One Quantum Measurement Sends Out Many Different Fronts

A local measurement is not quite like giving a chain a kick. It produces a definite outcome, and that information can immediately change what someone who knows the outcome predicts at distant sites, because the ground state was already correlated. At the same time, the measurement removes some of the measured site's correlations with the rest of the chain. The later dynamics carries both effects outward.

Nicholas Musat, Stephan Haas and Rishabh Jha solve this problem for quadratic, or free-fermion, models: the Kitaev chain, a spin chain called the cluster chain in a transverse field, and a Kitaev chain with power-law couplings. They calculate the root-mean-square change in a chosen observable across measurement outcomes. This measure retains the outcome-dependent signal even when averaging the outcomes would cancel it. Their exact treatment follows how pairs of quasiparticle waves contribute to each observable.

The key distinction is between a front's kinematics and its strength. The fastest quasiparticles set where a front travels; the way the energy band bends near their maximum speed sets how quickly it widens. But the measured operator must create those fast excitations, and the observed operator must register them. If the latter response vanishes for the fastest modes, a weaker contribution from nearby, slower modes can control the front instead. Thus two observables in the same system can share a location and width but have different heights, decay powers and scaled profiles.

Airy functions are important in particular cases, not a universal answer for every front in the paper. For a smooth fastest-velocity maximum with cubic leading band curvature, the width grows as the cube root of time. In the nearest-neighbor Kitaev chain at equal hopping and pairing, and in several stated cluster-chain protocols, the leading profiles include Ai(s)², Q(s), or B(s), depending on whether the leading observable weight survives or cancels. Here Q and B are specific integrals of the Airy function, not synonyms for Ai². At the Kitaev critical point, both density and bond-energy responses instead have a Q profile and decay as t⁻¹. For the critical cluster chain, measuring the symmetry-odd spin Z rather than the symmetry-even X changes the stabilizer front's decay from t⁻¹ to t⁻²⁄³ at the same Hamiltonian; its profile is a separate function built from Q(u)[1−Q(u)]. These are distinct analyzed protocols, not one profile rule.

The paper also shows the model dependence within the Kitaev example. For nearest-neighbor couplings with hopping equal to pairing, the density and bond-energy front powers exchange between topological and trivial phases: in the topological phase the density is blind to the fastest modes and its front decays as t⁻⁴⁄³, while the bond-energy front decays as t⁻²⁄³; in the trivial phase those powers swap. At the critical point both decay as t⁻¹. Changing the band by making hopping unequal to pairing restores a leading density contribution after a crossover, so the special cancellation is not generally robust to that change.

Long-range couplings provide another counterexample to a single width or profile. For the paper's power-law model, different ranges of the exponent α and different gap-closing momenta produce different spreading laws. Some fronts have widths proportional to t¹⁄(α−1), or approximately t log t at α = 2, rather than t¹⁄³. The associated scaled forms can involve a fractional-dispersion kernel or a Fresnel integral; the critical closing at k = π for α above about 2.283 is a distinct case with a Q profile. The long-range section also identifies cases where an algebraic tail contributes alongside the moving front.

These are exact results for the specified quadratic models and measurement protocols, not a theorem that fixes fronts in arbitrary quantum materials. The authors identify finite temperature, interactions, repeated measurements and periodic driving as open directions. They discuss possible quantum-simulator ingredients, but do not report an experimental demonstration. The work is an arXiv preprint, not evidence of peer review.