ∣ → 1 − , when a dark state is formed in the system [55 ] . For simplicity, we now focus on S + S_+ S + , corresponding to the even-parity sector. For this case, resonance occurs when [53 , 55 ]
exp ( − i E ℓ τ ) = exp ( − i E ℓ ′ τ ) , (141) \exp(-iE_{\ell}\tau)=\exp(-iE_{\ell'}\tau), \tag*{(141)} exp ( − i E ℓ τ ) = exp ( − i E ℓ ′ τ ) , (141)
where the associated energy eigenstates, ∣ E ℓ ⟩ |E_{\ell}\rangle ∣ E ℓ ⟩ and ∣ E ℓ ′ ⟩ |E_{\ell'}\rangle ∣ E ℓ ′ ⟩ , both have even parity. The corresponding dark state is
∣ D ℓ ℓ ′ ( + ) ⟩ = a ℓ ′ ∣ E ℓ ⟩ − a ℓ ∣ E ℓ ′ ⟩ p ℓ + p ℓ ′ , (142) |D_{\ell\ell'}^{(+)}\rangle=\frac{a_{\ell'}|E_{\ell}\rangle-a_{\ell}|E_{\ell'}\rangle}{\sqrt{p_{\ell}+p_{\ell'}}}, \tag*{(142)} ∣ D ℓ ℓ ′ ( + ) ⟩ = p ℓ + p ℓ ′ a ℓ ′ ∣ E ℓ ⟩ − a ℓ ∣ E ℓ ′ ⟩ , (142)
with a ℓ = ⟨ d + ∣ E ℓ ⟩ a_{\ell}=\langle d_+|E_{\ell}\rangle a ℓ = ⟨ d + ∣ E ℓ ⟩ , p ℓ = ∣ ⟨ d + ∣ E ℓ ⟩ ∣ 2 p_{\ell}=|\langle d_+|E_{\ell}\rangle|^2 p ℓ = ∣ ⟨ d + ∣ E ℓ ⟩ ∣ 2 .
Close to the resonance condition (141) , we define
Δ ℓ ℓ ′ = Arg [ exp ( − i ( E ℓ − E ℓ ′ ) τ ) ] , (143) \Delta_{\ell\ell'}=\operatorname{Arg}\left[\exp\left(-i(E_{\ell}-E_{\ell'})\tau\right)\right], \tag*{(143)} Δ ℓ ℓ ′ = Arg [ exp ( − i ( E ℓ − E ℓ ′ ) τ ) ] , (143)
where recall Arg ( ⋅ ) \operatorname{Arg}(\cdot) Arg ( ⋅ ) stands for the argument of the complex number. At exact resonance, Δ ℓ ℓ ′ = 0 \Delta_{\ell\ell'}=0 Δ ℓ ℓ ′ = 0 , while close to resonance ∣ Δ ℓ ℓ ′ ∣ ≪ 1 |\Delta_{\ell\ell'}|\ll1 ∣ Δ ℓ ℓ ′ ∣ ≪ 1 .
Following earlier studies on single-target problems [73 , 74 ] , we use the following near-resonance approximations
∣ Λ + ( l e f t ) ⟩ ≈ ∣ Λ + ( r i g h t ) ⟩ ≈ ∣ D ℓ ℓ ′ ( + ) ⟩ , ∣ Λ + ∣ 2 ≈ 1 − p ℓ p ℓ ′ ( p ℓ + p ℓ ′ ) 3 Δ ℓ ℓ ′ 2 , (144) |\Lambda_+^{(\mathrm{left})}\rangle\approx|\Lambda_+^{(\mathrm{right})}\rangle\approx|D_{\ell\ell'}^{(+)}\rangle,\qquad|\Lambda_+|^2\approx1-\frac{p_\ell p_{\ell'}}{(p_\ell+p_{\ell'})^3}\Delta_{\ell\ell'}^2, \tag*{(144)} ∣ Λ + ( left ) ⟩ ≈ ∣ Λ + ( right ) ⟩ ≈ ∣ D ℓ ℓ ′ ( + ) ⟩ , ∣ Λ + ∣ 2 ≈ 1 − ( p ℓ + p ℓ ′ ) 3 p ℓ p ℓ ′ Δ ℓ ℓ ′ 2 , (144)
where ∣ D ℓ ℓ ′ ( + ) ⟩ |D_{\ell\ell'}^{(+)}\rangle ∣ D ℓ ℓ ′ ( + ) ⟩ and p ℓ p_\ell p ℓ are given in Eqs. (142) and (144), respectively. Plugging these approximate expressions in Eq. (140) for T max ≫ 1 T_{\max}\gg1 T m a x ≫ 1 , we have
∑ { Λ + } ∑ { Λ ~ + } ( Λ + ) T max ( Λ ~ + ) T max ⟨ ψ 0 ( + ) ∣ Λ ~ + ( l e f t ) ⟩ ⟨ Λ ~ + ( r i g h t ) ∣ Λ + ( r i g h t ) ⟩ ⟨ Λ + ( l e f t ) ∣ ψ 0 ( + ) ⟩ ≈ ∑ ℓ ∑ ℓ ′ < ℓ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( + ) ⟩ ∣ 2 exp [ − p ℓ p ℓ ′ ( p ℓ + p ℓ ′ ) 3 Δ ℓ ℓ ′ 2 T max ] I ℓ ℓ ′ ( + ) , (145) \begin{aligned}
\sum_{\{\Lambda_+\}}\sum_{\{\widetilde{\Lambda}_+\}}(\Lambda_+)^{T_{\max}}(\widetilde{\Lambda}_+)^{T_{\max}}\langle\psi_0^{(+)}|\widetilde{\Lambda}_+^{(\mathrm{left})}\rangle\langle\widetilde{\Lambda}_+^{(\mathrm{right})}|\Lambda_+^{(\mathrm{right})}\rangle\langle\Lambda_+^{(\mathrm{left})}|\psi_0^{(+)}\rangle\\
&\approx\sum_\ell\sum_{\ell'<\ell}|\langle\psi_0|D_{\ell\ell'}^{(+)}\rangle|^2\exp\left[-\frac{p_\ell p_{\ell'}}{(p_\ell+p_{\ell'})^3}\Delta_{\ell\ell'}^2T_{\max}\right]\mathcal{I}_{\ell\ell'}^{(+)}, \tag*{(145)}
\end{aligned} { Λ + } ∑ { Λ + } ∑ ( Λ + ) T m a x ( Λ + ) T m a x ⟨ ψ 0 ( + ) ∣ Λ + ( left ) ⟩ ⟨ Λ + ( right ) ∣ Λ + ( right ) ⟩ ⟨ Λ + ( left ) ∣ ψ 0 ( + ) ⟩ ≈ ℓ ∑ ℓ ′ < ℓ ∑ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( + ) ⟩ ∣ 2 exp [ − ( p ℓ + p ℓ ′ ) 3 p ℓ p ℓ ′ Δ ℓ ℓ ′ 2 T m a x ] I ℓ ℓ ′ ( + ) , (145)
In the second line, we have introduced the indicator function I ℓ ℓ ′ ( + ) \mathcal{I}_{\ell\ell'}^{(+)} I ℓ ℓ ′ ( + ) , which is equal to unity when both ∣ E ℓ ⟩ |E_\ell\rangle ∣ E ℓ ⟩ and ∣ E ℓ ′ ⟩ |E_{\ell'}\rangle ∣ E ℓ ′ ⟩ belong to the even-parity sector, and vanishes otherwise. A similar analysis on the odd-parity sector gives
∑ { Λ − } ∑ { Λ ~ − } ( Λ − ) T max ( Λ ~ − ) T max ⟨ ψ 0 ( − ) ∣ Λ ~ − ( l e f t ) ⟩ ⟨ Λ ~ − ( r i g h t ) ∣ Λ − ( r i g h t ) ⟩ ⟨ Λ − ( l e f t ) ∣ ψ 0 ( − ) ⟩ ≈ ∑ ℓ ∑ ℓ ′ < ℓ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( − ) ⟩ ∣ 2 exp [ − p ℓ p ℓ ′ ( p ℓ + p ℓ ′ ) 3 Δ ℓ ℓ ′ 2 T max ] I ℓ ℓ ′ ( − ) , (146) \begin{aligned}
\sum_{\{\Lambda_-\}}\sum_{\{\widetilde{\Lambda}_-\}}(\Lambda_-)^{T_{\max}}(\widetilde{\Lambda}_-)^{T_{\max}}\langle\psi_0^{(-)}|\widetilde{\Lambda}_-^{(\mathrm{left})}\rangle\langle\widetilde{\Lambda}_-^{(\mathrm{right})}|\Lambda_-^{(\mathrm{right})}\rangle\langle\Lambda_-^{(\mathrm{left})}|\psi_0^{(-)}\rangle\\
&\approx\sum_\ell\sum_{\ell'<\ell}|\langle\psi_0|D_{\ell\ell'}^{(-)}\rangle|^2\exp\left[-\frac{p_\ell p_{\ell'}}{(p_\ell+p_{\ell'})^3}\Delta_{\ell\ell'}^2T_{\max}\right]\mathcal{I}_{\ell\ell'}^{(-)}, \tag*{(146)}
\end{aligned} { Λ − } ∑ { Λ − } ∑ ( Λ − ) T m a x ( Λ − ) T m a x ⟨ ψ 0 ( − ) ∣ Λ − ( left ) ⟩ ⟨ Λ − ( right ) ∣ Λ − ( right ) ⟩ ⟨ Λ − ( left ) ∣ ψ 0 ( − ) ⟩ ≈ ℓ ∑ ℓ ′ < ℓ ∑ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( − ) ⟩ ∣ 2 exp [ − ( p ℓ + p ℓ ′ ) 3 p ℓ p ℓ ′ Δ ℓ ℓ ′ 2 T m a x ] I ℓ ℓ ′ ( − ) , (146)
where I ℓ ℓ ′ ( − ) \mathcal{I}_{\ell\ell'}^{(-)} I ℓ ℓ ′ ( − ) is the indicator function for the odd-parity sector and ∣ D ℓ ℓ ′ ( − ) ⟩ |D_{\ell\ell'}^{(-)}\rangle ∣ D ℓ ℓ ′ ( − ) ⟩ is the dark state corresponding to S ℓ ℓ ′ − \mathcal{S}_{\ell\ell'}^{-} S ℓ ℓ ′ −
∣ D ℓ ℓ ′ ( − ) ⟩ = a ℓ ′ ∣ E ℓ ⟩ − a ℓ ∣ E ℓ ′ ⟩ p ℓ + p ℓ ′ , |D_{\ell\ell'}^{(-)}\rangle=\frac{a_{\ell'}|E_\ell\rangle-a_\ell|E_{\ell'}\rangle}{\sqrt{p_\ell+p_{\ell'}}}, ∣ D ℓ ℓ ′ ( − ) ⟩ = p ℓ + p ℓ ′ a ℓ ′ ∣ E ℓ ⟩ − a ℓ ∣ E ℓ ′ ⟩ ,
where both ∣ E ℓ ⟩ |E_\ell\rangle ∣ E ℓ ⟩ and ∣ E ℓ ′ ⟩ |E_{\ell'}\rangle ∣ E ℓ ′ ⟩ have odd parity. {#eq:149}
Substituting Eqs. (145) -(149) in Eq. (140) , we have
P L ( ψ 0 ) + P R ( ψ 0 ) ≈ 1 − ∑ ℓ ∑ ℓ ′ < ℓ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( + ) ⟩ ∣ 2 exp [ − p ℓ p ℓ ′ ( p ℓ + p ℓ ′ ) 3 Δ ℓ ℓ ′ 2 T max ] I ℓ ℓ ′ ( + ) − ∑ ℓ ∑ ℓ ′ < ℓ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( − ) ⟩ ∣ 2 exp [ − p ℓ p ℓ ′ ( p ℓ + p ℓ ′ ) 3 Δ ℓ ℓ ′ 2 T max ] I ℓ ℓ ′ ( − ) . (147) \begin{aligned}
P_L(\psi_0)+P_R(\psi_0)&\approx1-\sum_\ell\sum_{\ell'<\ell}|\langle\psi_0|D_{\ell\ell'}^{(+)}\rangle|^2\exp\left[-\frac{p_\ell p_{\ell'}}{(p_\ell+p_{\ell'})^3}\Delta_{\ell\ell'}^2T_{\max}\right]\mathcal{I}_{\ell\ell'}^{(+)}\\
&\quad-\sum_\ell\sum_{\ell'<\ell}|\langle\psi_0|D_{\ell\ell'}^{(-)}\rangle|^2\exp\left[-\frac{p_\ell p_{\ell'}}{(p_\ell+p_{\ell'})^3}\Delta_{\ell\ell'}^2T_{\max}\right]\mathcal{I}_{\ell\ell'}^{(-)}. \tag*{(147)}
\end{aligned} P L ( ψ 0 ) + P R ( ψ 0 ) ≈ 1 − ℓ ∑ ℓ ′ < ℓ ∑ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( + ) ⟩ ∣ 2 exp [ − ( p ℓ + p ℓ ′ ) 3 p ℓ p ℓ ′ Δ ℓ ℓ ′ 2 T m a x ] I ℓ ℓ ′ ( + ) − ℓ ∑ ℓ ′ < ℓ ∑ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ( − ) ⟩ ∣ 2 exp [ − ( p ℓ + p ℓ ′ ) 3 p ℓ p ℓ ′ Δ ℓ ℓ ′ 2 T m a x ] I ℓ ℓ ′ ( − ) . (147)
We emphasize that this expression is an asymptotic result, derived for large T max T_{\max} T m a x and in the vicinity of the finite-τ \tau τ resonances that satisfy Eq. (141) .
