First Hitting Times on a Quantum Computer:
Tracking vs. Local Monitoring, Topological Effects, and Dark States

We consider a tight-binding chiral quantum walk [34] on a triangle graph with complex edge weights in the presence of a magnetic flux (Fig. 1) for two different monitoring protocols. The localized states are $\ket{\bf{0}},\ket{\bf{1}}$ and $\ket{\bf{2}}$ and the initial state is $\ket{\psi_{{\rm in}}}$. The measurement-free evolution is defined via unitary dynamics

$\displaystyle U=\exp(-iH\tau)$

(1)

with the time intervall $\tau$ between detection attempts, and $\hbar=1$. The Hamiltonian $H$ is

$$H=-\gamma\ e^{i\alpha}\left(\ket{\bf{0}}\bra{\bf{1}}+\ket{\bf{1}}\bra{\bf{2}}+\ket{\bf{2}}\bra{\bf{0}}\right)+{\rm h.c.}\ .$$

(2)

Here h.c. stands for hermitian conjugate. $\alpha$ and $\gamma$ are control parameters that model the effect of magnetic flux and hopping amplitude, respectively. The time-reversal symmetry is broken when $\exp(i\alpha)\neq 1$. The eigenvalues of the Hamiltonian are

$$E_{k}=-2\gamma\cos\left({2\pi k\over 3}+\alpha\right),\ k=0,1,2\ .$$

(3)

When $\alpha=0$, we have a pair of distinct energy levels, and more generally when the flux $\alpha\neq 0$, we typically remove the two-fold degeneracy and the number of distinct energy levels is three [35].

The removal of energy degeneracy will have a profound effect on the mean return time and the dark states in the system as we will soon show.

To understand the physical properties of the monitored quantum walks, it is useful to visualize at which parameters $\alpha$ and $\gamma\tau$ (our control parameters), the eigenvalues of $U$ are degenerate

$\displaystyle\exp(-iE_{k}\tau)=\exp(-iE_{l}\tau)\ ,(k\neq l),\ k,l=0,1,2$

(4)

in a phase factor matching diagram; see Fig. 2. The analytical results are given in Appendix A. For $\alpha=0$ two phase factors match due to the degeneracies of the eigenenergies, while three phase factors match for $\gamma\tau=2\pi j/3$ with $j\in\mathbb{N}$. Other points where three phase factors match are, e.g., $(\alpha,\gamma\tau)=(\pi/6,2\pi/\sqrt{3}),(\pi/3,2\pi j/3),(\pi/2,2\pi/\sqrt{3})$, etc. We will later see how this matching phase factor diagram is related to the hitting time problem as well as to the dark states. We define a target state $\ket{\bf{0}}$ which we monitor each $\tau$ units of time using two measurement protocols. These are described in the next two subsections.

First, we consider the stroboscopic measurement scheme [9], which we call the local measurement protocol or the on-site protocol. Here, measurements are performed every $\tau$ units of time. Between measurements the dynamics is controlled by the unitary time evolution $U=\exp(-iH\tau)$. The measurement is made locally at $\ket{\bf{0}}$ to detect this state; see Fig. 1 (left panel). Here, the measurements yield either a no or a yes, that is, the quantum walker is either not found or is found on $\ket{\bf{0}}$. Mathematically, the measurement is described by the projector $\ket{\bf 0}\bra{\bf 0}$. A typical string of measurements is, for instance,

$$\{\mbox{no},\mbox{no},\mbox{no},\mbox{yes},\cdots\}.$$

In this case, the first hitting time is $4\tau$. The first click yes gives the random number of measurements, denoted $n$, needed to detect the quantum walker in the target state $\ket{\bf{0}}$. In this case, unless we deal with a two-state system, the information obtained in this way does not specify the state function of the walker after each null measurement. Hence, after obtaining a no-result, we do not know what the amplitudes are for states $\ket{\bf{1}}$ and $\ket{\bf{2}}$.

