A quantum phase reference must repay both its size and the precision that keeps it coherent. For complementary binomial signals with cyclic phase noise and an unknown common phase, we derive exact resource laws for binary discrimination. A contracting variational operator gives the ideal error exponent over every independent reference state and joint measurement. An entropy crossing identifies the sharp two-charge recovery threshold, p_G = 0.197884450826..., with a certified rational enclosure. Above this threshold, two charges attain the entire global ideal partial-recovery curve; at p = 0.3 and equal phase-use costs, the optimal charged exponent is 67.57% above counting. A finite polynomial comparison proves balanced-population exponent optimality for every bias and visibility. Under intermittent complete phase randomization, a tapered multi-charge profile attains a matching converse at a finite linear span and can outperform every two-charge design at the same noise rate and resource prices. Within the two-charge architecture, charging control precision yields an exact matching principle for monotone costs and three distinct regimes: bulk effort sublinear in log-precision preserves the ideal price law, linear effort renormalizes it, and superlinear effort makes counting optimal. Separately charged, diverging local precision costs also remove the advantage in an explicit independent-dephasing implementation. Finite converses, constructive receivers and reproducible certificates connect the attainable quantum gain to the resources required to realize it. This record archives the same scientific work as the manuscript titled Finite phase references: sharp resource and precision limits for quantum discrimination, under an alternative repository title. It is an author preprint; journal peer review has not been completed. The main manuscript, two proof supplements, LaTeX sources and computational materials are included. Version P19_9994. Access to all files is restricted, with no automatic release date. Bibliographic metadata are public.
A finite quantum phase reference can restore information lost by counting, but occupied support and phase noise impose different limits. For complementary binomial signals with sublinear cyclic order, we establish these limits and compare designs under a declared resource budget. At critical uniform-dephasing quality, bounded support attains the full coherent error exponent precisely above the bias threshold p* = (1 − 1/√2)/2. At and below it, every fixed support bound leaves a positive gap, while arbitrarily slow support growth suffices. Near the threshold from above, a geometric construction needs only logarithmically many charges in the inverse distance to the threshold. Four charges are necessary and sufficient for 3/10 ≤ p < 1/2. Gaussian recovery also depends on how rapidly the cyclic order grows. Widths decaying as a power faster than the inverse square root of signal size recover the ideal exponent at any fixed span ratio when cyclic order grows slower than the logarithm of that size. When cyclic order grows faster than this logarithm, every fixed polynomial precision loses its exponential advantage. The ideal optimized rate survives logarithmic precision costs and a declared polynomial charge for the largest protected subspace dimension in an occupied block. At a common budget, an explicit four-charge reference outperforms the fully optimized two-charge class. Explicit constructions and converses over all reference states distinguish occupied support, register span and noise as separate resources. Author preprint, version V2.7. The main manuscript, complete proofs, LaTeX sources, computational resource and alternative figure/table descriptions are included. The scientific content and author declarations are preserved; journal-specific packaging and the editorial cover letter are excluded. All files are embargoed until 15 February 2027, when they will automatically become publicly accessible. Bibliographic metadata are public.
We quantify the extra Bayes risk of locating a quantum change point and estimating its amplitude with one measurement. The joint and marginal optimizations use the full record and the same joint prior, with both parameters unknown. For product records on a known pure state curve, position has clipped absolute loss in resolution units, and amplitude offset has squared loss in fixed calibrated units. Ideal steps exhibit certified conflict at finite signal strength, including in phase records of sixteen qubits, and recover both informative marginal optima at strong signal. For finite offset prior width, a retained quantum boundary region serves position, and the remaining suffix serves amplitude. Physical recovery holds when strength grows more slowly than the square root of record length. Known smooth rises also admit leading recovery in the coherent experiment as strength grows and width shrinks. For the quintic profile, we establish two residual cost scaling laws. The residual cost is
Θ(s−2) as
s→∞. Taking the strong signal limit at fixed
w yields a limiting cost of
Θ(w3) as
w→0. In the resolved local model, quadratic position loss admits a common optimum within the covariant class. These results link onset geometry and decision loss to Bayesian measurement compatibility in quantum change point estimation.