Nearest-neighbor hopping Hamiltonian: For this case, we use Eq. (57) in Eq. (144) to get a ℓ = 2 / N + 1 sin k ℓ a_\ell=2/\sqrt{N+1}\sin k_\ell a ℓ = 2/ N + 1 sin k ℓ and p ℓ = 4 / ( N + 1 ) sin 2 k ℓ p_\ell=4/(N+1)\sin^2 k_\ell p ℓ = 4/ ( N + 1 ) sin 2 k ℓ where k ℓ = π ℓ / ( N + 1 ) k_\ell=\pi\ell/(N+1) k ℓ = π ℓ / ( N + 1 ) . Moreover, we saw before in Eq. (55) that the energy eigenstate ∣ E ℓ ⟩ |E_\ell\rangle ∣ E ℓ ⟩ has even parity for odd ℓ \ell ℓ and odd parity for even ℓ \ell ℓ . Using these results in Eq. (147) gives
P L ( ψ 0 ) + P R ( ψ 0 ) ≈ 1 − ∑ ℓ = 1 N ∑ ℓ ′ < ℓ 1 + ( − 1 ) ℓ − ℓ ′ 2 ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ⟩ ∣ 2 exp [ − N + 1 4 sin 2 k ℓ sin 2 k ℓ ′ ( sin 2 k ℓ + sin 2 k ℓ ′ ) 3 Δ ℓ ℓ ′ 2 T max ] . (148) P_L(\psi_0)+P_R(\psi_0)\approx1-\sum_{\ell=1}^{N}\sum_{\ell'<\ell}\frac{1+(-1)^{\ell-\ell'}}{2}|\langle\psi_0|D_{\ell\ell'}\rangle|^2\exp\left[-\frac{N+1}{4}\frac{\sin^2 k_\ell\sin^2 k_{\ell'}}{(\sin^2 k_\ell+\sin^2 k_{\ell'})^3}\Delta_{\ell\ell'}^2T_{\max}\right]. \tag*{(148)} P L ( ψ 0 ) + P R ( ψ 0 ) ≈ 1 − ℓ = 1 ∑ N ℓ ′ < ℓ ∑ 2 1 + ( − 1 ) ℓ − ℓ ′ ∣ ⟨ ψ 0 ∣ D ℓ ℓ ′ ⟩ ∣ 2 exp [ − 4 N + 1 ( sin 2 k ℓ + sin 2 k ℓ ′ ) 3 sin 2 k ℓ sin 2 k ℓ ′ Δ ℓ ℓ ′ 2 T m a x ] . (148)
where the factor ( 1 + ( − 1 ) ℓ − ℓ ′ ) / 2 (1+(-1)^{\ell-\ell'})/2 ( 1 + ( − 1 ) ℓ − ℓ ′ ) /2 ensures that only energy eigenstates with the same parity, i.e., with ℓ \ell ℓ and ℓ ′ \ell' ℓ ′ both odd or both even, contribute to the sum. We also define ∣ D ℓ ℓ ′ ⟩ |D_{\ell\ell'}\rangle ∣ D ℓ ℓ ′ ⟩ as
∣ D ℓ ℓ ′ ⟩ = sin k ℓ ′ ∣ E ℓ ⟩ − sin k ℓ ∣ E ℓ ′ ⟩ sin 2 k ℓ + sin 2 k ℓ ′ . (149) |D_{\ell\ell'}\rangle=\frac{\sin k_{\ell'}|E_\ell\rangle-\sin k_\ell|E_{\ell'}\rangle}{\sqrt{\sin^2 k_\ell+\sin^2 k_{\ell'}}}. \tag*{(149)} ∣ D ℓ ℓ ′ ⟩ = sin 2 k ℓ + sin 2 k ℓ ′ sin k ℓ ′ ∣ E ℓ ⟩ − sin k ℓ ∣ E ℓ ′ ⟩ . (149)
The theoretical prediction in Eq. (148) is shown in Figure 11 (d)-Figure 11 (f), where we compare it with the numerical simulations. Away from the Zeno limit, we find excellent agreement between the analytical expression and the simulation results. Close to the Zeno limit, however, noticeable deviations appear. This is expected, since the derivation of Eq. (148) relies on the resonance condition in Eq. (141) , which is not applicable in the Zeno limit. In this limit, all quasienergies become simultaneously degenerate.
However, for the other resonances, Eq. (141) is satisfied, and consequently there is excellent agreement between Eq. (148) and the simulations. Close to such a resonance, ∣ Δ ℓ ℓ ′ ∣ ≪ 1 \lvert\Delta_{\ell\ell'}\rvert\ll1 ∣ Δ ℓ ℓ ′ ∣ ≪ 1 for the corresponding pair of energy levels, so that Δ ℓ ℓ ′ 2 T max \Delta_{\ell\ell'}^{2}T_{\max} Δ ℓ ℓ ′ 2 T m a x is finite inside the exponential factor in Eq. (148) . Therefore, the terms in the summation in Eq. (148) are significant. As a result, P L ( ψ 0 ) + P R ( ψ 0 ) < 1 P_L(\psi_0) + P_R(\psi_0) < 1 P L ( ψ 0 ) + P R ( ψ 0 ) < 1 over a finite range of sampling times close to the resonance, giving rise to the broadening effect, as seen in Figure 11 (d)-Figure 11 (f). As one moves away from resonance, Δ ℓ ℓ ′ \Delta_{\ell\ell'} Δ ℓ ℓ ′ becomes finite, and one gets Δ ℓ ℓ ′ 2 T max ≫ 1 \Delta_{\ell\ell'}^{2}T_{\max} \gg1 Δ ℓ ℓ ′ 2 T m a x ≫ 1 for sufficiently large T max T_{\max} T m a x . The corresponding terms in the summation are exponentially suppressed. Consequently, the normalization approaches P L ( ψ 0 ) + P R ( ψ 0 ) = 1 P_L(\psi_0) + P_R(\psi_0) = 1 P L ( ψ 0 ) + P R ( ψ 0 ) = 1 , as observed in Figure 11 (d)-Figure 11 (f).
Conclusion and Future Directions
In conclusion, we have constructed a general theoretical framework for the splitting problem of a discrete-state, continuous-time quantum system, and subjected to projective measurements at two distinct, orthogonal target states at regular time intervals τ \tau τ . We provide explicit tools for analyzing how unitary evolution and repeated measurements combine to shape the competition between the two possible outcomes. Our renewal framework, as exemplified in Eq. (6) , expresses the associated first detection amplitudes in terms of the measurement-free unitary evolution. The splitting probability is then given by Eq. (4) or equivalently by Eqs. (30) and (35) . These results are valid for arbitrary time-independent, Hermitian Hamiltonian H H H , target states ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ , ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , the initial state ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ and the sampling time τ \tau τ .
For Hamiltonians possessing reflection symmetry and localized measurements performed at the left and right endpoints of a one-dimensional chain, ∣ D 1 ⟩ = ∣ 1 ⟩ \lvert D_1\rangle= \lvert1\rangle ∣ D 1 ⟩ = ∣ 1 ⟩ and ∣ D 2 ⟩ = ∣ N ⟩ \lvert D_2\rangle= \lvert N\rangle ∣ D 2 ⟩ = ∣ N ⟩ , the renewal framework reduces to a mapping that relates the splitting probabilities to a pair of single-target first-detection problems. For this class of Hamiltonians, we obtained exact expressions of the splitting probabilities, P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) and P R ( ψ 0 ) P_R(\psi_0) P R ( ψ 0 ) , in Eqs. (52) . To get some insights into these expressions, we examined a tight-binding chain with nearest-neighbor hopping Hamiltonian in Eq. (2) . An interesting feature emerges in the Zeno limit, where both P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) and P R ( ψ 0 ) P_R(\psi_0) P R ( ψ 0 ) , for any system size N N N , become independent of the detailed structure of the initial state ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ in the interior of the chain and depend only on the projection of ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ on the target boundary sites, see Eq. (11) .
For localized initial condition, ∣ ψ 0 ⟩ = ∣ x 0 ⟩ \lvert\psi_0\rangle= \lvert x_0\rangle ∣ ψ 0 ⟩ = ∣ x 0 ⟩ , we found a surprising phase-transition-like behavior in the splitting dynamics at a critical sampling time τ c = π / ( 2 γ ) \tau_c = \pi/(2\gamma) τ c = π / ( 2 γ ) in the limit of large system size for the nearest-neighbor hopping Hamiltonian. This flat-to-fluctuating transition separates two qualitatively distinct regimes: For 0 ≤ τ < τ c 0 \leq\tau< \tau_c 0 ≤ τ < τ c , the splitting probabilities have a flat value 1 / 2 1/2 1/2 for all x 0 x_0 x 0 and τ \tau τ , while for τ > τ c \tau> \tau_c τ > τ c , both P L ( x 0 ) P_L(x_0) P L ( x 0 ) and P R ( x 0 ) P_R(x_0) P R ( x 0 ) differ from 1 / 2 1/2 1/2 and have a fluctuating dependence on x 0 x_0 x 0 and τ \tau τ . This is illustrated in Figure 6 . Beyond localized initial conditions, we showed that this transition can also persist for initial states involving coherent superpositions over multiple sites. Interestingly, its detailed form can be modified by the spatial separation and relative phase between the components of the initial state. In particular, the splitting probability exhibits an oscillatory dependence on the initial phase, with a parity-dependent phase shift, as shown in Eq. (128) and Figure 10 .