Second, we consider the tracking protocol [32], with the target state $\ket{\bf{0}}$ where we record every $\tau$ units of time the position of the walker on the graph; see Fig. 1 (right panel). In this stroboscopic protocol, measurements are made at times $\tau,2\tau,\dots,20\tau$. In a typical realization of the process, we find a result, which is given by the eigenvalues of the position operator, for example

$$\{\bf{1},\bf{2},\bf{0},\cdots\}$$

or using the same initial condition and the same unitary

$$\{\bf{1},\bf{2},\bf{1},\bf{2},\bf{1},\bf{0},\cdots\}.$$

The first hitting time is clearly random and in these two examples it is $3\tau$ and $6\tau$, respectively.

From the basic postulates of quantum measurement theory [36], we know the state of the system immediately after each measurement. This will be the eigenstate corresponding to the eigenvalue just recorded, assuming strong measurements. For the first example, we will therefore find

$$\{\ket{\bf{1}},\ket{\bf{2}},\ket{\bf{0}},\cdots\}.$$

This implies that with this type of measurement we know precisely what the state of the system along the measured path is.

In the return problem the target state and the initial state are the same. This choice is also called the reccurence time problem. It will be studied with both protocols. Thus, we start with state $\ket{\psi_{\rm in}}=\ket{\bf{0}}$. Just before the first measurement at time $\tau$, the state function is $\ket{\psi(\tau)}=U\ket{\psi_{\rm in}}$. For the tracking protocol, we then measure the location of the walker on the triangle graph. Suppose that the first measurement yields $\bf{1}$, i.e., the system is projected to state $\ket{\bf{1}}$. We then continue and evolve the system until the second measurement according to $U\ket{\bf{1}}$, and measure again. For the on-site protocol the new state is the projection to the other accessible states if we do not detect the quantum walker on site $\ket{\bf 0}$. Note that when we study dark state physics, we use other initial conditions, see below.

To record the trajectory on a quantum computer we have to encode the tight-binding Hamiltonian into a qubit Hamiltonian:

	$\displaystyle H=-\frac{1}{2}\gamma[\cos(\alpha)(\sigma_{1}^{x}+\sigma_{2}^{x}+\sigma_{1}^{z}\sigma_{2}^{x}+\sigma_{1}^{x}\sigma_{2}^{z}+\sigma_{1}^{x}\sigma_{2}^{x}+\sigma_{1}^{y}\sigma_{2}^{y})$
	$\displaystyle+\sin(\alpha)(\sigma_{1}^{y}-\sigma_{2}^{y}+\sigma_{1}^{y}\sigma_{2}^{z}-\sigma_{1}^{z}\sigma_{2}^{y}+\sigma_{1}^{x}\sigma_{2}^{y}-\sigma_{1}^{y}\sigma_{2}^{x})]$

where $\sigma_{x}$, $\sigma_{y}$ and $\sigma_{z}$ are the Pauli matrices. The Hamiltonian $H$ defines two disconnected subspaces, the first composed of states $\ket{00},\ket{01},\ket{10}$ and the second of $\ket{11}$. The unitary evolution operator $U(\tau)=\exp(-iH\tau)$ is decomposed into unitary gates using Cartan’s decomposition [37]. This allows us to vary $\tau$ without increasing the depth of the circuits.

For the on-site protocol, we need to detect the state $\ket{\bf{0}}$ while we do not want to receive the information that allows us to distinguish the states $\ket{\bf{1}}$ and $\ket{\bf{2}}$. This can be achieved by measuring only one qubit, which we define as the right qubit in the Dirac notation. We use the following mapping between the qubits and the representation of the spatial states: $\ket{01}\rightarrow\ket{\bf{0}},\ket{10}\rightarrow\ket{\bf{2}}\ \mbox{and}\ \ket{00}\to\ket{\bf{1}}$. Measuring the right qubit in state $\ket{0}$ does not give any information to distinguish the states $\ket{\bf{1}}=\ket{10}$ and $\ket{\bf{2}}=\ket{00}$ but measuring the right qubit in state $\ket{1}$ the system is in $\ket{\bf{0}}=\ket{01}$ with certainty.

