betaFeedback

WorksPhysics and Astronomy

Finding Low-Energy States of the Bernasconi Model via Simulated Transverse-Field Quantum Dynamics

Abstract

The Low-Autocorrelation Binary Sequence (LABS) problem, the ground-state problem of the Bernasconi model, has a rugged landscape with deep, isolated local minima. We study real-time transverse-field dynamics initialized in these minima as a means of reaching lower-energy configurations. Using GPU statevector simulation for N = 20-30 spins, we compare the escape probability per verified quantum sample with the success rate per energy evaluation of steepest-descent kicks, random restarts and tabu search. Using the same field and sampling window across all system sizes, we find that, relative to the best classical baseline, the sampler achieves a factor of 11.9-35.2 for the deepest traps and 13.4-27.2 for the next energy tier. These ratios describe a constant-factor improvement in sample efficiency over the sizes studied. At the adopted field, broad Hamming-space spreading coexists with a strong preference for low-energy states. This regime follows a sharp increase in displacement as the field is raised. To connect this behaviour to the landscape, a short-time expansion relates the onset of escape to target distance and intermediate energies. Bounds on isotropic spin flips and on the free transverse mixer quantify the role of the Bernasconi potential in the observed transfer. At N = 28, the ground-state probability at t = 30 exceeds a canonical cost prediction fitted to the same mean energy by a factor of about 1.9. The cost of generating each quantum sample is assessed separately through a hardware estimate, which places a direct circuit implementation two to three orders of magnitude beyond present processor capabilities.

Introduction

We study whether coherent spin dynamics can improve the escape from a deep classical minimum. The Bernasconi model provides a concrete setting in which to connect this question to optimisation of binary sequences and to earlier work on quantum search.

Binary sequences with small aperiodic autocorrelation are handy in radar ranging, pulse compression and spread-spectrum communication. Their design leads to the Low-Autocorrelation Binary Sequence (LABS) problem, whose quality measure is the merit factor [1, 2, 3]. Constructive families and correlation bounds have clarified the attainable merit factors [4, 5]. Even with this progress, finding the optimal sequence at a given length remains a demanding discrete search problem.

Bernasconi formulated this search as the ground-state problem of a long-range Ising model [6]. Its deterministic interactions produce a rugged energy landscape with many local minima, making it a practical setting for studying glassy behaviour without quenched disorder [7, 8, 9]. The distribution and structure of these minima help explain the difficulty of local search [10]. To assess search methods on this landscape, exhaustive and branch-and-bound calculations provide exact ground states and reliable benchmarks at finite sequence lengths, although their computational cost grows exponentially with length [11, 12].

Classical methods address this difficulty through structured local search, restarts and combinations of global and local moves. For example, relaxation-based search, memetic algorithms and self-avoiding walks have produced strong LABS solvers [13, 14, 15, 16]. The inclusion of LABS in the Quantum Optimization Benchmarking Library also makes it a common setting for comparing quantum and classical optimization [17]. Within this broader search problem, escape from a local minimum is a basic step shared by many heuristics. Iterated local search and basin hopping perturb the current configuration and then descend again, while multistart methods explore fresh basins [18, 19, 20]. The difficulty of this step becomes clear from the scarcity of other beneficial proposals. For the deepest traps studied here, the fraction of lower-energy configurations falls from 7.6×10−67.6 \times10^{-6} at N=20N = 20 to 1.5×10−81.5 \times10^{-8} at N=30N = 30 spins. This scarcity makes the distribution of escape proposals central to the performance of local search.

Thermal and quantum fluctuations offer different ways to explore such landscapes. Simulated annealing uses thermal acceptance of uphill moves [21], while quantum annealing introduces a kinetic term that couples classical configurations [22, 23]. Adiabatic algorithms, quantum annealing studies and finite-range tunnelling benchmarks establish the role of this coupling in optimization dynamics [24, 25, 26]. The corresponding algorithms and implementations are discussed in Refs. [27, 28]. A recurring physical issue in these studies is localization in configuration space, including its effect on adiabatic optimization [29].