The fluctuating regime at τ > τ c \tau> \tau_c τ > τ c for localized case also exhibits a characteristic spatial structure. The splitting probabilities oscillate with the initial position x 0 x_0 x 0 , with dominant wave numbers Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) given in Eq. (13) . These dominant modes form the spectral branches seen in the power-spectrum heat map in Figure 4 , with the first branch emerging at τ c \tau_c τ c and additional branches appearing at larger sampling times. The onset of these oscillations can be understood from the inter-parity phase-matching condition in Eq. (82) , or equivalently Eq. (87) , between momentum modes. While for τ > τ c \tau> \tau_c τ > τ c there exist distinct momentum modes satisfying this phase-matching condition, for τ < τ c \tau< \tau_c τ < τ c no such modes exist. This leads to the spectral interpretation of the oscillations in the splitting probability and the associated flat-to-fluctuating transition. Finally, we derived the resonance conditions in Eq. (10) that determine the pointwise discontinuous jumps in the splitting probability at special sampling times. At these sampling times, P L ( x 0 ) + P R ( x 0 ) < 1 P_L(x_0) + P_R(x_0) < 1 P L ( x 0 ) + P R ( x 0 ) < 1 . For a finite number of measurements, these pointwise discontinuous jumps are broadened into dips of finite width, as described by Eq. (148) and illustrated in Figure 11 .
For the classical case, these features do not arise, even when the classical random walker is monitored stroboscopically with time period τ \tau τ [43 ] . As studied in detail in the Supplementary Material of [43 ] , a continuous-time classical random walk hops on a one-dimensional lattice chain, with its boundaries monitored only at the times τ , 2 τ , 3 τ , … \tau, 2\tau, 3\tau, \ldots τ , 2 τ , 3 τ , … . If the walker is located at any of the boundaries at the measurement times, it is absorbed and the process is stopped. Otherwise, it continues to evolve until a successful detection is made. Here, τ → 0 \tau\to0 τ → 0 corresponds to the continuous monitoring and leads to the gambler’s ruin formula for the splitting probabilities in Eq. (1) . By contrast, the τ = 0 \tau= 0 τ = 0 for quantum case leads to the freezing of dynamics with no detection, unless the starting position coincides with the detectors; see Eq. (70) . For arbitrary τ \tau τ , the classical splitting probabilities vary smoothly with the sampling time, without exhibiting the phase-transition-like behavior found in the quantum case. Moreover, the classical proximity effect is always preserved, with the classical walker having a higher probability of detection at the boundary closer to its initial position, whereas in the quantum case, detection at the more distant boundary can also be favored. Overall, our work establishes a systematic treatment of splitting probabilities in quantum monitored dynamics with two targets, clarifying the role of quantum measurements and interference effects. Looking ahead, our work points to several future directions. While we concentrated on splitting probabilities between two different targets, in many cases it is useful to characterize the hitting times at various targets in the system and their probabilities [2]. Our renewal framework, in principle, can be used to compute these quantities, and are left for the future. Extending our framework for other measurement protocols such as weak measurements [83 ] or random sampling time protocols [59 , 65 ] , and comparing them with our results is a promising future direction. Furthermore, our analysis has focused on local monitoring at fixed target states. It would be interesting to investigate the splitting problem under a tracking protocol, in which the system is monitored over the full set of states rather than only at the designated targets [63 , 73 ] . An interesting question is whether spatial oscillations also emerge only beyond a critical sampling time under such a protocol. Another direction is to determine the τ \tau τ -dependence of the amplitudes and phases of the spatial oscillations, beyond the characteristic wavenumbers obtained here for large system sizes. The scaling of the special resonant sampling times with system size is another aspect worth exploring. It would also be interesting to explore the splitting problems for other graphs and many-body systems [66 , 84 ] . In particular, the transition point τ c \tau_c τ c is dependent only on the width of the energy band via Eq. (12) , since the discrete energy levels merge into a continuum in the thermodynamic limit. Therefore, the transition between flat to fluctuating regimes is likely to be go beyond the current model.
Finally, the splitting protocol considered in our paper can be implemented in experimental platforms like quantum computer [72 , 73 , 74 ] or optical waveguides [69 , 70 , 71 ] . In quantum-computing implementations, the quantum-walk dynamics can be mapped to a spin model through the Jordan-Wigner transformation and periodic detection protocol can be implemented by mid-circuit measurements. Similarly, in optical waveguides, quantum walks can be effectively implemented by using electric fields across an array of coupled waveguides, and periodic measurements can be realized using a segmented waveguides. We anticipate that incorporating noise, as encountered in realistic systems, may reveal further insights [74 , 85 ] .
Acknowledgment
This research was supported by the Israel Science Foundation (grant No. 2311/25).
Derivation of the quantum renewal equation
We first recast the first-detection amplitude ϕ n ( D α ) ( ψ 0 ) \phi_{n}^{(D_{\alpha})}(\psi_0) ϕ n ( D α ) ( ψ 0 ) in Eq. (5) as
ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ θ n ⟩ , (150) \phi_{n}^{(D_{\alpha})}(\psi_0)=\langle D_{\alpha}|\theta_n\rangle, \tag*{(150)} ϕ n ( D α ) ( ψ 0 ) = ⟨ D α ∣ θ n ⟩ , (150)
where we define
∣ θ n ⟩ = U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ . (151) |\theta_n\rangle=U(\tau)\left[(1-\hat{D})U(\tau)\right]^{n-1}|\psi_0\rangle. \tag*{(151)} ∣ θ n ⟩ = U ( τ ) [ ( 1 − D ^ ) U ( τ ) ] n − 1 ∣ ψ 0 ⟩ . (151)
Here U ( τ ) = e − i H ^ τ U(\tau)=e^{-i\hat{H}\tau} U ( τ ) = e − i H ^ τ is the measurement-free unitary operator, and D ^ \hat{D} D ^ is the projection operator onto the detection subspace, as defined in Eq. (21) .
Proving the renewal equation (28) is equivalent to showing
∣ θ n ⟩ = U ( n τ ) ∣ ψ 0 ⟩ − ∑ l = 1 n − 1 U ( ( n − l ) τ ) D ^ ∣ θ l ⟩ . (152) |\theta_n\rangle=U(n\tau)|\psi_0\rangle-\sum_{l=1}^{n-1}U((n-l)\tau)\hat{D}|\theta_l\rangle. \tag*{(152)} ∣ θ n ⟩ = U ( n τ ) ∣ ψ 0 ⟩ − l = 1 ∑ n − 1 U (( n − l ) τ ) D ^ ∣ θ l ⟩ . (152)
Using the expression of ∣ θ n ⟩ |\theta_n\rangle ∣ θ n ⟩ in Eq. (151) , we get for n = 1 n=1 n = 1 and n = 2 n=2 n = 2
∣ θ 1 ⟩ = U ( τ ) ∣ ψ 0 ⟩ , (153) |\theta_1\rangle=U(\tau)|\psi_0\rangle, \tag*{(153)} ∣ θ 1 ⟩ = U ( τ ) ∣ ψ 0 ⟩ , (153)
∣ θ 2 ⟩ = U ( 2 τ ) ∣ ψ 0 ⟩ − U ( τ ) D ^ ∣ θ 1 ⟩ . (154) |\theta_2\rangle=U(2\tau)|\psi_0\rangle-U(\tau)\hat{D}|\theta_1\rangle. \tag*{(154)} ∣ θ 2 ⟩ = U ( 2 τ ) ∣ ψ 0 ⟩ − U ( τ ) D ^ ∣ θ 1 ⟩ . (154)
Both of these results are consistent with the recursive relation in Eq. (152) . For general n n n , we will establish this relation by induction.
Assume that Eq. (152) holds for some arbitrary n n n . Then, using Eq. (151) , we obtain
∣ θ n + 1 ⟩ = U ( τ ) ∣ θ n ⟩ − U ( τ ) D ^ ∣ θ n ⟩ , (155) \lvert\theta_{n+1}\rangle= U(\tau)\lvert\theta_n\rangle- U(\tau)\hat{D}\lvert\theta_n\rangle, \tag*{(155)} ∣ θ n + 1 ⟩ = U ( τ ) ∣ θ n ⟩ − U ( τ ) D ^ ∣ θ n ⟩ , (155)
= U ( ( n + 1 ) τ ) ∣ ψ 0 ⟩ − ∑ l = 1 n − 1 U ( ( n − l + 1 ) τ ) D ^ ∣ θ l ⟩ − U ( τ ) D ^ ∣ θ n ⟩ , (156) = U((n+1)\tau)\lvert\psi_0\rangle- \sum_{l=1}^{n-1} U((n-l+1)\tau)\hat{D}\lvert\theta_l\rangle- U(\tau)\hat{D}\lvert\theta_n\rangle, \tag*{(156)} = U (( n + 1 ) τ ) ∣ ψ 0 ⟩ − l = 1 ∑ n − 1 U (( n − l + 1 ) τ ) D ^ ∣ θ l ⟩ − U ( τ ) D ^ ∣ θ n ⟩ , (156)
= U ( ( n + 1 ) τ ) ∣ ψ 0 ⟩ − ∑ l = 1 n U ( ( n + 1 − l ) τ ) D ^ ∣ θ l ⟩ . (157) = U((n+1)\tau)\lvert\psi_0\rangle- \sum_{l=1}^{n} U((n+1-l)\tau)\hat{D}\lvert\theta_l\rangle. \tag*{(157)} = U (( n + 1 ) τ ) ∣ ψ 0 ⟩ − l = 1 ∑ n U (( n + 1 − l ) τ ) D ^ ∣ θ l ⟩ . (157)
This is essentially Eq. (152) evaluated at ( n + 1 ) (n+1) ( n + 1 ) . Hence, by induction, this relation holds for arbitrary n n n .
Details of the numerical simulations
To compute the splitting probabilities numerically, we employ a Monte Carlo simulation of the stroboscopic detection protocol for a given Hamiltonian H H H , and the measurement interval τ \tau τ . The procedure can be summarized as follows:
Starting from the initial state ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ , the system evolves under the unitary operator U ( τ ) U(\tau) U ( τ ) for a duration τ \tau τ . The probability of detecting the system in the first target state ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ is given by ω 1 ( τ ) = ∣ ⟨ D 1 ∣ U ( τ ) ∣ ψ 0 ⟩ ∣ 2 \omega_1(\tau)=\lvert\langle D_1\lvert U(\tau)\lvert\psi_0\rangle\rvert^2 ω 1 ( τ ) = ∣⟨ D 1 ∣ U ( τ ) ∣ ψ 0 ⟩ ∣ 2 . To implement the measurement, we draw a random number, r 1 r_1 r 1 , uniformly from the range [ 0 , 1 ] [0,1] [ 0 , 1 ] . If r 1 ≤ ω 1 ( τ ) r_1\leq\omega_1(\tau) r 1 ≤ ω 1 ( τ ) , the system is detected at ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ , and the process terminates.