For the tracking protocol, we measure the states of both qubits, and thus after each measurement we find the outcome $00$, $01$ or $10$. For error prevention reasons, we use a different mapping between the qubit and the spatial state representation: $\ket{00}\rightarrow\ket{\bf{0}},\ket{01}\rightarrow\ket{\bf{1}}\ \mbox{and}\ \ket{10}\to\ket{\bf{2}}$. The reason is that experimentally there is an asymmetry in the measured spectrum. For example, it is much more likely to migrate from a state $\ket{01}$ to $\ket{00}$ than from $\ket{00}$ to $\ket{01}$ due to decay processes. Therefore, we define $\ket{00}$ as our initial state.

As an error suppression strategy, we use dynamical decoupling by inserting two $X$-gates at specific intervals on the qubit, which is not measured, to keep it coherent [38]. We compute the time for the first detection or the first hitting time $n\tau$ and evaluate its statistical properties as a function of $\alpha$ and $\gamma$. Effectively, the measurement process has ended when we find the target state for the first time. The result obtained from the measurements is a string of size $N$ which yields what we call a trajectory (a string of $N$ bits for the on-site protocol and $N$ times two bits for the tracking protocol for 32,000 runs). In practice, we cannot stop the experiments in the middle of the process, i.e., currently there is no conditional abort option on IBM quantum computers. We therefore shorten the trajectories using classical post-processing at the bit position where the target site was measured.

Now we define statistical observables of interest. They can be estimated from the sample paths and will depend on the parameters of the model, such as $\gamma\tau$, $\alpha$, the initial state and the measurement protocol. Let $F_{n}$ be the probability that the quantum walker is detected for the first time in the target state at the $n$-th measurement [9, 32]. Thus $F_{1}$ is the probability that the target state is recorded at the first measurement event, while $F_{2}$ describes the case where the state was recorded for the first time in the second attempt. $\langle n\rangle$ is the conditional mean number of events until detection

$\displaystyle\langle n(N)\rangle=\frac{\sum_{n=1}^{N}nF_{n}}{\sum_{n=1}^{N}F_{n}}\ ,$

(5)

for $\sum_{n=1}^{N}F_{n}\neq 0$. Hence, $\tau\langle n(N)\rangle$ is the mean time until detection. From now on, we will call this physical quantity mean return time for brevity. We also define the total detection probability $P_{{\rm det}}$

$\displaystyle P_{{\rm det}}(N)=\sum_{n=1}^{N}F_{n}\ ,$

(6)

which is the probability that a detection event was recorded. The estimation of this probability comes from the number of strings without detection. In an ideal situation with infinite resolution, these are obtained through $\lim_{N\to\infty}P_{{\rm det}}(N)=\sum_{n=1}^{\infty}F_{n}$ and if the latter is unity, that is, if the system is detected with probability one, $\lim_{N\to\infty}\langle n(N)\rangle=\sum_{n=1}^{\infty}nF_{n}$. As mentioned in the Introduction, even for the triangle model under study we can find cases where $P_{{\rm det}}<1$ due to destructive interference, as explained below.

In the experiment, we obtain the statistical properties of the first detection time for a finite number of measurements $N$ giving rise to finite resolution effects. Per trajectory, the first detection time is $n\tau$, where the integer $n$ is the random number of measurements until the first detection of the target state. Clearly in our experiments $n\leq N$. Note that the obtained strings are vectors of size $N$. It is possible that the string did not contain the target state $\ket{\bf{0}}$, a fact that will become important later, close to topological transitions.

The theorem of recurrence for on-site measurements reads [5, 6]

$$\langle n\rangle=\mbox{number of distinct phase factors}\ \ \exp(-iE_{k}\tau)$$

(7)

where $E_{k}$ are the eigenstates of $H$ [Eq. (3)] for the triangle model. From Eq. (3) it becomes clear why the phase factor matching diagram is a very useful tool. For example, for the choice of parameters corresponding to blue circles in Fig. 2 we expect $\langle n\rangle=1$, while for the colored lines (besides blue) $\langle n\rangle=2$, and anywhere else $\langle n\rangle=3$. This holds in general under the condition that the energy states, i.e., eigenstates of $H$, denoted $\ket{E_{k}}$ have finite overlap with the detected state. For example, in the Zeno limit $\tau\to 0$, the three phase factors merge, and hence $\langle n\rangle$ is equal to one. This is expected for the return problem, since we start at the measured state, and hence the first measurement detects it when $\tau=0$. Phase factor matching conditions correspond to removal of a state from the effective Hilbert space. If phase factor match $\exp(-iE_{1}\tau)=\exp(-iE_{2}\tau)$, the state $\psi\sim\langle{\bf 0}|E_{1}\rangle|E_{2}\rangle-\langle{\bf 0}|E_{2}\rangle|E_{1}\rangle$ cannot be detected. Thus, the phase factor matching choices of $\gamma\tau$ and $\alpha$ reduce the dimension of the Hilbert space, which can be shown to lead to faster detection compared to the cases where the phase factors do not match.