With the Hamiltonian held fixed, the same coupling defines a continuous-time quantum walk on the graph of spin configurations. This viewpoint brings the escape problem into the framework of coherent transport and search [30, 31, 32], with the hypercube supplying the natural graph for independent spin flips [33]. In addition to that, quantum-walk studies of spin-glass optimization connect the competition between the cost Hamiltonian and the driver to the probability of finding low-energy states [34, 35].

Our protocol is closely related to population transfer, in which evolution starts from a low-energy classical configuration and redistributes amplitude among configurations of comparable energy [36, 37, 38]. Work on the quantum random energy model relates this process to localization, mobility edges and extended states [39, 40, 41]. Numerical studies further examine how the field and evolution time determine the transfer [42]. The choice of a convenient initial classical state also connects this approach to quantum local search and reverse annealing [43, 44]. Quantum-generated proposals have been studied in Markov chain Monte Carlo [45], and classical warm starts have been incorporated into variational quantum optimization [46].

For LABS itself, the quantum approximate optimization algorithm (QAOA) [47, 48] has been compared with strong classical solvers, with evidence of a scaling advantage reported for a fixed-parameter QAOA-based approach [49]. Along a different route, a recent variational study uses Pauli correlation encoding to represent larger LABS instances with few qubits [50]. These developments motivate a complementary question about a single optimization move. Given a deep classical minimum, how effectively can transverse-field dynamics produce a configuration of lower energy?

We address this question directly on the deterministic Bernasconi landscape. The initial states exhaust the symmetry-distinct minima in the two lowest excited energy tiers. Full-statevector simulations for N∈{20,22,24,26,28,30}N \in\{20,22,24,26,28,30\} compare the resulting sample yield with classical escape methods through their successes per energy evaluation. To understand the resulting escape probabilities, we develop a short-time expansion, geometric bounds on isotropic kicks and the free mixer, constraints from energy conservation, and a comparison with a canonical cost distribution fixed by the measured mean energy. These results connect the numerical observations to the distance, number and energies of the accessible configurations.

Model and escape problem

Our first step is to make the escape problem precise. We introduce the classical energy and the transverse-field dynamics, and then identify the initial traps and the lower-energy configurations that count as successful escape. This gives a common setting for the quantum and classical searches.

Energy and spin dynamics

We represent a binary sequence by a configuration of NN spins, s=(s1,…,sN)s=(s_1,\ldots,s_N), with si∈{−1,+1}s_i \in\{-1,+1\}. Its overlap with a shifted copy is measured by the aperiodic autocorrelation. For a lag k∈{1,…,N−1}k \in\{1,\ldots,N-1\}, this is

Ck(s)=∑i=1N−ksisi+k.(1)C_k(s)=\sum_{i=1}^{N-k}s_i s_{i+k}. \tag*{(1)}

To favour sequences with small correlations at all nonzero lags, we combine these overlaps into the classical Bernasconi energy,

E(s)=∑k=1N−1Ck(s)2.(2)E(s)=\sum_{k=1}^{N-1}C_k(s)^2. \tag*{(2)}

The merit factor F(s)=N2/[2E(s)]F(s)=N^2/[2E(s)] measures sequence quality, so maximizing F(s)F(s) is equivalent to minimizing E(s)E(s) [2, 3]. Expanding Eq. (2) gives the long-range spin interactions [8, 9, 12],

E(s)=N(N−1)2+2∑k=1N−1∑1≤i<j≤N−ksisi+ksjsj+k.(3)E(s)=\frac{N(N-1)}{2}+2\sum_{k=1}^{N-1}\sum_{1\le i<j\le N-k}s_i s_{i+k}s_j s_{j+k}. \tag*{(3)}

The energy is invariant under global spin inversion (si↦−si)(s_i \mapsto-s_i), sequence reversal (si↦sN−i+1)(s_i \mapsto s_{N-i+1}) and alternating sign reversal (si↦(−1)isi)(s_i \mapsto(-1)^i s_i). These operations generate an order-8 symmetry group [12]. We therefore choose one canonical representative from each orbit when identifying distinct initial traps.