If this detection attempt is unsuccessful, i.e., r 1 > ω 1 r_1>\omega_1 r 1 > ω 1 , the state is immediately collapsed to
∣ ψ ( τ + ) ⟩ = ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣ ) U ( τ ) ∣ ψ 0 ⟩ 1 − ω 1 ( τ ) . \lvert\psi(\tau^+)\rangle= \frac{(1-\lvert D_1\rangle\langle D_1\rvert)U(\tau)\lvert\psi_0\rangle}{\sqrt{1-\omega_1(\tau)}}. ∣ ψ ( τ + )⟩ = 1 − ω 1 ( τ ) ( 1 − ∣ D 1 ⟩ ⟨ D 1 ∣) U ( τ ) ∣ ψ 0 ⟩ .
where τ + \tau^+ τ + recall is the time instant immediately after a failed measurement of ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ . An immediate measurement is then performed at the second target state ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ . The probability of detection is ω 2 ( τ ) = ∣ ⟨ D 2 ∣ ψ ( τ + ) ⟩ ∣ 2 \omega_2(\tau)=\lvert\langle D_2\lvert\psi(\tau^+)\rangle\rvert^2 ω 2 ( τ ) = ∣⟨ D 2 ∣ ψ ( τ + )⟩ ∣ 2 . To employ the measurement, we again generate another random number r 2 r_2 r 2 uniformly in the range [ 0 , 1 ] [0,1] [ 0 , 1 ] . If r 2 ≤ ω 2 ( τ ) r_2\leq\omega_2(\tau) r 2 ≤ ω 2 ( τ ) , the system is detected at ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , and the process terminates.
On the other hand, if r 2 > ω 2 ( τ ) r_2>\omega_2(\tau) r 2 > ω 2 ( τ ) , this detection attempt also fails. The state is then collapsed with respect to ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , yielding
∣ ψ ( τ + + ) ⟩ = ( 1 − ∣ D 2 ⟩ ⟨ D 2 ∣ ) 1 − ω 2 ( τ ) ∣ ψ ( τ + ) ⟩ . \lvert\psi(\tau^{++})\rangle= \frac{(1-\lvert D_2\rangle\langle D_2\rvert)}{\sqrt{1-\omega_2(\tau)}}\lvert\psi(\tau^+)\rangle. ∣ ψ ( τ ++ )⟩ = 1 − ω 2 ( τ ) ( 1 − ∣ D 2 ⟩ ⟨ D 2 ∣) ∣ ψ ( τ + )⟩ .
Taking this normalized state in place of the original ∣ ψ 0 ⟩ \lvert\psi_0\rangle ∣ ψ 0 ⟩ , the entire procedure is then repeated at the next stroboscopic time step.
This recursive process continues until the system is successfully detected at either ∣ D 1 ⟩ \lvert D_1\rangle ∣ D 1 ⟩ or ∣ D 2 ⟩ \lvert D_2\rangle ∣ D 2 ⟩ , or the maximum T max T_{\max} T m a x number of measurement attempts is reached. Subsequently, the simulation is carried out for R R R independent realizations, from which the splitting probabilities, P D 1 ( ψ 0 ) P_{D_1}(\psi_0) P D 1 ( ψ 0 ) and P D 2 ( ψ 0 ) P_{D_2}(\psi_0) P D 2 ( ψ 0 ) , are estimated as ensemble averages. If R α R_\alpha R α denotes the total number of realizations in which the system is detected at the ∣ D α ⟩ \lvert D_\alpha\rangle ∣ D α ⟩ state (with α ∈ { 1 , 2 } \alpha\in\{1,2\} α ∈ { 1 , 2 } ), then the splitting probabilities are
P D 1 ( ψ 0 ) = R 1 R , P D 2 ( ψ 0 ) = R 2 R . (158) P_{D_1}(\psi_0)=\frac{R_1}{R},\qquad P_{D_2}(\psi_0)=\frac{R_2}{R}. \tag*{(158)} P D 1 ( ψ 0 ) = R R 1 , P D 2 ( ψ 0 ) = R R 2 . (158)
Note that R 1 + R 2 ≤ R R_1+R_2\leq R R 1 + R 2 ≤ R , with strict inequality in the case of special measurement times. In our simulation, we fix T max = 500 , 1000 , 1500 , 2000 , 10 4 T_{\max}=500,1000,1500,2000,10^4 T m a x = 500 , 1000 , 1500 , 2000 , 1 0 4 and R = 10 4 R=10^4 R = 1 0 4 .
τ = 0 \tau=0 τ = 0 case
In this appendix, we will derive the splitting detection probabilities P L ( ψ 0 ) P_L(\psi_0) P L ( ψ 0 ) and P R ( ψ 0 ) P_R(\psi_0) P R ( ψ 0 ) for τ = 0 \tau=0 τ = 0 , as was quoted in Eq. (69) in the main text. Our starting point are the auxiliary probabilities ρ ± ( ψ 0 ) \rho_{\pm}(\psi_0) ρ ± ( ψ 0 ) in Eqs. (93) and (94) . Since all quasienergies are identical, we have
ρ + ( ψ 0 ) = ∣ ∑ m = 1 m ∈ o d d N ⟨ d + ∣ E m ⟩ ⟨ E m ∣ ψ 0 ⟩ ∣ 2 ∑ m = 1 m ∈ o d d N ∣ ⟨ d + ∣ E m ⟩ ∣ 2 . (159) \rho_{+}(\psi_{0}) = \frac{\left|\sum_{\substack{m=1\\m\in\mathrm{odd}}}^{N}\langle d_{+}|E_{m}\rangle\langle E_{m}|\psi_{0}\rangle\right|^{2}}{\sum_{\substack{m=1\\m\in\mathrm{odd}}}^{N}|\langle d_{+}|E_{m}\rangle|^{2}}. \tag*{(159)} ρ + ( ψ 0 ) = ∑ m = 1 m ∈ odd N ∣ ⟨ d + ∣ E m ⟩ ∣ 2 ∑ m = 1 m ∈ odd N ⟨ d + ∣ E m ⟩ ⟨ E m ∣ ψ 0 ⟩ 2 . (159)
Using ∣ d + ⟩ |d_{+}\rangle ∣ d + ⟩ from Eq. (57) , the denominator in Eq. (159) simplifies to
∑ m = 1 m ∈ o d d N ∣ ⟨ d + ∣ E m ⟩ ∣ 2 = 4 N + 1 ∑ m = 1 m ∈ o d d N sin ( π m N + 1 ) 2 = 1. (160) \sum_{\substack{m=1\\m\in\mathrm{odd}}}^{N}|\langle d_{+}|E_{m}\rangle|^{2}=\frac{4}{N+1}\sum_{\substack{m=1\\m\in\mathrm{odd}}}^{N}\sin\left(\frac{\pi m}{N+1}\right)^{2}=1. \tag*{(160)} m = 1 m ∈ odd ∑ N ∣ ⟨ d + ∣ E m ⟩ ∣ 2 = N + 1 4 m = 1 m ∈ odd ∑ N sin ( N + 1 π m ) 2 = 1. (160)
On the other hand, the numerator depends on the initial state ∣ ψ 0 ⟩ |\psi_{0}\rangle ∣ ψ 0 ⟩
∑ m = 1 m ∈ o d d N ⟨ d + ∣ E m ⟩ ⟨ E m ∣ ψ 0 ⟩ = ⟨ d + ∣ ψ 0 ⟩ = 1 2 ( ⟨ 1 ∣ ψ 0 ⟩ + ⟨ N ∣ ψ 0 ⟩ ) . (161) \sum_{\substack{m=1\\m\in\mathrm{odd}}}^{N}\langle d_{+}|E_{m}\rangle\langle E_{m}|\psi_{0}\rangle=\langle d_{+}|\psi_{0}\rangle=\frac{1}{\sqrt{2}}\left(\langle1|\psi_{0}\rangle+\langle N|\psi_{0}\rangle\right). \tag*{(161)} m = 1 m ∈ odd ∑ N ⟨ d + ∣ E m ⟩ ⟨ E m ∣ ψ 0 ⟩ = ⟨ d + ∣ ψ 0 ⟩ = 2 1 ( ⟨ 1∣ ψ 0 ⟩ + ⟨ N ∣ ψ 0 ⟩ ) . (161)
The simplified forms of the numerator and denominator now yield
ρ + ( ψ 0 ) = 1 2 ∣ ⟨ 1 ∣ ψ 0 ⟩ + ⟨ N ∣ ψ 0 ⟩ ∣ 2 . (162) \rho_{+}(\psi_{0})=\frac{1}{2}\left|\langle1|\psi_{0}\rangle+\langle N|\psi_{0}\rangle\right|^{2}. \tag*{(162)} ρ + ( ψ 0 ) = 2 1 ∣ ⟨ 1∣ ψ 0 ⟩ + ⟨ N ∣ ψ 0 ⟩ ∣ 2 . (162)
The same procedure also yields the other auxiliary probability
ρ − ( ψ 0 ) = 1 2 ∣ ⟨ 1 ∣ ψ 0 ⟩ − ⟨ N ∣ ψ 0 ⟩ ∣ 2 . (163) \rho_{-}(\psi_{0})=\frac{1}{2}\left|\langle1|\psi_{0}\rangle-\langle N|\psi_{0}\rangle\right|^{2}. \tag*{(163)} ρ − ( ψ 0 ) = 2 1 ∣ ⟨ 1∣ ψ 0 ⟩ − ⟨ N ∣ ψ 0 ⟩ ∣ 2 . (163)
Turning to Eq. (52) , we also need the interference term ζ + − ( ψ 0 ) \zeta_{+-}(\psi_{0}) ζ +− ( ψ 0 ) to fully specify the splitting probabilities. This, in turn, requires the auxiliary generating functions χ ~ ± ( z ) \widetilde{\chi}_{\pm}(z) χ ± ( z ) via Eq. (53) . For τ = 0 \tau=0 τ = 0 , Eq. (46) simplifies to
χ ~ ± ( z ) ≃ z ⟨ d ± ∣ ψ 0 ⟩ ≃ z 2 ( ⟨ 1 ∣ ψ 0 ⟩ ± ⟨ N ∣ ψ 0 ⟩ ) . (164) \widetilde{\chi}_{\pm}(z)\simeq z\langle d_{\pm}|\psi_{0}\rangle\simeq\frac{z}{\sqrt{2}}\left(\langle1|\psi_{0}\rangle\pm\langle N|\psi_{0}\rangle\right). \tag*{(164)} χ ± ( z ) ≃ z ⟨ d ± ∣ ψ 0 ⟩ ≃ 2 z ( ⟨ 1∣ ψ 0 ⟩ ± ⟨ N ∣ ψ 0 ⟩ ) . (164)
The interference term in Eq. (53) then becomes
ζ + − ( ψ 0 ) ≃ 1 2 ( ∣ ⟨ 1 ∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 ) . (165) \zeta_{+-}(\psi_{0})\simeq\frac{1}{2}\left(\left|\langle1|\psi_{0}\rangle\right|^{2}-\left|\langle N|\psi_{0}\rangle\right|^{2}\right). \tag*{(165)} ζ +− ( ψ 0 ) ≃ 2 1 ( ∣ ⟨ 1∣ ψ 0 ⟩ ∣ 2 − ∣ ⟨ N ∣ ψ 0 ⟩ ∣ 2 ) . (165)
Substituting this expression into Eq. (52) , together with ρ ± ( ψ 0 ) \rho_{\pm}(\psi_{0}) ρ ± ( ψ 0 ) from Eqs. (162) and (163) , we obtain the expressions for the splitting probabilities given in Eq. (69) .