We provide more details in Appendix B.

Except for special model parameters, the mean return time $\langle n\rangle$ is the number of states in the system, which is three for the triangle model under study. The result $\langle n\rangle=3$ can be viewed as a classical result. More specifically, consider a classical random walk coupled to a heat bath in the infinite temperature limit. In this case, in a steady state, all states are equally likely, namely the occupation probability is $1/3$. Then from the classical Kac theorem for the return problem $\langle n\rangle=3$ [31]. At least in the classical domain, this is easy to understand, since the probability of measuring the walker in the target state in the first measurement is $1/3$, in the second $(2/3)(1/3)$ etc. and hence $\langle n\rangle=\sum_{n=1}^{\infty}n(2/3)^{n-1}(1/3)=3$. Thus, tracking, for most of the choices of $\tau$, drives the system to a classical limit. The quantum aspect of the hitting-time process is found for special values of $\tau$ that capture the revival of the wave packet and hence the underlying periodicity of the quantum dynamics. Namely, there exist revival times where the initial wave function returns to its original state in addition to an unimportant phase. Then if we measure at that time, the quantum walker is detected with probability one in the target state in the first measurement and hence $\langle n\rangle=1$. In particular, when $\tau=0$ we have $\langle n\rangle=1$. In the diagram in Fig. 2 these special sampling times correspond to three phase factors matching, shown as blue dots and blue lines. Clearly, the idealized theory predicts that as we vary $\gamma\tau$ and $\alpha$, when three phase factors match, we will see a dip or a transition in $\langle n\rangle$. Thus, $\langle n\rangle$ will jump from $\langle n\rangle=3$ to $\langle n\rangle=1$ and back. In Appendix C, following Ref. [32] we provide more details of the theory by introducing the corresponding Markov or stochastic matrix.

Fig. 3 (a) shows the parameter regime (red dashed line) utilized for the computation of the mean return time as well as the classical simulation of the mean return time as a function of $\gamma\tau$ and $\alpha$ for $N=20$. In comparison to Fig. 2 the areas for $\langle n\rangle=1$ (dark blue) and $\langle n\rangle=2$ (medium blue) are broadened. The figure clearly shows how the conditions for phase factor matching can lead to jumps in $\langle n\rangle$.

For the experimental computation of $\langle n\rangle$ we utilize the quantum circuit in Fig. 3 (b). The on-site protocol can be emulated by measuring the upper qubit after each unitary $U$.

Let us first discuss the results of the asymptotic theory using Eq. (7). In Fig. 3 (c) - (f), we consider $\langle n\rangle$ versus the dimensionless sampling time $\gamma\tau$, as well as the magnetic flux $\alpha$ (black solid line). For almost any $\tau$, on average, we return after two measurements, i.e., $\langle n\rangle=2$. This holds when the magnetic flux is turned off ($\alpha=0$). In contrast, when $\alpha\neq 0$ we have besides very special choices of $\tau$, $\langle n\rangle=3$ which is related to the fact that the magnetic flux $\alpha>0$ removes the degeneracy by breaking the time-reversal symmetry and the rotational invariance. In addition, one observes typical sudden plunges in $\langle n\rangle$. These are found for special values of $\gamma\tau$ and $\alpha$. As mentioned in the Introduction, these transitions, e.g., dips in $\langle n\rangle$ when plotted versus $\gamma\tau$ or $\alpha$ are related to a discontinuous change of the topology of the underlying generating function [5, 6, 11].