Having specified the energy landscape, we introduce dynamics between its configurations by replacing each spin with a Pauli-ZZ operator Z^i\hat{Z}_i. The resulting diagonal cost Hamiltonian satisfies H^B∣s⟩=E(s)∣s⟩\hat{H}_B|s\rangle=E(s)|s\rangle, where ∣s⟩|s\rangle is the computational-basis state representing the string ss. During evolution, this term assigns an energy-dependent phase to each configuration, while a transverse field of strength gg produces the transitions between them. The full Hamiltonian is therefore

H^(g)=1NH^B−g∑i=1NX^i,(4)\hat{H}(g)=\frac{1}{N}\hat{H}_B-g\sum_{i=1}^{N}\hat{X}_i, \tag*{(4)}

where X^i\hat{X}_i is the Pauli-XX operator acting on spin ii, and the factor 1/N1/N ensures that both terms scale extensively. The transverse operator ∑iX^i\sum_i\hat{X}_i couples states differing by a single spin flip, allowing probability amplitude to move across the Boolean hypercube.

Let JJ denote the physical energy scale multiplying H^B/N\hat{H}_B/N. We express Hamiltonians in units of JJ and measure evolution time tt in units of ℏ/J\hbar/J, where ℏ\hbar is the reduced Planck constant. All quoted fields and times are numerical values in these units, and we set J=ℏ=1J=\hbar=1 throughout.

Traps and lower-energy states

The escape problem begins at a configuration that resists a local energy-lowering move. We call s0s_0 a local minimum, or trap, when every single spin flip increases or preserves its energy,

E(s′)≥E(s0)∀s′ such that dH(s′,s0)=1.(5)E(s')\ge E(s_0)\quad\forall s'\text{ such that }d_H(s',s_0)=1. \tag*{(5)}

where dH(sa,sb)=∑i=1N1−sa,isb,i2d_H(s_a,s_b)=\sum_{i=1}^{N}\frac{1-s_{a,i}s_{b,i}}{2} is the Hamming distance.

To quantify the difficulty of escaping a trap s0s_0 of energy E0=E(s0)E_0 = E(s_0), we define the minimal barrier distance dmin⁡d_{\min} as the minimum Hamming distance to any configuration of strictly lower energy,

dmin⁡(s0)=min⁡s{dH(s,s0)∣E(s)<E0}.(6)d_{\min}(s_0) = \min_{s}\{d_H(s,s_0) \mid E(s) < E_0\}. \tag*{(6)}

Distance alone gives only part of the target geometry. We must also count how many lower-energy configurations lie at each distance. We denote the set of such configurations by S<={s:E(s)<E0}S_{<} = \{s : E(s) < E_0\}. For an integer distance d∈{0,…,N}d \in\{0,\ldots,N\}, let MdM_d be the number of configurations in this set at distance dd from s0s_0, and let MM be their total number. Using vertical bars to denote the number of elements in a set,

Md=∣{s∈S<:dH(s,s0)=d}∣,M=∑d=0NMd.(7)M_d = \left|\{s \in S_{<} : d_H(s,s_0) = d\}\right|,\qquad M = \sum_{d=0}^{N} M_d. \tag*{(7)}

Uniform sampling of all configurations therefore has escape probability Pblind=M/2NP_{\mathrm{blind}} = M/2^N. These counts include every bitstring, including symmetry partners, whereas the seed counts in Table 1 refer to symmetry-distinct orbits.

NNE∗E^*TierE0E_0Orbitsdmin⁡d_{\min} rangeSD hit fraction
2026Level 13446–81.03%
Level 23875–72.13%
2239Level 147294–93.10%
Level 2551012–69.79%
2436Level 144180.05%
Level 248190.07%
2645Level 15366–90.30%
Level 261204–90.69%
2850Level 1581100.05%
Level 2622110.03%
3059Level 167410–130.05%
Level 275194–110.19%

Table 1. Classification of exhaustive local minima tiers evaluated across system sizes N∈{20,22,24,26,28,30}N \in\{20,22,24,26,28,30\}. For each NN, the exact global optimum E∗E^* [12] is shown. Level 1 comprises all symmetry-distinct local minima at the lowest energy E0∗E_0^* strictly above E∗E^*. Level 2 comprises all orbits at the next energy level above E0∗E_0^*. SD hit fraction is the fraction of 10510^5 random-start steepest descents that end in any orbit of that tier.