Dominant wavenumber in the power spectrum of ζ + − ( x 0 ) \zeta_{+-}(x_{0}) ζ +− ( x 0 )
In this Appendix, we derive the large-N N N expression of the dominant wavenumber Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) in Eq. (86) , at which the power spectrum ∣ ζ ~ ( Q , N ) ∣ 2 |\widetilde{\zeta}(Q,N)|^{2} ∣ ζ ( Q , N ) ∣ 2 in Eq. (71) for the nearest-neighbor Hamiltonian is maximum. Our starting point is Eq. (76) , which we rewrite here for simplicity
ζ ~ ( Q , N ) = ∑ s = 1 N ∑ s ′ = 1 s ′ o d d s ′ e v e n N C s , s ′ ( τ , N ) 4 [ e − i ( Q − k s + k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q − k s + k s ′ ) ] sin [ 1 2 ( Q − k s + k s ′ ) ] + e − i ( Q + k s − k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q + k s − k s ′ ) ] sin [ 1 2 ( Q + k s − k s ′ ) ] − e − i ( Q − k s − k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q − k s − k s ′ ) ] sin [ 1 2 ( Q − k s − k s ′ ) ] − e − i ( Q + k s + k s ′ ) ( N + 1 ) 2 sin [ N 2 ( Q + k s + k s ′ ) ] sin [ 1 2 ( Q + k s + k s ′ ) ] ] , (166) \begin{aligned}
\widetilde{\zeta}(Q,N)={}&\sum_{s=1}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{odd}\ s'\ \mathrm{even}}}^{N}\frac{C_{s,s'}(\tau,N)}{4}\Biggl[
e^{-i\frac{(Q-k_s+k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q-k_s+k_{s'})\right]}{\sin\left[\frac{1}{2}(Q-k_s+k_{s'})\right]}\\
&+e^{-i\frac{(Q+k_s-k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q+k_s-k_{s'})\right]}{\sin\left[\frac{1}{2}(Q+k_s-k_{s'})\right]}\\
&-e^{-i\frac{(Q-k_s-k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q-k_s-k_{s'})\right]}{\sin\left[\frac{1}{2}(Q-k_s-k_{s'})\right]}\\
&-e^{-i\frac{(Q+k_s+k_{s'})(N+1)}{2}}\frac{\sin\left[\frac{N}{2}(Q+k_s+k_{s'})\right]}{\sin\left[\frac{1}{2}(Q+k_s+k_{s'})\right]}
\Biggr], \tag*{(166)}
\end{aligned} ζ ( Q , N ) = s = 1 ∑ N s ′ = 1 s ′ odd s ′ even ∑ N 4 C s , s ′ ( τ , N ) [ e − i 2 ( Q − k s + k s ′ ) ( N + 1 ) sin [ 2 1 ( Q − k s + k s ′ ) ] sin [ 2 N ( Q − k s + k s ′ ) ] + e − i 2 ( Q + k s − k s ′ ) ( N + 1 ) sin [ 2 1 ( Q + k s − k s ′ ) ] sin [ 2 N ( Q + k s − k s ′ ) ] − e − i 2 ( Q − k s − k s ′ ) ( N + 1 ) sin [ 2 1 ( Q − k s − k s ′ ) ] sin [ 2 N ( Q − k s − k s ′ ) ] − e − i 2 ( Q + k s + k s ′ ) ( N + 1 ) sin [ 2 1 ( Q + k s + k s ′ ) ] sin [ 2 N ( Q + k s + k s ′ ) ] ] , (166)
where recall k s = π s / ( N + 1 ) k_s=\pi s/(N+1) k s = π s / ( N + 1 ) and C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is given in Eq. (73) . Notice that Eq. (166) contains terms of the form
e − i y ( N + 1 ) 2 sin ( N y 2 ) sin ( y 2 ) = cos [ y ( N + 1 ) 2 ] sin ( N y 2 ) sin ( y 2 ) − i sin [ y ( N + 1 ) 2 ] sin ( N y 2 ) sin ( y 2 ) . (167) e^{-i\frac{y(N+1)}{2}}\frac{\sin\left(\frac{Ny}{2}\right)}{\sin\left(\frac{y}{2}\right)}
=\cos\left[\frac{y(N+1)}{2}\right]\frac{\sin\left(\frac{Ny}{2}\right)}{\sin\left(\frac{y}{2}\right)}
-i\sin\left[\frac{y(N+1)}{2}\right]\frac{\sin\left(\frac{Ny}{2}\right)}{\sin\left(\frac{y}{2}\right)}. \tag*{(167)} e − i 2 y ( N + 1 ) sin ( 2 y ) sin ( 2 N y ) = cos [ 2 y ( N + 1 ) ] sin ( 2 y ) sin ( 2 N y ) − i sin [ 2 y ( N + 1 ) ] sin ( 2 y ) sin ( 2 N y ) . (167)
where y y y can take possible values y = 0 , ± π / ( N + 1 ) , ± 2 π / ( N + 1 ) … y=0,\pm\pi/(N+1),\pm2\pi/(N+1)\ldots y = 0 , ± π / ( N + 1 ) , ± 2 π / ( N + 1 ) … . Let us look at the real part of Eq. (167) . For y → 0 y\to0 y → 0 , the ratio sin ( N y / 2 ) / sin ( y / 2 ) \sin(Ny/2)/\sin(y/2) sin ( N y /2 ) / sin ( y /2 ) approaches N N N , while the cosine factor becomes unity, so that the real part scales as N N N . By contrast, for y = m π / ( N + 1 ) y=m\pi/(N+1) y = mπ / ( N + 1 ) with odd m m m , the real part vanishes since cos ( m π / 2 ) = 0 \cos(m\pi/2)=0 cos ( mπ /2 ) = 0 , whereas for nonzero even m m m , the real part is of order unity. Thus, for large N N N , the contribution from y = 0 y=0 y = 0 dominates the real part of Eq. (167) , and we obtain
Re [ e − i y ( N + 1 ) / 2 sin ( N y / 2 ) sin ( y / 2 ) ] ≃ N δ y , 0 . (168) \operatorname{Re}\left[e^{-iy(N+1)/2}\frac{\sin(Ny/2)}{\sin(y/2)}\right]\simeq N\delta_{y,0}. \tag*{(168)} Re [ e − i y ( N + 1 ) /2 sin ( y /2 ) sin ( N y /2 ) ] ≃ N δ y , 0 . (168)
Let us next examine the large-N N N behavior of the imaginary part of Eq. (167) . For y → 0 y\to0 y → 0 , the imaginary part vanishes, unlike the real part. For y = m π / ( N + 1 ) y=m\pi/(N+1) y = mπ / ( N + 1 ) with even m m m , the imaginary part again vanishes since sin ( m π / 2 ) = 0 \sin(m\pi/2)=0 sin ( mπ /2 ) = 0 . For odd m m m , however, it is nonzero and can be written as − cot [ m π / 2 ( N + 1 ) ] -\cot[m\pi/2(N+1)] − cot [ mπ /2 ( N + 1 )] . For large N N N with fixed m m m , cot [ m π / 2 ( N + 1 ) ] ≃ 2 ( N + 1 ) / ( m π ) \cot[m\pi/2(N+1)]\simeq2(N+1)/(m\pi) cot [ mπ /2 ( N + 1 )] ≃ 2 ( N + 1 ) / ( mπ ) , showing that these contributions are of order N N N . Thus, in the large-N N N limit, the leading contribution to the imaginary part in Eq. (167) arises from odd m m m , while the remaining terms are subleading
Im [ e − i y ( N + 1 ) / 2 sin ( N y / 2 ) sin ( y / 2 ) ] ≃ − 2 N π ∑ m = 1 m o d d ∞ 1 m [ δ y , m π / ( N + 1 ) − δ y , − m π / ( N + 1 ) ] . (169) \operatorname{Im}\left[e^{-iy(N+1)/2}\frac{\sin(Ny/2)}{\sin(y/2)}\right]\simeq-\frac{2N}{\pi}\sum_{\substack{m=1\\m\ \mathrm{odd}}}^{\infty}\frac{1}{m}\left[\delta_{y,m\pi/(N+1)}-\delta_{y,-m\pi/(N+1)}\right]. \tag*{(169)} Im [ e − i y ( N + 1 ) /2 sin ( y /2 ) sin ( N y /2 ) ] ≃ − π 2 N m = 1 m odd ∑ ∞ m 1 [ δ y , mπ / ( N + 1 ) − δ y , − mπ / ( N + 1 ) ] . (169)
Plugging Eqs. (168) –(169) in Eq. (167) , we obtain
e − i y ( N + 1 ) / 2 sin ( N y / 2 ) sin ( y / 2 ) ≃ N δ y , 0 + 2 i N π ∑ m = 1 m o d d ∞ 1 m [ δ y , − m π / ( N + 1 ) − δ y , m π / ( N + 1 ) ] . (170) e^{-iy(N+1)/2}\frac{\sin(Ny/2)}{\sin(y/2)}
\simeq N\delta_{y,0}+\frac{2iN}{\pi}\sum_{\substack{m=1\\m\ \mathrm{odd}}}^{\infty}\frac{1}{m}\left[\delta_{y,-m\pi/(N+1)}-\delta_{y,m\pi/(N+1)}\right]. \tag*{(170)} e − i y ( N + 1 ) /2 sin ( y /2 ) sin ( N y /2 ) ≃ N δ y , 0 + π 2 i N m = 1 m odd ∑ ∞ m 1 [ δ y , − mπ / ( N + 1 ) − δ y , mπ / ( N + 1 ) ] . (170)
Substituting this form in Eq. (166) gives
ζ ~ ( Q , N ) ≃ N 4 ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) [ δ Q , ∣ k s − k s ′ ∣ − δ Q , π − ∣ π − k s − k s ′ ∣ + 2 i π ∑ m = 1 m o d d ∞ 1 m { δ Q , ∣ k s − k s ′ ∣ − m π / ( N + 1 ) − δ Q , ∣ k s − k s ′ ∣ + m π / ( N + 1 ) − δ Q , π − ∣ π − k s − k s ′ ∣ − m π / ( N + 1 ) + δ Q , π − ∣ π − k s − k s ′ ∣ + m π / ( N + 1 ) } ] . (171) \begin{aligned}
\widetilde{\zeta}(Q,N)\simeq{}&\frac{N}{4}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\Bigg[\delta_{Q,\lvert k_s-k_{s'}\rvert}-\delta_{Q,\pi-\lvert\pi-k_s-k_{s'}\rvert}\\
&+\frac{2i}{\pi}\sum_{\substack{m=1\\m\ \mathrm{odd}}}^{\infty}\frac{1}{m}\Bigg\{\delta_{Q,\lvert k_s-k_{s'}\rvert-m\pi/(N+1)}-\delta_{Q,\lvert k_s-k_{s'}\rvert+m\pi/(N+1)}\\
&\qquad-\delta_{Q,\pi-\lvert\pi-k_s-k_{s'}\rvert-m\pi/(N+1)}+\delta_{Q,\pi-\lvert\pi-k_s-k_{s'}\rvert+m\pi/(N+1)}\Bigg\}\Bigg]. \tag*{(171)}
\end{aligned} ζ ( Q , N ) ≃ 4 N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) [ δ Q , ∣ k s − k s ′ ∣ − δ Q , π − ∣ π − k s − k s ′ ∣ + π 2 i m = 1 m odd ∑ ∞ m 1 { δ Q , ∣ k s − k s ′ ∣ − mπ / ( N + 1 ) − δ Q , ∣ k s − k s ′ ∣ + mπ / ( N + 1 ) − δ Q , π − ∣ π − k s − k s ′ ∣ − mπ / ( N + 1 ) + δ Q , π − ∣ π − k s − k s ′ ∣ + mπ / ( N + 1 ) } ] . (171)
where the symbol ‘≃ \simeq ≃ ’ indicates that this expression is valid in the large-N N N limit. Recall that the wavenumber can be written as Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) where j ∈ { 1 , 2 , … , N } j\in\{1,2,\ldots,N\} j ∈ { 1 , 2 , … , N } . Eq. (171) then becomes
ζ ~ ( Q = π j N + 1 , N ) ≃ N 4 ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) [ δ j , ∣ s − s ′ ∣ − δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ + 2 i π ∑ m = 1 m o d d ∞ 1 m { δ j , ∣ s − s ′ ∣ − m − δ j , ∣ s − s ′ ∣ + m − δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ − m + δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ + m } ] . (172) \begin{aligned}
\widetilde{\zeta}\left(Q=\frac{\pi j}{N+1},N\right)\simeq{}&\frac{N}{4}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\Bigg[\delta_{j,\lvert s-s'\rvert}-\delta_{j,N+1-\lvert N+1-s-s'\rvert}\\
&+\frac{2i}{\pi}\sum_{\substack{m=1\\m\ \mathrm{odd}}}^{\infty}\frac{1}{m}\Bigg\{\delta_{j,\lvert s-s'\rvert-m}-\delta_{j,\lvert s-s'\rvert+m}\\
&\qquad-\delta_{j,N+1-\lvert N+1-s-s'\rvert-m}+\delta_{j,N+1-\lvert N+1-s-s'\rvert+m}\Bigg\}\Bigg]. \tag*{(172)}
\end{aligned} ζ ( Q = N + 1 π j , N ) ≃ 4 N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) [ δ j , ∣ s − s ′ ∣ − δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ + π 2 i m = 1 m odd ∑ ∞ m 1 { δ j , ∣ s − s ′ ∣ − m − δ j , ∣ s − s ′ ∣ + m − δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ − m + δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ + m } ] . (172)
where recall C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is a real function given in Eq. (73) .
Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with odd j j j
From Eq. (172) , s s s is odd and s ′ s' s ′ is even. As a result, both ∣ s − s ′ ∣ \lvert s-s'\rvert ∣ s − s ′ ∣ and N + 1 − ∣ N + 1 − s − s ′ ∣ N+1-\lvert N+1-s-s'\rvert N + 1 − ∣ N + 1 − s − s ′ ∣ are odd integers, while ∣ s − s ′ ∣ ± m \lvert s-s'\rvert\pm m ∣ s − s ′ ∣ ± m and N + 1 − ∣ N + 1 − s − s ′ ∣ ± m N+1-\lvert N+1-s-s'\rvert\pm m N + 1 − ∣ N + 1 − s − s ′ ∣ ± m are all even since m m m is odd. Using this result in Eq. (172) , we then see that for odd j j j , only the real part of Eq. (172) contributes while for even j j j , only the imaginary part contributes.
We first consider ζ ~ ( Q , N ) \widetilde{\zeta}(Q,N) ζ ( Q , N ) for Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with odd j j j , for which Eq. (172) reduces to
ζ ~ ( Q , N ) ≃ T 1 ( Q , N ) + T 2 ( Q , N ) , (173) \widetilde{\zeta}(Q,N)\simeq\mathcal{T}_1(Q,N)+\mathcal{T}_2(Q,N), \tag*{(173)} ζ ( Q , N ) ≃ T 1 ( Q , N ) + T 2 ( Q , N ) , (173)
with
T 1 ( Q , N ) = N 4 ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) δ j , ∣ s − s ′ ∣ , (174) \mathcal{T}_1(Q,N)=\frac{N}{4}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\delta_{j,\lvert s-s'\rvert}, \tag*{(174)} T 1 ( Q , N ) = 4 N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) δ j , ∣ s − s ′ ∣ , (174)
T 2 ( Q , N ) = − N 4 ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ . (175) \mathcal{T}_2(Q,N)=-\frac{N}{4}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\delta_{j,N+1-\lvert N+1-s-s'\rvert}. \tag*{(175)} T 2 ( Q , N ) = − 4 N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) δ j , N + 1 − ∣ N + 1 − s − s ′ ∣ . (175)
We demonstrated in § 7 and Figure 5 (a) that ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ in Eq. (173) and ∣ T 1 ( Q , N ) ∣ \lvert\mathcal{T}_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ in Eq. (174) attain their maxima at the same wavenumber Q max ( τ ) Q_{\max}(\tau) Q m a x ( τ ) . From Eq. (174) , ∣ T 1 ( Q , N ) ∣ \lvert\mathcal{T}_1(Q,N)\rvert ∣ T 1 ( Q , N )∣ is maximum when C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is maximum. As shown in Appendix § 17 § 18 , the maximum of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is obtained when s s s and s ′ s' s ′ in the large-N N N limit satisfy
2 γ τ ∣ cos ( k s ) − cos ( k s ′ ) ∣ = 2 π r , r = 1 , 2 , … , (176) 2\gamma\tau\lvert\cos(k_s)-\cos(k_{s'})\rvert=2\pi r,\qquad r=1,2,\ldots, \tag*{(176)} 2 γ τ ∣ cos ( k s ) − cos ( k s ′ )∣ = 2 π r , r = 1 , 2 , … , (176)
k s + k s ′ = π . (177) k_s+k_{s'}=\pi. \tag*{(177)} k s + k s ′ = π . (177)
Using Eq. (174) , the corresponding Fourier mode is Q max ( τ ) = ∣ k s − k s ′ ∣ Q_{\max}(\tau)=\lvert k_s-k_{s'}\rvert Q m a x ( τ ) = ∣ k s − k s ′ ∣ . Next using the trigonometric relation
∣ cos ( k s ) − cos ( k s ′ ) ∣ = 2 sin ( ∣ k s − k s ′ ∣ 2 ) sin ( k s + k s ′ 2 ) , (178) \lvert\cos(k_s)-\cos(k_{s'})\rvert=2\sin\left(\frac{\lvert k_s-k_{s'}\rvert}{2}\right)\sin\left(\frac{k_s+k_{s'}}{2}\right), \tag*{(178)} ∣ cos ( k s ) − cos ( k s ′ )∣ = 2 sin ( 2 ∣ k s − k s ′ ∣ ) sin ( 2 k s + k s ′ ) , (178)
in Eqs. (176) –(177) , we obtain
Q max ( τ ) = 2 sin − 1 ( π r 2 γ τ ) , r = 1 , 2 , … (179) Q_{\max}(\tau)=2\sin^{-1}\left(\frac{\pi r}{2\gamma\tau}\right),\qquad r=1,2,\ldots \tag*{(179)} Q m a x ( τ ) = 2 sin − 1 ( 2 γ τ π r ) , r = 1 , 2 , … (179)
Thus, for odd j j j , the dominant wavenumber at which ∣ ζ ~ ( Q , N ) ∣ 2 \lvert\widetilde{\zeta}(Q,N)\rvert^2 ∣ ζ ( Q , N ) ∣ 2 is maximal is determined by Eq. (179) in the large-N N N limit.
Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with even j j j
We now consider ζ ~ ( Q , N ) \widetilde{\zeta}(Q,N) ζ ( Q , N ) for Q = π j / ( N + 1 ) Q=\pi j/(N+1) Q = π j / ( N + 1 ) with even j j j . As discussed above, in this case only the imaginary part of Eq. (172) contributes, giving
ζ ~ ( Q , N ) ≃ i N 2 π ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) ∑ m = 1 m o d d ∞ 1 m [ δ j + m , ∣ s − s ′ ∣ − δ j − m , ∣ s − s ′ ∣ − δ j + m , N + 1 − ∣ N + 1 − s − s ′ ∣ + δ j − m , N + 1 − ∣ N + 1 − s − s ′ ∣ ] . (180) \widetilde{\zeta}(Q,N)\simeq\frac{iN}{2\pi}\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\sum_{\substack{m=1\\m\ \mathrm{odd}}}^{\infty}\frac{1}{m}\left[\delta_{j+m,\lvert s-s'\rvert}-\delta_{j-m,\lvert s-s'\rvert}-\delta_{j+m,N+1-\lvert N+1-s-s'\rvert}+\delta_{j-m,N+1-\lvert N+1-s-s'\rvert}\right]. \tag*{(180)} ζ ( Q , N ) ≃ 2 π i N s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) m = 1 m odd ∑ ∞ m 1 [ δ j + m , ∣ s − s ′ ∣ − δ j − m , ∣ s − s ′ ∣ − δ j + m , N + 1 − ∣ N + 1 − s − s ′ ∣ + δ j − m , N + 1 − ∣ N + 1 − s − s ′ ∣ ] . (180)
For fixed s s s and s ′ s' s ′ , the terms in the m m m -summation decay as 1 / m 1/m 1/ m . Hence, the leading contribution is obtained by taking m = 1 m=1 m = 1
∣ ζ ~ ( Q , N ) ∣ ≃ N 2 π ∣ ∑ σ = ± 1 σ [ ∑ s = 1 s o d d N ∑ s ′ = 1 s ′ e v e n N C s , s ′ ( τ , N ) { δ j + σ , ∣ s − s ′ ∣ − δ j + σ , N + 1 − ∣ N + 1 − s − s ′ ∣ } ] ∣ . (181) \left\lvert\widetilde{\zeta}(Q,N)\right\rvert\simeq\frac{N}{2\pi}\left\lvert\sum_{\sigma=\pm1}\sigma\left[\sum_{\substack{s=1\\s\ \mathrm{odd}}}^{N}\sum_{\substack{s'=1\\s'\ \mathrm{even}}}^{N}C_{s,s'}(\tau,N)\left\{\delta_{j+\sigma,\lvert s-s'\rvert}-\delta_{j+\sigma,N+1-\lvert N+1-s-s'\rvert}\right\}\right]\right\rvert. \tag*{(181)} ζ ( Q , N ) ≃ 2 π N σ = ± 1 ∑ σ s = 1 s odd ∑ N s ′ = 1 s ′ even ∑ N C s , s ′ ( τ , N ) { δ j + σ , ∣ s − s ′ ∣ − δ j + σ , N + 1 − ∣ N + 1 − s − s ′ ∣ } . (181)
Since j j j is even and σ = ± 1 \sigma=\pm1 σ = ± 1 , ( j + σ ) (j+\sigma) ( j + σ ) is odd. Therefore, for each σ \sigma σ , the two terms inside the square brackets in Eq. (181) have exactly the same structure as the two terms obtained in Eq. (173) for the odd-j j j case. Following the same line of reasoning, the dominant contribution to ∣ ζ ~ ( Q , N ) ∣ \lvert\widetilde{\zeta}(Q,N)\rvert ∣ ζ ( Q , N )∣ in Eq. (181) is then given by j max + σ = ∣ s − s ′ ∣ j_{\max}+\sigma=\lvert s-s'\rvert j m a x + σ = ∣ s − s ′ ∣ where the pair ( s , s ′ ) (s,s') ( s , s ′ ) satisfies Eqs. (176) –(177) . Translating it to Q max ( τ ) = π j max / ( N + 1 ) Q_{\max}(\tau)=\pi j_{\max}/(N+1) Q m a x ( τ ) = π j m a x / ( N + 1 ) , we obtain
Q max ( τ ) = 2 sin − 1 ( π r 2 γ τ ) , r = 1 , 2 , … , (182) Q_{\max}(\tau)=2\sin^{-1}\left(\frac{\pi r}{2\gamma\tau}\right),\qquad r=1,2,\ldots, \tag*{(182)} Q m a x ( τ ) = 2 sin − 1 ( 2 γ τ π r ) , r = 1 , 2 , … , (182)
in the large-N N N limit even for the even-j j j case. Thus, comparing this result with Eq. (179) , we see that in the large-N N N limit the dominant wavenumber at which the power spectrum attains its maximum is the same for both odd and even j j j , namely Q max ( τ ) = 2 sin − 1 ( π r 2 γ τ ) Q_{\max}(\tau)=2\sin^{-1}\left(\frac{\pi r}{2\gamma\tau}\right) Q m a x ( τ ) = 2 sin − 1 ( 2 γ τ π r ) .