Purity does not change under a phase rotation, but estimating it from successive optical pulses can still require relative-phase information. This article determines the best possible precision when an unknown single-mode source is independently prepared, undergoes photon loss, and is then rotated by a hidden correlated phase trajectory. A uniform precision law. The normal source state satisfies
Tr(ρN)≤E and
Tr(ρNs)≤M, with fixed
E,M>0 and
s≥1. The phase starts uniformly on the circle, independently of the entire sequence of independent increments, whose known Fourier spectrum is
exp(−γ∣q∣β), with
γ>0 and
0<β≤2. The target is the purity
Pη=Tr[Lη(ρ)2] after pure loss, not the purity of the phase-averaged marginal. For
m source preparations, set
p=1−η,
Lm=[log(m+e)]1/β and
qm=⌈[logm/(2γ)]1/β⌉. The minimax expected absolute error obeys
Rm(Pη)≍Fm(p),Fm(p)=Lm−sηqm+Lm−2s(1+pLm)−2s. The comparison constants are uniform for
η∈[1/2,1], including transmissivities approaching one as the budget grows. An ideal adjacent-pair interference measurement, independent of the transmissivity, attains the order. The converse allows arbitrary joint measurements on the complete correlated quantum history. The same law holds for the second Rényi entropy
−logPη. What loss changes. Vacuum–high-photon coherence and coherence between broad high-photon packets give two different hard pairs of legal sources. Without loss the rate is
(logm)−s/β; at fixed nonzero loss it is
(logm)−4s/β. Each loss level has its own output target. The faster rate therefore does not mean improved recovery of the original input purity. Both constructions retain their actual moment budgets, the same source within each experiment and complete photon tails. Mean cost, stopping and unbiased readings. Allowing arbitrary quantum memory, feedback and an observable stopping time leaves the order unchanged when only the worst-case mean number of requested raw source slots is bounded by
B:
RBstop(Pη)≍F⌊B⌋(1−η). Every requested slot is charged, including discarded slots and failed outcomes. An exactly unbiased purity estimator attains the same absolute-error order with finite mean cost and almost-sure termination, using rare, unbounded signed readings. Under the specified independent common-phase reset and auxiliary-reset permissions for repeated blocks, exact unbiasedness and uniformly finite mean cost rule out uniformly finite variance. This is a uniform obstruction over sources, not a claim of infinite variance at every source; exact unbiasedness is not asserted for the Rényi entropy. The analysis builds on swap/parity measurements, drift diagnostics, quadratic-functional deconvolution and randomized telescoping. Its contribution is the matching quantum-history law, its uniform loss dependence and its return to the actual mean preparation budget. The integrated 14-page PDF contains two figure groups, complete proofs and references. The figures illustrate asymptotic rate functions, not experimental data or calibrated sample requirements; detector, phase-control, storage and elapsed-time costs are separate from the source-slot optimum.
Financial crashes, cascading failures in infrastructure, and critical errors in AI systems are frequently triggered by events that occur with extremely small probability. Efficiently discovering and sampling events with probability below a threshold is therefore of critical interest. Yet this task is highly non-trivial using existing classical or quantum methods. Being rare, such events require an immense sampling overhead to collect sufficient data samples. Moreover, because the rare events are not known in advance, they cannot be flagged for amplification using standard techniques. Here, we introduce a quantum algorithm for rare-event discovery and sampling without first learning which events are rare. The algorithm achieves the optimal quantum scaling with the rarity threshold. We further demonstrate that this can achieve a quadratic speedup for heavy-tailed systems whose tail has nonvanishing total mass, and translates into a robust polynomial speedup for stationary stochastic processes, with the exponent determined by its entropy-rate structure.
Recovery from a coherent rotation with an unobserved angle depends on both the encoding and what calibration reveals about the noise. We determine their combined limit for a known finite-dimensional generator and one final recovery. Two polynomial projections organize the answer. A physical Krylov chain separates correctable error amplitudes from their first state-dependent distortion; projection onto lower-angle powers identifies the part that a coherent decoder can cancel. Under explicit rank and moment conditions, their combination gives the exact leading optimal entanglement infidelity at weak noise and a recovery attaining it for any finite logical dimension. The optimal worst-input error lies between one and
d times that infidelity at leading order, for logical dimension
d. When the angle distribution is unknown, we optimize one decoder for every distribution compatible with the calibration. When spectral and calibration frequencies are integer multiples of a common frequency, moment positivity turns this into an exact finite optimization. Small measured means can still allow rare finite rotations. If the sole calibration datum is
βr=E[sin2r(θ/2)], with
θ the wrapped rotation phase at that common frequency, the optimal worst-prior infidelity is
O(βr) exactly when generator powers through degree
r−1 are correctable. When the first uncorrectable power has degree
m<r, the optimal loss scales as
βrm/r, with a sharp coefficient fixed by a universal moment problem. In a loss-free four-photon memory, population measurements at two calibration durations retain the fully informed quartic error coefficient for a fixed symmetric narrow-noise family with positive variance and finite fourth moment; a single duration permits loss quadratic in the noise width.