A large dmin⁡d_{\min} places the nearest successful configurations many spin flips away. Within the Hamming shell at distance dd, which contains (Nd)\binom{N}{d} strings, the target fraction is Md/(Nd)M_d/\binom{N}{d}, so escape must overcome both distance and scarcity.

To include the deepest traps systematically, we enumerate all one-flip local minima from the full table of 2N2^N energies and group them by energy after symmetry reduction. Let E∗E^* denote the global minimum energy and E0∗E_0^* the lowest local-minimum energy strictly above it. We call the symmetry-distinct minima at E0∗E_0^* Level 1, while Level 2 consists of those at the next local-minimum energy above E0∗E_0^*. This exhaustive selection matters because random-start descent rarely reaches some of these low-lying basins. For example, it reaches the tiers at E=44E = 44 for N=24N = 24 and E=53E = 53 for N=26N = 26 in only 0.05% and 0.30% of runs, respectively.

The benchmark covers N∈{20,22,24,26,28,30}N \in\{20,22,24,26,28,30\}, from 220≈1.05×1062^{20} \approx1.05 \times10^6 to 230≈1.07×1092^{30} \approx1.07 \times10^9 configurations, where the lower-energy target set becomes increasingly sparse. For comparison, a 2,000-shot budget already covers more than 3% of the 216=65,5362^{16} = 65{,}536 configurations at N=16N = 16, where blind random sampling finds a ground state with 62% probability.

Table 1 gives the known global optimum E∗E^* [12], the two trap energies, their orbit counts and the range of dmin⁡d_{\min} for each size. The steepest-descent (SD) hit fractions show how frequently random-start descent reaches these tiers. With this classification in place, we evaluate every dynamical observable and escape ratio below on the orbits in this table.

Simulation and comparison

We now turn the escape problem into a common sampling test for quantum dynamics and classical search. To do so, we specify the evolution and its observables,

and compare the classical success rates through their energy evaluation cost. We then estimate the circuit resources needed to generate the quantum samples.

Time evolution and measured quantities

To follow coherent evolution under (4), we simulate the full statevector ∣ψ(t)⟩∈C2N\lvert\psi(t)\rangle\in\mathbb{C}^{2^N} using NVIDIA cuStateVec [51] accelerator by custom CUDA kernels. We initialize the state in the computational-basis configuration of the trap, ∣ψ(0)⟩=∣s0⟩\lvert\psi(0)\rangle= \lvert s_0\rangle.

Write the diagonal part as H^Z=H^B/N\hat{H}_Z = \hat{H}_B/N and the transverse part as H^X=−g∑iX^i\hat{H}_X = -g\sum_i \hat{X}_i. For a time step δt\delta t, define the symmetric second-order product formula [52, 53, 54, 55],

U^2(δt)=e−iδt2H^Ze−iδtH^Xe−iδt2H^Z,(8)\hat{U}_2(\delta t) = e^{-i\frac{\delta t}{2}\hat{H}_Z}e^{-i\delta t\hat{H}_X}e^{-i\frac{\delta t}{2}\hat{H}_Z}, \tag*{(8)}

which approximates e−iδtH^(g)e^{-i\delta t\hat{H}(g)} with a local error of O(δt3)\mathcal{O}(\delta t^3).

