1D Coined Quantum Walk

• 1D coined quantum walk is a discrete-time unitary evolution on a line where movement depends on a two-state coin and exhibits interference between amplitudes.
• It demonstrates ballistic spreading with oscillatory, multiscale probability distributions that sharply contrast classical diffusive behavior.
• The framework facilitates state engineering, localization studies, and advances in quantum computation through various coin generalizations and disorder effects.

A one-dimensional coined quantum walk is a discrete-time unitary evolution on a line in which a walker’s translation is conditioned on an internal coin degree of freedom. In its standard form, the model lives on a tensor-product Hilbert space of position and a two-dimensional coin, evolves by repeated coin rotations and conditional shifts, and differs from a classical random walk because amplitudes rather than probabilities interfere. The resulting dynamics are typically ballistic rather than diffusive, with highly structured long-time distributions rather than Gaussian smoothing. The model has a founding role in the study of quantum computation, and its one-dimensional version has become a prototype for classification theory, asymptotic analysis, localization studies, generalized coin constructions, and state-engineering protocols (Luo et al., 2015, Hantzko et al., 31 Jul 2025).

1. Standard formulation and elementary dynamics

The standard one-dimensional coined walk is defined on a composite Hilbert space $\mathcal H=\mathcal H_p\otimes\mathcal H_c$, with position basis $\{\ket{x}:x\in\mathbb Z\}$ and a two-state coin basis written in the literature as $\{\ket{\uparrow},\ket{\downarrow}\}$, $\{\ket{L},\ket{R}\}$, $\{\ket{0},\ket{1}\}$, or analogous labels. A general time step has the form “coin, then conditional shift,” often written as

$$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$

with

$$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$

In the Hadamard case, a common symmetric initial state is

$$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$

and after $N$ steps one may write

$$\ket{\psi_N}=\sum_{i=-N}^{N}\big(a_i\ket{\downarrow}+b_i\ket{\uparrow}\big)\ket{i}, \qquad P_i=a_i^2+b_i^2, \qquad \sum_i P_i=1.$$

A general $\{\ket{x}:x\in\mathbb Z\}$0 coin used in the line-walk literature is

$\{\ket{x}:x\in\mathbb Z\}$1

with Hadamard, Grover, and Fourier coins appearing as special parameter choices. In that parametrization, $\{\ket{x}:x\in\mathbb Z\}$2 controls spreading, $\{\ket{x}:x\in\mathbb Z\}$3 controls asymmetry, and $\{\ket{x}:x\in\mathbb Z\}$4 does not affect the final position probability distribution, even though it appears in intermediate amplitudes (Jayakody et al., 2021, Luo et al., 2015).

The key physical distinction from the classical random walk is interference. In the Hadamard walk, the coin operator can be decomposed as $\{\ket{x}:x\in\mathbb Z\}$5, so that the $\{\ket{x}:x\in\mathbb Z\}$6-step evolution may be viewed as $\{\ket{x}:x\in\mathbb Z\}$7: right- and left-moving amplitudes combine constructively or destructively across many paths. This produces linear spreading,

$\{\ket{x}:x\in\mathbb Z\}$8

whereas the classical one-dimensional random walk has a binomial law that becomes Gaussian with width $\{\ket{x}:x\in\mathbb Z\}$9. The same contrast appears in other analyses of the line walk: coherent step-to-step evolution generates ballistic profiles and strong coin-position entanglement after the first step, while classical or classically correlated models do not share the same transport law (Luo et al., 2015, Jayakody et al., 2021).

2. Long-time probability structure on the line

The long-time distribution of the standard one-dimensional coined walk is not a smooth curve. A detailed numerical study of walks up to $\{\ket{\uparrow},\ket{\downarrow}\}$0 steps showed oscillatory structure on all length scales, from lattice-scale alternation to slow modulations near the center. The distribution develops two dominant lobes with maxima near

$\{\ket{\uparrow},\ket{\downarrow}\}$1

and the effective support converges to the normalized interval

$\{\ket{\uparrow},\ket{\downarrow}\}$2

but without an abrupt cutoff. Instead, $\{\ket{\uparrow},\ket{\downarrow}\}$3 is a crossover point where the envelope changes rapidly and the oscillations disappear. The location of maximal probability satisfies $\{\ket{\uparrow},\ket{\downarrow}\}$4, the peak scales as

$\{\ket{\uparrow},\ket{\downarrow}\}$5

the full width at half maximum of the leading peak scales as $\{\ket{\uparrow},\ket{\downarrow}\}$6, the first 10 peaks near the center have width $\{\ket{\uparrow},\ket{\downarrow}\}$7, and the last 10 peaks near the edge have width $\{\ket{\uparrow},\ket{\downarrow}\}$8. Near the edge the extracted envelope behaves essentially like a square-root singularity in the distance from the edge; near the center it is approximately quadratic; beyond the edge the probability decays in an unconventional exponential or pseudo-Gaussian form rather than with classical Gaussian exponents. The discrete Fourier transform confirms multiscale structure: $\{\ket{\uparrow},\ket{\downarrow}\}$9, the number of Fourier peaks grows linearly with $\{\ket{L},\ket{R}\}$0, the envelope is algebraic near both $\{\ket{L},\ket{R}\}$1 and $\{\ket{L},\ket{R}\}$2, and beating near $\{\ket{L},\ket{R}\}$3 mirrors beating in real space (Luo et al., 2015).