Maximum of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N )
We now derive the conditions for maximizing C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) with respect to s s s and s ′ s' s ′ , as quoted in Eqs. (82) –(84) . Recall that C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is determined by the functions μ s ( η ) \mu_s(\eta) μ s ( η ) and ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) through Eq. (73) , with their explicit forms given in Eqs. (74) and (75) . Let us see how these functions depend on η \eta η for fixed s s s and s ′ s' s ′ . As illustrated in Figure 5 (b) for s = 19 s=19 s = 19 , s ′ = 60 s'=60 s ′ = 60 , N = 200 N=200 N = 200 , γ = 1 \gamma=1 γ = 1 and τ = 2 \tau=2 τ = 2 , the absolute values of both μ s ( η ) \mu_s(\eta) μ s ( η ) and ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) are sharply peaked as functions of η \eta η . We stated before that for large N N N , the peak locations of μ s ( η ) \mu_s(\eta) μ s ( η ) and ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) are given by
η μ s p e a k ≃ − 2 γ τ cos ( k s ) ( m o d 2 π ) , (183) \eta^{\mu_s\mathrm{peak}} \simeq-2\gamma\tau\cos(k_s)\pmod{2\pi}, \tag*{(183)} η μ s peak ≃ − 2 γ τ cos ( k s ) ( mod 2 π ) , (183)
η ν s ′ p e a k ≃ − 2 γ τ cos ( k s ′ ) ( m o d 2 π ) . (184) \eta^{\nu_{s'}\mathrm{peak}} \simeq-2\gamma\tau\cos(k_{s'})\pmod{2\pi}. \tag*{(184)} η ν s ′ peak ≃ − 2 γ τ cos ( k s ′ ) ( mod 2 π ) . (184)
To prove this, we set η = ( η μ s p e a k + ϵ ) \eta=(\eta^{\mu_s\mathrm{peak}}+\epsilon) η = ( η μ s peak + ϵ ) and expand the expression of μ s ( η ) \mu_s(\eta) μ s ( η ) in Eq. (74) for ϵ → 0 \epsilon\to0 ϵ → 0 . We obtain
μ s ( η ) ≃ sin ( k s ) e 2 i γ τ cos ( k s ) sin 2 ( k s ) − i ϵ 2 ∑ m ≠ s m o d d sin 2 ( k m ) + ϵ 2 ∑ m ≠ s m o d d sin 2 ( k m ) cot [ γ τ { cos ( k m ) − cos ( k s ) } ] . (185) \mu_s(\eta)\simeq\frac{\sin(k_s)e^{2i\gamma\tau\cos(k_s)}}{\sin^2(k_s)-\frac{i\epsilon}{2}\sum_{\substack{m\ne s\\m\ \mathrm{odd}}}\sin^2(k_m)+\frac{\epsilon}{2}\sum_{\substack{m\ne s\\m\ \mathrm{odd}}}\sin^2(k_m)\cot\left[\gamma\tau\{\cos(k_m)-\cos(k_s)\}\right]}. \tag*{(185)} μ s ( η ) ≃ sin 2 ( k s ) − 2 i ϵ ∑ m = s m odd sin 2 ( k m ) + 2 ϵ ∑ m = s m odd sin 2 ( k m ) cot [ γ τ { cos ( k m ) − cos ( k s )} ] sin ( k s ) e 2 iγ τ c o s ( k s ) . (185)
We next use
∑ m ≠ s m o d d sin 2 ( k m ) = N + 1 4 − sin 2 ( k s ) , (186) \sum_{\substack{m\ne s\\m\ \mathrm{odd}}}\sin^2(k_m)=\frac{N+1}{4}-\sin^2(k_s), \tag*{(186)} m = s m odd ∑ sin 2 ( k m ) = 4 N + 1 − sin 2 ( k s ) , (186)
∑ m ≠ s m o d d sin 2 ( k m ) cot [ γ τ { cos ( k m ) − cos ( k s ) } ] ≃ N α s ( τ ) , (187) \sum_{\substack{m\ne s\\m\ \mathrm{odd}}}\sin^2(k_m)\cot\left[\gamma\tau\{\cos(k_m)-\cos(k_s)\}\right]\simeq N\alpha_s(\tau), \tag*{(187)} m = s m odd ∑ sin 2 ( k m ) cot [ γ τ { cos ( k m ) − cos ( k s )} ] ≃ N α s ( τ ) , (187)
where α s ( τ ) \alpha_s(\tau) α s ( τ ) is an order-unity quantity in the large-N N N limit. Eq. (185) now becomes
μ s ( η ) ≃ sin ( k s ) e 2 i γ τ cos ( k s ) sin 2 ( k s ) − i ϵ N 8 + ϵ N α s ( τ ) 2 . (188) \mu_s(\eta)\simeq\frac{\sin(k_s)e^{2i\gamma\tau\cos(k_s)}}{\sin^2(k_s)-\frac{i\epsilon N}{8}+\frac{\epsilon N\alpha_s(\tau)}{2}}. \tag*{(188)} μ s ( η ) ≃ sin 2 ( k s ) − 8 i ϵ N + 2 ϵ N α s ( τ ) sin ( k s ) e 2 iγ τ c o s ( k s ) . (188)
For ϵ N ≫ 1 \epsilon N\gg1 ϵ N ≫ 1 , or equivalently ϵ ≫ 1 / N \epsilon\gg1/N ϵ ≫ 1/ N , the denominator in Eq. (188) becomes large and hence μ s ( η ) \mu_s(\eta) μ s ( η ) decays to zero. Thus, μ s ( η ) \mu_s(\eta) μ s ( η ) is appreciable only in a narrow region around η = η μ s p e a k \eta=\eta^{\mu_s\mathrm{peak}} η = η μ s peak , whose width shrinks with increasing N N N . This explains the strongly localized structure of ∣ μ s ( η ) ∣ 2 |\mu_s(\eta)|^2 ∣ μ s ( η ) ∣ 2 seen in Figure 5 (b).
Similarly, expanding ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) around η = ( η ν s ′ p e a k + ϵ ) \eta=(\eta^{\nu_{s'}\mathrm{peak}}+\epsilon) η = ( η ν s ′ peak + ϵ ) gives
ν s ′ ( η ) ≃ sin ( k s ′ ) e − 2 i γ τ cos ( k s ′ ) sin 2 ( k s ′ ) + i ϵ N 8 − ϵ N β s ′ ( τ ) 2 , (189) \nu_{s'}(\eta)\simeq\frac{\sin(k_{s'})e^{-2i\gamma\tau\cos(k_{s'})}}{\sin^2(k_{s'})+\frac{i\epsilon N}{8}-\frac{\epsilon N\beta_{s'}(\tau)}{2}}, \tag*{(189)} ν s ′ ( η ) ≃ sin 2 ( k s ′ ) + 8 i ϵ N − 2 ϵ N β s ′ ( τ ) sin ( k s ′ ) e − 2 iγ τ c o s ( k s ′ ) , (189)
where
β s ′ ( τ ) = 1 N ∑ m ≠ s ′ m e v e n sin 2 ( k m ) cot [ γ τ { cos ( k s ′ ) − cos ( k m ) } ] . (190) \beta_{s'}(\tau)=\frac{1}{N}\sum_{\substack{m\ne s'\\m\ \mathrm{even}}}\sin^2(k_m)\cot\left[\gamma\tau\{\cos(k_{s'})-\cos(k_m)\}\right]. \tag*{(190)} β s ′ ( τ ) = N 1 m = s ′ m even ∑ sin 2 ( k m ) cot [ γ τ { cos ( k s ′ ) − cos ( k m )} ] . (190)
Following the same argument as for μ s ( η ) \mu_s(\eta) μ s ( η ) , we find that ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) is appreciable only in a narrow region around η = η ν s ′ p e a k \eta=\eta^{\nu_{s'}\mathrm{peak}} η = η ν s ′ peak , whose width likewise shrinks with increasing N N N .