The implementation applies a full diagonal step followed by a transverse step. Denote this update by U^code(δt)=e−iδtH^Xe−iδtH^Z\hat{U}_{\mathrm{code}}(\delta t) = e^{-i\delta t\hat{H}_X}e^{-i\delta t\hat{H}_Z}. Writing D^=e−iδtH^Z/2\hat{D} = e^{-i\delta t\hat{H}_Z/2} for a half diagonal step gives U^code=D^−1U^2D^\hat{U}_{\mathrm{code}} = \hat{D}^{-1}\hat{U}_2\hat{D}. For an integer number mm of steps, this relation becomes U^codem=D^−1U^2mD^\hat{U}_{\mathrm{code}}^m = \hat{D}^{-1}\hat{U}_2^m\hat{D}. The initial basis state acquires only a common phase under D^\hat{D}, while the final D^−1\hat{D}^{-1} changes only the phases of the evolved state’s basis amplitudes. Consequently, the computational-basis probabilities agree exactly with those from (8) at every recorded time. The escape probabilities, Hamming displacements and energy distributions therefore inherit the symmetric formula’s accuracy.

The full diagonal step applies exp⁡(−iδtE(s)/N)\exp(-i\delta t E(s)/N) in parallel across all 2N2^N states using a custom kernel with hardware population counts (__popcll). The transverse mixer factorizes into single-qubit rotations, with I^\hat{I} denoting the identity operator,

eiδtg∑iX^i=∏i=1N[cos⁡(gδt)I^+isin⁡(gδt)X^i],(9)e^{i\delta t g\sum_i \hat{X}_i}=\prod_{i=1}^{N}\left[\cos(g\delta t)\hat{I}+i\sin(g\delta t)\hat{X}_i\right], \tag*{(9)}

applied sequentially via custatevecApplyMatrix. Simulations employ a Trotter step δt=0.05\delta t=0.05 across a total horizon t∈[0,60]t\in[0,60] (1,200 Trotter steps).

Once the state has evolved, we use its amplitudes cs(t)=⟨s∣ψ(t)⟩c_s(t)=\langle s|\psi(t)\rangle to follow both escape and spatial spreading. Summing their squared magnitudes over the target set S<S_{<} defined in § 2 gives the escape probability P<(t)P_{<}(t). We evaluate its average ⟨P<⟩\langle P_{<}\rangle over a fixed time interval [tstart,tmax⁡]=[10.0,38.5][t_{\mathrm{start}},t_{\max}]=[10.0,38.5]. At the step δt=0.05\delta t=0.05, this interval contains 571 recorded times tj=10+0.05jt_j=10+0.05j, with j=0,…,570j=0,\ldots,570. The reported average assigns equal weight to these times,

P<(t)=∑s:E(s)<E0∣cs(t)∣2,(10)P_{<}(t)=\sum_{s:E(s)<E_0}|c_s(t)|^2, \tag*{(10)}
⟨P<⟩=1571∑j=0570P<(tj).(11)\langle P_{<}\rangle=\frac{1}{571}\sum_{j=0}^{570}P_{<}(t_j). \tag*{(11)}

The interval begins after the initial delay in escape and provides a common sampling protocol for every seed and system size. Equivalently, ⟨P<⟩\langle P_{<}\rangle is the success probability when the measurement time is chosen uniformly from the 571 recorded times in this interval.

The window average measures the yield of the move, while the approach to the first peak reveals how quickly that yield develops. Let tpeak,1t_{\mathrm{peak},1} denote the time of the first local maximum of P<(t)P_{<}(t), using the overall peak when the trajectory rises monotonically. The first-peak growth indicator is

dP<dt∣rise=P<(tpeak,1)tpeak,1.(12)\left.\frac{dP_{<}}{dt}\right|_{\mathrm{rise}}=\frac{P_{<}(t_{\mathrm{peak},1})}{t_{\mathrm{peak},1}}. \tag*{(12)}

The figures label this quantity the initial rise rate. It measures growth up to a peak whose time depends on gg. § 4.1 relates this finite-time indicator to the short-time onset of escape.

To relate this growth to motion through configuration space, we also measure the mean Hamming displacement from the initial trap,

⟨dH(t)⟩=∑s∣cs(t)∣2dH(s,s0).(13)\langle d_H(t)\rangle=\sum_s|c_s(t)|^2d_H(s,s_0). \tag*{(13)}

Together, the displacement and escape probability distinguish spatial spreading from the accumulation of weight on lower-energy configurations.

