Discrete-Time Quantum Random Walks

• Discrete-Time Quantum Random Walks are lattice-based unitary dynamics where a coin operation controls conditional shifts, leading to interference and ballistic transport.
• The framework accommodates various coin parameterizations, inhomogeneous coins, and continuum limits that reproduce phenomena like Dirac dynamics and resonant tunneling.
• Applications span quantum simulation, neutrino oscillations, and spatial epidemic models, providing insights into disorder effects, entanglement generation, and transport characteristics.

A discrete-time quantum random walk (QRW) framework is a lattice-based unitary dynamics in which a finite-dimensional internal degree of freedom—the coin, spin, or chirality—controls conditional motion on a discrete position space, so that one step is generated by a coin operation followed by a conditional shift. In the standard coined formulation this yields the familiar operator structure $U=S(C\otimes I)$, but the same framework also admits space-inhomogeneous coins, scattering states on $\ell^\infty$, Hamiltonian realizations, CPTP and random-unitary extensions, Markov and CMV representations, and application-specific embeddings ranging from neutrino oscillations to spatial epidemic models (Shikano, 2013, Jayakody et al., 2021, Moqadam et al., 2015, Doliwa et al., 2023).

1. Canonical coined architecture

The canonical one-dimensional discrete-time QRW is defined on a tensor-product Hilbert space

$\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$

with position basis $\{|n\rangle:n\in\mathbb{Z}\}$ and a two-dimensional coin basis, written either as $\{|L\rangle,|R\rangle\}$, $\{|\uparrow\rangle,|\downarrow\rangle\}$, or $\{|{\bf H}\rangle,|{\bf T}\rangle\}$ depending on the paper. The conditional shift preserves the coin state while moving the walker left or right; in one common notation,

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$

and the one-step unitary is $U=W(I_p\otimes C)$ for a homogeneous coin $C$ (Shikano, 2013).

A frequently used general $\ell^\infty$0 coin, up to global phase, is

$\ell^\infty$1

which makes explicit how $\ell^\infty$2 controls mixing while $\ell^\infty$3 control interference phases. Within this parameterization, the Hadamard, Grover, and Fourier coins appear as particular parameter choices, and the position marginal is independent of $\ell^\infty$4 although it depends on $\ell^\infty$5 and $\ell^\infty$6 (Jayakody et al., 2021).

The same coined structure extends directly to higher dimensions. On the two-dimensional square lattice, the position space is $\ell^\infty$7, while the coin can be implemented with two qubits,

$\ell^\infty$8

with the four basis states associated to the four cardinal directions. A standard choice is $\ell^\infty$9, and the conditional shift acts by

$\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$0

so that again $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$1 defines one step (Manna et al., 6 Sep 2025).

Space-inhomogeneous coined walks preserve the same architecture but replace a global coin by local unitaries $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$2. A useful decomposition is

$\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$3

where $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$4 and $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$5 are the parts of the amplitude that move left and right. The update rule becomes

$\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$6

which is the same coined DTQW written in scattering form (Matsue et al., 2017).

2. Asymptotic transport, weak limits, and continuum connections

For homogeneous, time-independent coins, the defining large-scale signature of the standard one-dimensional DTQW is ballistic transport. If $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$7 denotes the measured position after $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$8 steps, the natural scaling is $\mathcal{H}=\mathcal{H}_p\otimes\mathcal{H}_c,$9, not $\{|n\rangle:n\in\mathbb{Z}\}$0, and the weak limit law is explicitly non-Gaussian. For coin

$\{|n\rangle:n\in\mathbb{Z}\}$1

Konno’s limit theorem states

$\{|n\rangle:n\in\mathbb{Z}\}$2

where the limiting density has compact support on $\{|n\rangle:n\in\mathbb{Z}\}$3, edge singularities, and an asymmetry term determined jointly by the coin and the initial coin state (Shikano, 2013).