This asymptotic picture rules out a common oversimplification: the one-dimensional coined walk is not merely a pair of smooth outward-moving Gaussians. Its ballistic character coexists with persistent oscillations, algebraic envelopes, edge concentration, and a crossover to rapid decay outside the main support. On the infinite lattice, another asymptotic result is the explicit return law

$\{\ket{L},\ket{R}\}$4

which yields $\{\ket{L},\ket{R}\}$5 and is independent of the initial coin state; for the Hadamard walk $\{\ket{L},\ket{R}\}$6 this becomes $\{\ket{L},\ket{R}\}$7 for even $\{\ket{L},\ket{R}\}$8. On finite one-dimensional lattices, the time-averaged distribution can be nonuniform and parity-sensitive, and for sufficiently large thresholds the average mixing time satisfies $\{\ket{L},\ket{R}\}$9 (Xu, 2010).

3. Classification, equivalence, and analytic formalisms

A major structural result is that many apparently different simple coined walks on the line are unitarily equivalent. For translation-invariant one-dimensional walks with an arbitrary $\{\ket{0},\ket{1}\}$0 coin, local unitary transformations on coin and position reduce the full three-parameter Euler-angle family to a one-parameter canonical family $\{\ket{0},\ket{1}\}$1. In that sense, all nonequivalent simple walks are classified by the middle Euler angle $\{\ket{0},\ket{1}\}$2, and the remaining two coin parameters can be absorbed into basis changes and quasi-momentum shifts. The same equivalence framework shows that electric quantum walks are equivalent to walks with a time-dependent coin toss operator (Goyal et al., 2013).

A more recent distributional classification sharpens this picture by focusing directly on the induced spatial probability laws. For non-trivial coins, there are exactly two initial coin states, up to global phase, that generate symmetric walks, characterized by

$\{\ket{0},\ket{1}\}$3

Modulo distributional equivalence, symmetric walks admit a bijective parametrization by $\{\ket{0},\ket{1}\}$4; all coined walks admit a surjective parametrization by $\{\ket{0},\ket{1}\}$5; and limiting distributions also admit a bijective parametrization on a reduced parameter domain. In the same framework, corrected closed-form expressions for walk amplitudes were derived, and quadratic variance growth was identified as the hallmark of non-trivial coined quantum walks, in contrast with the linear variance of correlated classical random walks except at extremal correlation points (Hantzko et al., 31 Jul 2025).

Several complementary analytic techniques have been developed for the one-dimensional coined walk. The path-integral approach encodes trajectories as strings of forward and backward steps, classifies them by switch structure, and yields closed-form amplitudes for arbitrary step number. The effective-Hamiltonian approach reconstructs the discrete step operator from

$\{\ket{0},\ket{1}\}$6

or in position space,

$\{\ket{0},\ket{1}\}$7

interpreting the walk as a Weyl Hamiltonian plus a Dirac-comb potential. For inhomogeneous line walks, the generating-function and CGMV methods link continued fractions, Schur recursion, and the spectral measure of the associated CMV matrix, making detailed asymptotics accessible (Joshi et al., 2018, Sarkar et al., 2015, Konno et al., 2013).

4. Inhomogeneity, disorder, decoherence, and localization

Localization in one-dimensional coined quantum walks is not a single phenomenon. In a half-line model with site-dependent coins

$\{\ket{0},\ket{1}\}$8

the coin approaches the identity as $\{\ket{0},\ket{1}\}$9, yet the walk exhibits localization near the origin with a power-law tail,

$$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$0

which decays like $$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$1, together with “bottom localization” at the ballistic front in the dual distribution. The weak limit takes the form

$$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$2

with no continuous part. For spatially random i.i.d. $$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$3 coins on $$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$4, sufficient transfer-matrix conditions imply dynamical localization and strong spectral localization: transition amplitudes decay rapidly with distance uniformly in time, and the spectrum is almost surely pure point. This includes coin distributions that are Haar-continuous and certain two-point distributions whose support contains the Hadamard coin (Konno et al., 2013, Ahlbrecht et al., 2011).

Other localization mechanisms arise from disorder and topology. In homogeneous walks, coherence in the full Hilbert space and especially in position space grows with ballistic spreading. Under spatial disorder, the distribution becomes strongly localized near the origin, the full-space coherence saturates at a very small value, and coin-space coherence oscillates in time; temporal disorder produces weaker localization; and in split-step walks

$$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$5

localized edge states appear at interfaces between distinct topological phases. In that setting, reduced position-space coherence, oscillatory coin-space coherence, and concentration of the interference indicator $$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$6 diagnose localization more precisely than spread alone (Singh et al., 2017).