Classical search and success rates

We compare quantum escape with the perturbation and local-search moves used in classical heuristics. Kick followed by descent is a standard component of iterated local search and basin hopping [18, 19]. Tabu search [56] complements these moves by keeping a short memory of the recent search, and underlies strong LABS algorithms [14, 15]. All baseline implementations use cached autocorrelation updates with O(N)O(N) cost per proposed spin flip. Counting cost-function evaluations supplies a common measure of sample yield, while the quantum evolution resources are assessed in § 3.3. Explicit resource accounting is also central to broader optimization benchmarks [17].

In kick plus steepest descent (Kick-SD), we flip dd randomly chosen spins of s0s_0 and descend to a local minimum. For each d∈{2,4,…,12}d\in\{2,4,\ldots,12\}, we run 50,000 trials. The success rate per evaluation is denoted by rSD∗(d)r_{\mathrm{SD}}^*(d), with one evaluation charged for the sampled string and NN for each descent step. The best kick rate is rSD∗=max⁡drSD(d)r_{\mathrm{SD}}^*=\max_d r_{\mathrm{SD}}(d). Restart plus steepest descent (Restart-SD) explores fresh basins by beginning from a uniformly random string. It uses 50,000 trials to estimate the rate rrestartr_{\mathrm{restart}} under the same accounting [20].

To allow the local search to move beyond repeated visits to the same configurations, we also use tabu search. It begins either directly at s0s_0 (Tabu-direct), after a dd-spin kick (Tabu-kick, optimized over dd), or at a uniformly random string (Tabu-restart). At each move, the tenure is chosen uniformly from [max⁡(2,⌊N/6⌋),max⁡(3,⌊N/3⌋)][\max(2,\lfloor N/6\rfloor),\max(3,\lfloor N/3\rfloor)]. An aspiration criterion accepts a tabu move when it strictly improves the best energy found. Each trial scans all NN single-flip neighbours per move, charging one evaluation for the initial string and NN for each scan, and stops upon reaching E<E0E<E_0 or after 100 moves. We run 50,000 trials per mode and per tested kick distance and denote the best rate by rtabu∗r_{\mathrm{tabu}}^*.

Each classical rate is the number of successful trials divided by the total number of energy evaluations across all trials. The trial count is fixed, while the total evaluation count depends on the search path and stopping time. This accounting compares the methods through their yield per energy evaluation.

We use the largest of these classical rates, denoted by rclassical∗r_{\mathrm{classical}}^*, to compare with the quantum yield. The quantum-to-classical escape ratio RR is therefore

R=⟨P<⟩rclassical∗,rclassical∗=max⁡(rSD∗,rrestart,rtabu∗).(14)R=\frac{\langle P_{<}\rangle}{r_{\mathrm{classical}}^*},\qquad r_{\mathrm{classical}}^*=\max(r_{\mathrm{SD}}^*,r_{\mathrm{restart}},r_{\mathrm{tabu}}^*). \tag*{(14)}

Thus RR measures the yield of lower-energy configurations per energy evaluation, with one evaluation assigned to verification of each quantum sample. State preparation and evolution determine the additional cost of generating that sample. Tabu search is the stronger classical baseline on most traps, exceeding the best steepest-descent variant by a factor of 1.2–2, with larger gains on shallow traps that Tabu-direct escapes from the initial state.

Classical rates are estimated from the successes accumulated in the 50,000 trials. For a run with zero successes, let NevalN_{\mathrm{eval}} be its total number of evaluations. We use $1/N_{\mathrm{eval}}$ as the one-hit resolution of the rate, and mark the corresponding RR as a lower-limit estimate under this convention. For tiers with several orbits, we report medians and 68% bootstrap intervals using $B=2000 resamples of the symmetry-distinct orbits [57]. A tier containing a single orbit is reported as a single value, with orbit count n=1n=1.