This ballistic behavior contrasts sharply with the classical central-limit scaling $\{|n\rangle:n\in\mathbb{Z}\}$4. In two dimensions, the same distinction appears in both variance scaling and profile shape: coined QRWs on regular lattices obey $\{|n\rangle:n\in\mathbb{Z}\}$5, while the classical walk obeys $\{|n\rangle:n\in\mathbb{Z}\}$6. The spatial profile is likewise different: classical diffusion yields an approximately Gaussian hill-shaped distribution with maximum at the origin, whereas the 2D QRW exhibits suppressed return probability at the origin and a bowl-shaped, interference-structured profile with higher weight on a ring-like front (Manna et al., 6 Sep 2025).

The coin parameters act as transport controls. In the one-dimensional general-coin formulation, $\{|n\rangle:n\in\mathbb{Z}\}$7 regulates spreading versus confinement, $\{|n\rangle:n\in\mathbb{Z}\}$8 controls asymmetry, and $\{|n\rangle:n\in\mathbb{Z}\}$9 does not affect the position marginal. For $\{|L\rangle,|R\rangle\}$0 or $\{|L\rangle,|R\rangle\}$1, probability accumulates near the extreme positions; for $\{|L\rangle,|R\rangle\}$2, the walk can localize near the origin; and for $\{|L\rangle,|R\rangle\}$3, asymmetric distributions arise even from an unbiased initial coin state (Jayakody et al., 2021).

A second major asymptotic theme is that DTQW can interpolate to continuous-time quantum transport. With the final-time-dependent coin

$\{|L\rangle,|R\rangle\}$4

and rescaled time $\{|L\rangle,|R\rangle\}$5, the limiting distribution changes from Konno’s DTQW law to the arcsine-type law of a CTQW; with a different small-angle coin $\{|L\rangle,|R\rangle\}$6, the continuum limit reproduces the $\{|L\rangle,|R\rangle\}$7-dimensional Dirac equation. This places DTQW as a superordinate discrete-time transport model from which CTQW and Dirac dynamics emerge as limiting descriptions (Shikano, 2013).

3. Inhomogeneity, scattering states, and resonant tunneling

A particularly informative inhomogeneous QRW framework is the double-barrier walk on the line. There the background coin is

$\{|L\rangle,|R\rangle\}$8

while two defect sites $\{|L\rangle,|R\rangle\}$9 and $\{|\uparrow\rangle,|\downarrow\rangle\}$0 carry a barrier coin

$\{|\uparrow\rangle,|\downarrow\rangle\}$1

The free regions propagate deterministically with phase accumulation, and the two defects act as localized quantum scatterers (Matsue et al., 2017).

The physically relevant states in this setting are not $\{|\uparrow\rangle,|\downarrow\rangle\}$2 wave packets but stationary scattering states in $\{|\uparrow\rangle,|\downarrow\rangle\}$3. For incidence from the left, the asymptotic form is

$\{|\uparrow\rangle,|\downarrow\rangle\}$4

with reflection and transmission probabilities

$\{|\uparrow\rangle,|\downarrow\rangle\}$5

The explicit amplitudes are

$\{|\uparrow\rangle,|\downarrow\rangle\}$6

so the denominator plays the role of the usual scattering denominator built from multiple internal reflections (Matsue et al., 2017).

The resonance condition is exact. Assuming $\{|\uparrow\rangle,|\downarrow\rangle\}$7,

$\{|\uparrow\rangle,|\downarrow\rangle\}$8

Writing $\{|\uparrow\rangle,|\downarrow\rangle\}$9 and $\{|{\bf H}\rangle,|{\bf T}\rangle\}$0, this becomes

$\{|{\bf H}\rangle,|{\bf T}\rangle\}$1

Perfect transmission is therefore a phase-matching effect: multiple reflections between the two barriers interfere destructively on the left and constructively on the right. The authors explicitly contrast this with classical random walks, which have no phase and therefore no nontrivial reflectionless resonance. They also identify this phenomenon as a “third characteristic” of quantum walks, distinct from both ballistic spreading and localization (Matsue et al., 2017).