Disorder does not always classicalize the walk in the same way. On one-dimensional percolation lattices, random missing edges act like decoherence: dynamic gaps drive a crossover from ballistic spreading to a classical-like binomial profile with crossover time

$$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$7

and long-time width scaling $$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$8; static gaps can trap the walker even more strongly, although tunneling across broken links delays the loss of quantum behavior. By contrast, a distinct “tunneling decoherence” model in which the walker is randomly shifted one extra site left or right after each coherent step preserves quadratic variance growth and nonzero coin-position entanglement. In that case

$$\mathcal U=\mathcal S(\mathcal C\otimes \mathcal I),$$9

and the decoherent position law is an explicit binomial mixture of shifted coherent distributions, so the profile is smoothed rather than classicalized. A further caution is that not every coin modification localizes the walk: with two or more time-independent unitary coin rotations before the shift, the line walk remains translationally invariant and ballistic, with $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$0 and no localization (Leung et al., 2010, Annabestani et al., 2010, Xue et al., 2014).

5. Generalized coin spaces and nonstandard one-dimensional walks

The standard two-state line walk admits several systematic generalizations. On weighted graphs, a coined walk with a single self-loop of real weight $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$1 at each site yields a three-state line walk with basis $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$2, local coin

$$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$3

and moving shift. This generalized lackadaisical walk is exactly equivalent to a continuous deformation of the three-state Grover walk under

$$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$4

with ballistic peak velocities

$$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$5

Increasing $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$6 therefore gives faster ballistic dispersion in the moving-shift realization (Wong, 2017).

Higher-dimensional internal coins on the line generate qualitatively new transport regimes. A walk with two entangled qubits as coin has four internal states $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$7 and a shift in which $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$8 moves right, $$\mathcal S = \ket{{\bf H}}\bra{{\bf H}}\otimes \sum_x \ket{x+1}\bra{x} + \ket{{\bf T}}\bra{{\bf T}}\otimes \sum_x \ket{x-1}\bra{x}.$$9 moves left, and $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$0 stay put. Depending on the initial coin entanglement and whether the two sub-coins are identical or not, the walk can display Gaussian or classical-like spreading, self-trapping, perfect transfer with highest velocity, a four-peaks regime, and strong variation in symmetry and entropy. In the separable limit the entropy is minimized; for maximally entangled initial states it is maximized. Position-dependent coins supply another control mechanism: when the coin angle depends on the site as $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$1, the same one-dimensional walk can become localized, periodic, bounded classical-like, semi-classical-like, or quantum-like, and in most cases the position-space entropy is smaller than for the corresponding position-independent coin. With two identical entangled position-dependent coins, complete localization can occur for certain Bell states because part of the shift acts trivially on $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$2 and $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$3 (Panahiyan et al., 2018, Ahmad et al., 2019).

Memory can also be folded into the internal state. In the two-step-memory Hadamard walk, the local state stores the previous two directions together with a coin bit, giving an effective eight-state walk on the line. Its probability distribution differs from both the standard Hadamard walk and the one-step-memory walk: it does not exhibit the strong localization of the one-step-memory model, but instead shows a double-peaked profile more reminiscent of the standard walk, with peaks roughly at

$$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$4

rather than at $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$5. Interference also appears later than in the usual Hadamard walk, reflecting the more complex path combinatorics (Zhou et al., 2019).

6. State engineering, platform independence, and conceptual boundaries

Beyond transport theory, the one-dimensional coined walk has been turned into a state-engineering primitive. With site-independent but step-dependent coins, followed by a final projection of the coin, coined quantum walks on a line can be used to prepare arbitrary superpositions of the walker’s sites. The reachable joint walker-coin states after $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$6 steps are characterized by boundary conditions

$$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$7

together with orthogonality-type constraints

$$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$8

Within that reachable set, essentially arbitrary target walker states over $$\ket{\psi_0}=\frac{1}{\sqrt 2}\big(\ket{\uparrow}+i\ket{\downarrow}\big)\otimes\ket{0},$$9 sites can be obtained probabilistically by choosing a suitable coin sequence and projecting the coin at the end. The required coin parameters can be reconstructed efficiently by backward iteration, and a concrete linear-optics implementation has been proposed in which the walker is encoded in photonic orbital angular momentum, the coin in polarization, the conditional shift by a $N$0-plate, and the final projection by standard polarization optics (Innocenti et al., 2017).

A useful conceptual boundary is supplied by comparison with models that do not reuse a single coherent coin across steps. In an ensemble-based “quantum random walk” with a different unrelated Hadamard coin at each step, the one-dimensional position law is an asymmetric binomial distribution,

$N$1

with

$N$2

There is directionality induced by the initial coin coherence, but no ballistic spreading because the steps are uncorrelated. This comparison makes precise what is distinctive about the coined quantum walk proper on the line: ballistic transport, non-Gaussian limit laws, and multiscale interference are consequences of coherent step-to-step memory carried by the repeated unitary evolution of the same coin-position system (Chen et al., 2019).