Quantization of the mean return time of the experimental result (light blue circles) is clearly evident for most values of $\gamma\tau$ and $\alpha$. Far from transitions in $\langle n\rangle$, theory predicts, and the experiments confirm, that for on-site measurements, with and without magnetic flux, $\langle n\rangle\approx 2$ [Fig. 3 (c)] and $\langle n\rangle\approx 3$ [Fig. 3 (d)-(f)], respectively. Notwithstanding these successes, theory ($N\rightarrow\infty$) and experiment depart close to the parameters in which the broadening of the dark blue and medium blue areas is obvious in Fig. 3 (a). When we choose path (III), in the ($\alpha$, $\gamma\tau$) plain, we are in the vicinity of a phase factor matching curve, though strictly not exactly on it. Since the number of measurements $N$ is finite in the experimental study, we are effectively witnessing a topological state with $\langle n\rangle=2$, instead of $\langle n\rangle=3$ predicted by theory, see Fig. 3 (e). We observe not only a broadening of the resonances, but also additional resonances for finite $N$ that are not present in the theoretical graph ($N\rightarrow\infty$) [Fig. 3 (d)].

Furthermore, experiment and asymptotic theory do not match when triplets of dips of $\langle n\rangle$ are in the vicinity of one another on path (II). Indeed, when $\alpha=0.5$, we see from the theoretical prediction in Fig. 3 (d) three nearby dips (black solid line), while in reality, we see one wide resonance (light blue circles). Thus, fine details are wiped out near the point where all three phase factors are almost matching.

What is the cause of the deviation between theory (black solid lines) and experiment (light blue circles) close to these topological transitions? It might be due to noise, readout errors, or finite resolution effects. Without a doubt, all of these effects may play some role. However, for the mean return time under study, we concluded that the main factor is the finite resolution of the experiment. The unitary dynamics and measurements implemented on the quantum computer work well on the time and size scale of our experiment, indicated by the fact that the experimental points lie close to the simulation on a classical computer (blue dashed lines). Finite resolution effects are discussed in detail in Sec. VI.

Fig. 4 (a) shows the parameter regime (red dashed line) utilized for the computation of the mean return time as well as the simulation of the mean return time as a function of $\gamma\tau$ and $\alpha$ for $N=20$ for the tracking protocol. Compared to Fig. 2 the areas for $\langle n\rangle=1$ (dark blue) are broadened and in comparison to the on site protocol Fig. 3 (a) we observe the dark color ($\langle n\rangle$) more sporadically, i.e., the lines where only two phase factors match are irrelevant. For the experimental computation of $\langle n\rangle$ we implement the quantum circuit in Fig. 4 (b), where both qubits are measured after applying the unitary $U$.

We start by discussing the theory for $N\rightarrow\infty$ (black solid line). The tracking protocol gives with and without flux ($\alpha$) $\langle n\rangle=3$, for nearly all $\tau$, according to theoretical predictions. Note that any small deviation of $\tau$ from the special revival times or $\tau=0$, that is, the Zeno limit, yields $\langle n\rangle=3$ as shown in Fig. 4 (c) - (f) as a black solid line. For the finite magnetic flux $\alpha=0.5$ in Fig. 4 (d), there are no special sampling times $\tau$ except $\tau=0$. (Similar for $\gamma\tau=3\pi/\sqrt{3}$ and finite $\alpha$ in Fig. 4 (e)). When three phase factors merge, the deviations between the theory valid for $N\to\infty$ and experiments (light blue circles) for $N=20$ are large. We obtain a large deviation for the parameters in which the broadening of the dark blue areas is observed in Fig. 4 (a) between experiment and assymptotic theory. The broadening is present in direction of $\gamma\tau$ and $\alpha$ which leads not only to a broadening of the resonances, but also to additional resonances for finite $N$ that are not present in the theoretical graph ($N\rightarrow\infty$) [Fig. 4 (d) and (e)] but in the finite $N$ simulation (blue dashed line). Similarly to the on-site protocol, the deviation of the experimental values from the asymptotic theory is due to finite-resolution effects. These are discussed in detail in Sec. VI.

Consider the case $\alpha=0$, the initial condition

$$\ket{\psi}_{0}={\ket{\bf{1}}+e^{i\phi}\ket{\bf{2}}\over\sqrt{2}}$$

(8)

and as before the hitting process ends when we detect the system in state $\ket{\bf{0}}$. We study the influence of the phase $\phi$ on the detection process.