The same model admits a quantum-graph interpretation. Stationary Schrödinger scattering on a line with $\{|{\bf H}\rangle,|{\bf T}\rangle\}$2-potentials at the vertices is mapped to a space-inhomogeneous DTQW with local coin

$\{|{\bf H}\rangle,|{\bf T}\rangle\}$3

Under this correspondence, the double-barrier resonance condition in the continuum problem matches the algebraic DTQW condition above, and the coined/scattering formulations become manifestly equivalent (Matsue et al., 2017).

4. Hamiltonian, Markov, and spectral reformulations

Several works recast the stepwise coined walk into alternative but equivalent frameworks. One route uses continuous-time Hamiltonian evolution sampled stroboscopically. In the resonator–qubit realization, the walker is encoded in a coherent state of a resonator and the coin in a qubit, with always-on Hamiltonian

$\{|{\bf H}\rangle,|{\bf T}\rangle\}$4

The first term generates coin rotations, the second generates conditional phase-space shifts. A Trotter–Suzuki approximation relates $\{|{\bf H}\rangle,|{\bf T}\rangle\}$5 to repeated DTQW steps, and the mismatch between the exact Hamiltonian evolution and the ideal coined walk is quantified by the Hellinger distance between the corresponding walker probability distributions. For standard site-to-site walking the distance is larger but bounded; for $\{|{\bf H}\rangle,|{\bf T}\rangle\}$6, where the walker effectively moves between sites, it remains below $\{|{\bf H}\rangle,|{\bf T}\rangle\}$7 for $\{|{\bf H}\rangle,|{\bf T}\rangle\}$8 (Moqadam et al., 2015).

A second Hamiltonian reformulation starts from the discrete step operator and reconstructs an effective generator. In one dimension the total Hamiltonian is written as

$\{|{\bf H}\rangle,|{\bf T}\rangle\}$9

with translational part

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$0

and local coin part

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$1

In position space this becomes a Weyl Hamiltonian plus a Dirac comb potential,

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$2

and the resulting time-evolution operator agrees with the standard coined walk. On square and graphene lattices, the same construction yields higher-dimensional additive single-step walks and clarifies why multiplicative constructions correspond effectively to multi-step constrained dynamics (Sarkar et al., 2015).

A different reformulation embeds a Hadamard walk into a higher-dimensional classical process. The four-state internal Markov chain

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$3

is combined with the interference matrix

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$4

so that

$$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$5

At the Markov level there is no interference; interference is recovered only after the linear map $$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$6, which subtracts the populations of sign-labeled internal states and reproduces the quantum amplitudes of the ordinary two-state walk (Bar-Haim, 2020).

The most systematic spectral reformulation is the half-line quantization based on Karlin–McGregor theory, Szegedy’s unitary, and orthogonal polynomials on the unit circle. A birth–death chain with transition matrix $$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$7 is described by orthogonal polynomials $$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$8 and a spectral measure $$W=\sum_{n\in\mathbb{Z}}\Big(|n-1,L\rangle\langle n,L|+|n+1,R\rangle\langle n,R|\Big),$$9 on $U=W(I_p\otimes C)$0; its Szegedy quantization has discriminant equal to the associated Jacobi matrix, and the unitary evolution on the cyclic subspace is represented by a CMV matrix with Verblunsky coefficients $U=W(I_p\otimes C)$1. The classical probabilities and the Verblunsky coefficients are related by

$U=W(I_p\otimes C)$2

with the corresponding $U=W(I_p\otimes C)$3 given by the Geronimus relation. The measures are connected by the classical Szegő map,

$U=W(I_p\otimes C)$4

which makes the random-walk and QRW polynomial systems spectrally equivalent (Doliwa et al., 2023).