We now turn to C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) in Eq. (73) . As we showed, μ s ( η ) \mu_s(\eta) μ s ( η ) and ν s ′ ( η ) \nu_{s'}(\eta) ν s ′ ( η ) are sharply localized around η μ s p e a k \eta^{\mu_s\mathrm{peak}} η μ s peak and η ν s ′ p e a k \eta^{\nu_{s'}\mathrm{peak}} η ν s ′ peak , respectively, their product in Eq. (73) gives an appreciable contribution to the integral only when the two peak locations coincide. If the peaks are separated, their overlap is negligible and C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is correspondingly small. Hence, to get pronounced values of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) require η μ s p e a k = η ν s ′ p e a k \eta^{\mu_s\mathrm{peak}}=\eta^{\nu_{s'}\mathrm{peak}} η μ s peak = η ν s ′ peak . For example, for N = 200 N=200 N = 200 , γ = 1 \gamma=1 γ = 1 , and τ = 2 \tau=2 τ = 2 , the pair ( s , s ′ ) = ( 33 , 150 ) (s,s')=(33,150) ( s , s ′ ) = ( 33 , 150 ) gives η μ s p e a k ≈ η ν s ′ p e a k ≈ 2.8 \eta^{\mu_s\mathrm{peak}}\approx\eta^{\nu_{s'}\mathrm{peak}}\approx2.8 η μ s peak ≈ η ν s ′ peak ≈ 2.8 , and correspondingly C s , s ′ ( τ , N ) ∼ 10 − 3 C_{s,s'}(\tau,N)\sim10^{-3} C s , s ′ ( τ , N ) ∼ 1 0 − 3 . On the other hand, for ( s , s ′ ) = ( 19 , 60 ) (s,s')=(19,60) ( s , s ′ ) = ( 19 , 60 ) , we have η μ s p e a k ≈ 2.4 \eta^{\mu_s\mathrm{peak}}\approx2.4 η μ s peak ≈ 2.4 and η ν s ′ p e a k ≈ 3.9 \eta^{\nu_{s'}\mathrm{peak}}\approx3.9 η ν s ′ peak ≈ 3.9 . These peak locations are well separated, and the corresponding value is much smaller, C s , s ′ ( τ , N ) ∼ 10 − 5 C_{s,s'}(\tau,N)\sim10^{-5} C s , s ′ ( τ , N ) ∼ 1 0 − 5 .
Using Eqs. (183) –(185) , we therefore find that pronounced values of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) require
η μ s p e a k = η ν s ′ p e a k , ⟹ 2 γ τ ∣ cos ( k s ) − cos ( k s ′ ) ∣ = 2 π r , r = 1 , 2 , … , (191) \eta^{\mu_s\mathrm{peak}}=\eta^{\nu_{s'}\mathrm{peak}},\qquad\Longrightarrow\qquad2\gamma\tau\left|\cos(k_s)-\cos(k_{s'})\right|=2\pi r,\qquad r=1,2,\ldots, \tag*{(191)} η μ s peak = η ν s ′ peak , ⟹ 2 γ τ ∣ cos ( k s ) − cos ( k s ′ ) ∣ = 2 π r , r = 1 , 2 , … , (191)
where recall s s s is odd and s ′ s' s ′ is even. Noting that E s = − 2 γ cos ( k s ) E_s=-2\gamma\cos(k_s) E s = − 2 γ cos ( k s ) , Eq. (191) gives the condition for maximizing C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) , as was quoted earlier in Eq. (82) .
The corresponding form of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) is obtained by plugging Eqs. (188) –(189) in Eq. (73) and changing the integral from η \eta η to ϵ \epsilon ϵ .
C s , s ′ ( τ , N ) ≃ sin ( k s ) sin ( k s ′ ) 4 π Re [ ∫ − ∞ ∞ d ϵ 1 [ sin 2 ( k s ) − i ϵ N 2 + ϵ N α s ( τ ) 2 ] [ sin 2 ( k s ′ ) + i ϵ N 2 + ϵ N α s ′ ( τ ) 2 ] ] . (192) C_{s,s'}(\tau,N) \simeq\frac{\sin(k_s)\sin(k_{s'})}{4\pi}\operatorname{Re}\left[\int_{-\infty}^{\infty} \mathrm{d}\epsilon\frac{1}{\left[\sin^2(k_s)-\frac{\mathrm{i}\epsilon N}{2}+\frac{\epsilon N\alpha_s(\tau)}{2}\right]\left[\sin^2(k_{s'})+\frac{\mathrm{i}\epsilon N}{2}+\frac{\epsilon N\alpha_{s'}(\tau)}{2}\right]}\right]. \tag*{(192)} C s , s ′ ( τ , N ) ≃ 4 π sin ( k s ) sin ( k s ′ ) Re ∫ − ∞ ∞ d ϵ [ sin 2 ( k s ) − 2 i ϵ N + 2 ϵ N α s ( τ ) ] [ sin 2 ( k s ′ ) + 2 i ϵ N + 2 ϵ N α s ′ ( τ ) ] 1 . (192)
We have extended the limits of integration to ( − ∞ , ∞ ) (-\infty,\infty) ( − ∞ , ∞ ) since, in the large-N N N limit, the integrand is sharply localized around ϵ = 0 \epsilon=0 ϵ = 0 and decays rapidly away from this region, as explained before. We have further used α s ( τ ) ≃ − β s ′ ( τ ) \alpha_s(\tau)\simeq-\beta_{s'}(\tau) α s ( τ ) ≃ − β s ′ ( τ ) in the large-N N N limit. This relation follows directly from Eqs. (187) –(190) by replacing the sums with integrals and using Eq. (191) .
Performing the integration in Eq. (192) , we get
C s , s ′ ( τ , N ) ≃ 4 N sin ( k s ) sin ( k s ′ ) [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] 2 + 16 α s ( τ ) 2 [ sin 2 ( k s ) − sin 2 ( k s ′ ) ] 2 . (193) C_{s,s'}(\tau,N) \simeq\frac{4}{N}\frac{\sin(k_s)\sin(k_{s'})\left[\sin^2(k_s)+\sin^2(k_{s'})\right]}{\left[\sin^2(k_s)+\sin^2(k_{s'})\right]^2+16\alpha_s(\tau)^2\left[\sin^2(k_s)-\sin^2(k_{s'})\right]^2}. \tag*{(193)} C s , s ′ ( τ , N ) ≃ N 4 [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] 2 + 16 α s ( τ ) 2 [ sin 2 ( k s ) − sin 2 ( k s ′ ) ] 2 sin ( k s ) sin ( k s ′ ) [ sin 2 ( k s ) + sin 2 ( k s ′ ) ] . (193)
The same expression was given earlier in Eq. (83) .
To determine the largest possible value of C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) , we minimize the denominator in Eq. (193) . Its second term is nonnegative and disappears when sin 2 ( k s ) = sin 2 ( k s ′ ) \sin^2(k_s)=\sin^2(k_{s'}) sin 2 ( k s ) = sin 2 ( k s ′ ) . Because s s s and s ′ s' s ′ are distinct, k s ≠ k s ′ k_s\ne k_{s'} k s = k s ′ , and hence this condition implies
k s + k s ′ = π , (194) k_s+k_{s'}=\pi, \tag*{(194)} k s + k s ′ = π , (194)
as obtained previously in Eq. (84) . The corresponding maximum value is C s , s ′ ( τ , N ) ≃ 2 / N C_{s,s'}(\tau,N)\simeq2/N C s , s ′ ( τ , N ) ≃ 2/ N . Thus, Eqs. (192) –(194) give the conditions for maximizing C s , s ′ ( τ , N ) C_{s,s'}(\tau,N) C s , s ′ ( τ , N ) with respect to s s s and s ′ s' s ′ in the large-N N N limit.
Derivation of Eq. (128)
We stated in Eq. (127) that, for the nearest-neighbor hopping Hamiltonian in Eq. (2) , the first-detection amplitude to the left boundary ϕ n ( L ) ( x 0 ) \phi_n^{(L)}(x_0) ϕ n ( L ) ( x 0 ) is real for odd x 0 x_0 x 0 and purely imaginary for even x 0 x_0 x 0 (similarly for ϕ n ( R ) ( x 0 ) \phi_n^{(R)}(x_0) ϕ n ( R ) ( x 0 ) ). To derive this result, let us introduce the sublattice operator
Γ ^ ∣ x ⟩ = ( − 1 ) x ∣ x ⟩ . (195) \hat{\Gamma}|x\rangle=(-1)^x|x\rangle. \tag*{(195)} Γ ^ ∣ x ⟩ = ( − 1 ) x ∣ x ⟩ . (195)
The Hamiltonian in Eq. (2) satisfies
Γ ^ H Γ ^ = − H , Γ ^ U ( τ ) Γ ^ = e i H τ = U ( τ ) ∗ . (196) \hat{\Gamma}H\hat{\Gamma}=-H,\qquad\hat{\Gamma}U(\tau)\hat{\Gamma}=e^{\mathrm{i}H\tau}=U(\tau)^*. \tag*{(196)} Γ ^ H Γ ^ = − H , Γ ^ U ( τ ) Γ ^ = e i H τ = U ( τ ) ∗ . (196)
Moreover, Γ ^ ∣ x L ⟩ = − ∣ x L ⟩ \hat{\Gamma}|x_L\rangle=-|x_L\rangle Γ ^ ∣ x L ⟩ = − ∣ x L ⟩ and Γ ^ ∣ x R ⟩ = ( − 1 ) N ∣ x R ⟩ \hat{\Gamma}|x_R\rangle=(-1)^N|x_R\rangle Γ ^ ∣ x R ⟩ = ( − 1 ) N ∣ x R ⟩ . Therefore, for the survival operator S = ( 1 − ∣ x L ⟩ ⟨ x L ∣ − ∣ x R ⟩ ⟨ x R ∣ ) U ( τ ) S=\left(1-|x_L\rangle\langle x_L|-|x_R\rangle\langle x_R|\right)U(\tau) S = ( 1 − ∣ x L ⟩ ⟨ x L ∣ − ∣ x R ⟩ ⟨ x R ∣ ) U ( τ ) , we have
Γ ^ S = S ∗ Γ ^ , ⟹ Γ ^ S n − 1 = ( S ∗ ) n − 1 Γ ^ . (197) \hat{\Gamma}S=S^*\hat{\Gamma},\qquad\Longrightarrow\qquad\hat{\Gamma}S^{n-1}=(S^*)^{n-1}\hat{\Gamma}. \tag*{(197)} Γ ^ S = S ∗ Γ ^ , ⟹ Γ ^ S n − 1 = ( S ∗ ) n − 1 Γ ^ . (197)
We then obtain
ϕ n ( L ) ( x 0 ) = ⟨ x L ∣ U ( τ ) S n − 1 ∣ x 0 ⟩ , = − ⟨ x L ∣ Γ ^ U ( τ ) S n − 1 ∣ x 0 ⟩ , = − ⟨ x L ∣ U ( τ ) ( S ∗ ) n − 1 Γ ^ ∣ x 0 ⟩ = ( − 1 ) 1 + x 0 ( ϕ n ( L ) ( x 0 ) ) ∗ . (198) \begin{aligned}
\phi_n^{(L)}(x_0)&=\langle x_L|U(\tau)S^{n-1}|x_0\rangle,\\
&=-\langle x_L|\hat{\Gamma}U(\tau)S^{n-1}|x_0\rangle,\\
&=-\langle x_L|U(\tau)(S^*)^{n-1}\hat{\Gamma}|x_0\rangle\\
&=(-1)^{1+x_0}\left(\phi_n^{(L)}(x_0)\right)^*. \tag*{(198)}
\end{aligned} ϕ n ( L ) ( x 0 ) = ⟨ x L ∣ U ( τ ) S n − 1 ∣ x 0 ⟩ , = − ⟨ x L ∣ Γ ^ U ( τ ) S n − 1 ∣ x 0 ⟩ , = − ⟨ x L ∣ U ( τ ) ( S ∗ ) n − 1 Γ ^ ∣ x 0 ⟩ = ( − 1 ) 1 + x 0 ( ϕ n ( L ) ( x 0 ) ) ∗ . (198)