For on-site measurement and when $\phi=0$ the system is detected with probability one, except for special values of sampling times. In contrast, if $\phi=\pi$, the detection probability is zero. The latter case is caused by destructive interference and the symmetry of the problem. When $\phi=\pi$ the current from $\ket{\bf{1}}$ to the detected state is minus the current from $\ket{\bf{2}}$, which means that for all times, including between measurements, the amplitude of the detected state is zero. This can break down when $\alpha\neq 0$, i.e., when the symmetry is broken. A different behavior is found for the tracking protocol. Then, after the first measurement, the wave packet is spatially localized in the system; hence, the symmetry is broken by the measurements, and the walker is eventually detected. Thus, a dark state in the on-site protocol can be bright for the tracking method.

A comparison between the two protocols is presented in the three-dimensional graphs in Fig. 5 (a) and (b). The dark color in the figure corresponds to dark states. For the on-site protocol we find using methods in [12]

$$\lim_{N\to\infty}P_{{\rm det}}(N)=\left\{\begin{array}[]{l l}0,&\gamma\tau=2\pi k/3\\ &\\ {1+\cos(\phi)\over 2},&\mbox{otherwise}\end{array}\right.$$

(9)

The first line for $k=0$ corresponds to the well-studied Zeno limit. The Schrödinger equation being first order in time implies that the amplitude in the detected state $\ket{\bf{0}}$, at time $\tau$, is proportional to $\tau$ but according to the rules of quantum measurement theory, the probability of detecting the quantum walker is proportional to $\tau^{2}$, and hence the walker is not detected at all. It should be noted, however, that for any small value of $\tau$ and when $N\to\infty$ the walker is eventually detected, which means that the limits of large $N$ and small $\tau$ do not commute. Other values of $k=1,2,3,...$ are choices of $\tau$ corresponding to revival times, when the wave function returns to its original state, and hence the stroboscopic measurements cannot click yes even after many measurements and the quantum walker is never detected. The second line in Eq. (9) corresponds to the destructive interference that is strongest when $\phi=\pi$. It is a robust effect in the sense that, for on-site measurements, it will hold for any choice of measurement times.

We recorded $P_{{\rm det}}$ on a quantum computer using on-site measurements; see Fig. 5 (a). We use $N=10$ measurements for each value of $\gamma\tau$ and $\phi$. The three-dimensional plot, Fig. 5 (a), clearly demonstrates the above mentioned features as horizontal and vertical dark stripes. In contrast, in the tracking protocol [Fig. 5 (b)], we see only the horizontal lines corresponding to the Zeno effect and the revivals, since, as mentioned, tracking breaks the symmetry in the system needed to maintain destructive interference effects.

Clearly, the results in Fig. 5 (a) and (b) using the quantum computer and in Fig. 12 in Appendix D using simulations indicate a very good agreement between theory and experiment.

When $\alpha\neq 0$ the detection probability plots presented so far, could in principle be replaced with a study of $P_{{\rm det}}$ versus of $\alpha,\phi$ and $\gamma\tau$. To simplify the matter, in the experiments and simulations below, the initial condition is

$$\ket{\psi}_{0}=\ket{\bf{1}}$$

and therefore the control parameters are $\gamma\tau$ and $\alpha$. In Fig. 5 (c) and (d), we plot the detection probability for the on-site and tracking protocols, respectively, and find a striking difference. For comparison, the corresponding simulation is shown in Fig. 12 in Appendix D.

The main features of these figures can be explained as follows. We start with the on-site protocol. Consider the choice of parameters for which two phase factors merge. For example, $\exp(-iE_{1}\tau)=\exp(-iE_{2}\tau)$. We also denote the corresponding eigenstates with $\ket{E_{1}}$ and $\ket{E_{2}}$, respectively. Then an initial state

$$\ket{\psi}_{{\rm in}}\sim\ \bra{\bf{0}}E_{1}\rangle\ket{E_{2}}-\bra{\bf{0}}E_{2}\rangle\ket{E_{1}}$$

(10)

is clearly orthogonal with respect to the detected state $\ket{\bf{0}}$. In this case, every unitary step yields the same global phase shift $\exp(-iE_{k}\tau)$ with $k=1,2$. Therefore, the amplitude of the target state will be zero for any subsequent measurement. In other words, this initial condition is completely dark and cannot be detected at all. Any initial state that is not orthogonal to this state cannot be detected with probability one; hence in our example, the initial state under study is not bright when two phase factors merge.