5. Disorder, randomization, and alternative meanings of “QRW”

The coined framework admits several distinct kinds of randomness. One analytically solvable model introduces a dichotomic random variable $U=W(I_p\otimes C)$5 into the $U=W(I_p\otimes C)$6 coin

$U=W(I_p\otimes C)$7

and averages over the two step operators through the CPTP map

$U=W(I_p\otimes C)$8

The resulting second moment is explicit, and the walk crosses over from the ballistic regime $U=W(I_p\otimes C)$9 to the diffusive regime $C$0 when

$C$1

The same averaged evolution can be dilated to a larger unitary walk, so the noisy process is also interpretable as a generalized QRW on an extended space (Ellinas et al., 2012).

A different type of dynamical disorder randomizes the coin at each time step while keeping the total evolution unitary along each realization. For a one-dimensional walk with random $C$2 coins, the reduced coin density matrix

$C$3

satisfies $C$4 asymptotically under broad randomness assumptions, so the coin–position entanglement entropy approaches its maximal value,

$C$5

independently of the initial state. Ordered walks do not share this universality: their asymptotic entanglement remains below one and depends strongly on the initial spin and position state (1305.4191).

Randomness can also be placed in the shift rather than the coin. In the discrete-time random-step quantum walk, the one-step shift is

$C$6

with $C$7 chosen randomly at each step from a fixed interval $C$8. The resulting single-run distributions are highly irregular. Deterministic fixed-length variants are also studied: in the unbiased case the left and right step sizes are equal and the peak positions obey the empirical rule “position of a peak $C$9,” while the standard deviation shows a sawtooth pattern; in the biased case unequal left and right step lengths generate asymmetric distributions and directional drift (Ahmad et al., 2020).

A terminological caution is necessary. In much of the literature “QRW” is simply another label for the standard coined DTQW, but one work defines QRW differently: the walker uses many unrelated coins, one per step, so there are no temporal correlations between steps. In one dimension the resulting position law is an asymmetric binomial distribution, coherence induces directional movement through effective transition probabilities, and for an initially decohered coin state the distribution is exactly the same as that of the classical random walk. This is a distinct framework, not a synonym for the standard single-coin DTQW (Chen et al., 2019).

6. Quantum simulation and application domains

The DTQW framework has been used as an explicit simulator of neutrino oscillations. A one-dimensional six-level walk is constructed from three two-level blocks,

$\ell^\infty$00

with

$\ell^\infty$01

The three mass sectors are then mixed by the PMNS matrix, so the flavor transition amplitudes take the standard oscillatory form

$\ell^\infty$02

The required internal space can be implemented as a single six-level system, a three-qubit system, or a qubit–qutrit system, and the model also permits direct study of spin–position entanglement during propagation (Mallick et al., 2016).

A more applied but explicitly qualified use appears in spatial epidemic modeling on a 2D square lattice. There the QRW is the standard 2D coined walk with $\ell^\infty$03, but the epidemiological layer uses the quantum circuit only to generate non-classical move statistics. Each infected site spawns a walker, susceptible sites are infected upon visits with probability $\ell^\infty$04, walkers persist for $\ell^\infty$05 steps, and the main observables are the single-run cluster size $\ell^\infty$06, its ensemble average $\ell^\infty$07, a heterogeneity parameter $\ell^\infty$08, and the empirical reproduction number $\ell^\infty$09. The simulations show low-$\ell^\infty$10 regimes in which QRW and classical random-walk SIR behave similarly, and higher-$\ell^\infty$11, higher-$\ell^\infty$12 regimes in which ballistic propagation and interference produce larger $\ell^\infty$13 and non-Gaussian spatial profiles. The authors explicitly describe this as a conceptual toy model rather than a literal quantum-mechanical model of biological transmission (Manna et al., 6 Sep 2025).

These examples show the range of the framework without changing its core structure. Whether the target phenomenon is ballistic transport, resonant tunneling, noisy entanglement generation, classical-walk quantization, neutrino flavor oscillation, or a quantum-inspired spreading process on a lattice, the common ingredients remain a finite internal space, a conditional shift on discrete position, and a unitary—or controlled random-unitary or CPTP—update rule that determines how interference, disorder, and geometry shape transport (Matsue et al., 2017, Moqadam et al., 2015, Doliwa et al., 2023).