When the three phase factors merge, a stronger effect will occur on the triangle graph. Then the wave function, every $\tau$ units of time, returns to its initial state and hence revives. Since in the case studied in this section, the initial state is orthogonal to the detected state, clearly one cannot detect the walker. The same holds for the tracking measurement. The walker cannot be detected in the first measurement, nor in any of the repeated measurements. Hence, for on-site measurements, nonbright states are found when two or three phase factors merge. For the tracking protocol, dark states are found only when three phase factors merge; otherwise, the state is detected with probability one. Therefore, to better understand the detection, we return to the phase factor matching diagram in Fig. 2, where we plot in the ($\gamma\tau$, $\alpha$) plane curves when two phase factors coincide. That is, $\exp(-iE_{1}\tau)=\exp(-iE_{2}\tau)$ gives one branch of this relation between $\gamma\tau$ and $\alpha$ and similarly for pairs $(E_{1},E_{3})$ and $(E_{2},E_{3})$. The plot of this information requires the energy levels that depend, of course, on $\alpha$. Different pairs of matching phase factors are presented in different colors. When three phase factors merge (circles), i.e., when curves cross each other in Fig. 2, we see that for the tracking protocol, the detection probability in Fig. 5 (d) is small or zero. These crossing events appear as dark blue circles in Fig. 5 (d). In the small $\gamma\tau$ limit, we see a clear dark stripe, representing Zeno dark states. When $\tau=0$ the three phase factors coincide.

When two phase factors match, in Fig. 2, we see for the on-site measurements, i.e., in Fig. 5 (c), the corresponding darker color. An issue for future study is whether one can distinguish between the colors in Fig. 2, using repeated measurements. If we want to remove curves in Fig. 5 (c), at least partially, then we have to choose an initial state orthogonal to the dark-energy state in Eq. (10) which will be bright corresponding to the merging of the two phase factors $\exp(-iE_{1}\tau)=\exp(-iE_{2}\tau)$. Adding a phase to the initial state, like in Eq. (8) can also add features to these plots. Can any of this be observed in the experiment? The results presented in Fig. 5 are very convincing in this regard.

If we consider the return problem, the detection probability is at its maximum, and hence a figure like Fig. 5 (c) would be colored white and, therefore, not very informative. In the dark area where $P_{\rm det}=0$ the mean return time is $\langle n\rangle=1$ for both protocols. On brighter lines where the two phase factors match, we get $\langle n\rangle=2$ (return) and $0<P_{\rm det}<1$ (transfer) for the on-site protocol. Also, here finite resolution effects are important, as for $N\rightarrow\infty$ the dark lines will become infinitely thin.

Finally, we pose the question of whether the total detection probability changes if we choose $\ket{\psi}_{0}=\ket{\bf{2}}$. The total detection probability for the transition from $\ket{\bf{1}}$ to $\ket{\bf{0}}$ and $\ket{\bf{2}}$ to $\ket{\bf{0}}$, respectively, is different and depends on the direction of the magnetic flux $\alpha$ for $N=10$. The difference of the total detection probabilities is shown in Fig. 6, where an asymmetry is observed around the broadened phase-matching areas and lines. The total detection time of the chiral quantum walker for the transfer depends on the direction (clockwise or counter clockwise) and also on the sign of the magnetic flux. Breaking of the time-reversal symmetry can therefore enhance or suppress the total detection and therefore provides a directional control for certain parameters.

We present an exact solution to the problem related to our experimental setup for both on-site and tracking measurement protocols. Previously, the broadening effect was studied in more generality for the on-site measurements, although the asymptotic behavior of large $N$ was studied there [39]. In the following we will discuss the broadening of the return time $\langle n\rangle$ for zero magnetic flux in detail.