The target counts also allow a direct geometric comparison with the perturbation itself. Consider a kick that first chooses a distance dd with probability wdw_d and then chooses uniformly among the (Nd)\binom{N}{d} configurations at that distance from s0s_0. Here wd≥0w_d \ge0 and ∑d=0Nwd=1\sum_{d=0}^{N} w_d = 1. We call this an isotropic kick and denote its probability of reaching S<S_{<} by PisoP_{\mathrm{iso}}. Using the counts in Eq. (7),

Piso=∑d=0NwdMd(Nd)≤max⁡0≤d≤NMd(Nd).(15)P_{\mathrm{iso}} = \sum_{d=0}^{N} w_d \frac{M_d}{\binom{N}{d}} \le\max_{0 \le d \le N} \frac{M_d}{\binom{N}{d}}. \tag*{(15)}

Global spin inversion preserves the Bernasconi energy, so every target at distance dd has a target partner at distance N−dN-d. Thus Md=MN−dM_d = M_{N-d}, and targets occur only for dmin⁡≤d≤N−dmin⁡d_{\min} \le d \le N-d_{\min}, with dmin⁡d_{\min} defined in Eq. (6). Throughout this interval, (Nd)≥(Ndmin⁡)\binom{N}{d} \ge\binom{N}{d_{\min}}. Denoting the resulting upper bound by BisoB_{\mathrm{iso}}, we obtain

Piso≤Biso≡M(Ndmin⁡).(16)P_{\mathrm{iso}} \le B_{\mathrm{iso}} \equiv\frac{M}{\binom{N}{d_{\min}}}. \tag*{(16)}

Because the bound applies to every distribution of kick distances, it also covers a radius chosen separately for each seed and arbitrary mixtures of radii. It concerns the probability of reaching a target in the kick itself. Subsequent descent and tabu moves use the energies of visited configurations, so their performance is compared through Eq. (14). Together, these comparisons allow us to ask how much of the quantum yield comes from reaching distant shells and how much comes from favouring useful configurations within them.

Circuit cost

The sample yield becomes practically beneficial on a processor once the required evolution can be implemented. For the Bernasconi Hamiltonian, the main circuit cost comes from its two- and four-body phases. Expanding H^B\hat{H}_{B} in Eq. (3) yields N(N−1)(N−2)/6N(N-1)(N-2)/6 four-spin products sisi+ksjsj+ks_i s_{i+k}s_j s_{j+k} before simplification. Products with j=i+kj=i+k reduce to two-body terms (si+k2=1s_{i+k}^{2}=1), and coincident index sets merge, leaving a mixture of two- and four-body Pauli-ZZ operators. At N=30N=30, the Hamiltonian decomposes into exactly 210 two-body and 1,925 four-body commuting terms (and 182 two-body and 1,547 four-body terms at N=28N=28). Synthesizing these terms on an idealized all-to-all topology requires 2 CNOTs per two-body and 6 CNOTs per four-body term, totaling ≈12,000\approx12{,}000 two-qubit gates per Trotter step naively (≈9,650\approx9{,}650 at N=28N=28).

Routing on a processor with constrained connectivity increases the gate count, as shown for N=28N=28 in Table 2. To assess how time discretization affects the escape probability and establish the largest step tolerable for hardware synthesis, Table 3 compares the window-averaged probability ⟨P<⟩[10,38.5]\langle P_{<}\rangle_{[10,38.5]} for coarse second-order Trotter steps with the reference at δt=0.05\delta t=0.05, using the deepest traps at N=16N=16, 20 and 24 (g=0.35g=0.35). A step of δt=0.25\delta t=0.25 retains 92–98% of the reference probability (a loss of only 2–8%). At δt=0.5\delta t=0.5, the loss grows to 31–84%, and steps of δt≥0.75\delta t \ge0.75 retain at most 12% of the escape probability. The rapid deterioration past δt≈0.5\delta t \approx0.5 matches the threshold δt≤0.5\delta t \le0.5 found for non-commuting Hamiltonians by commutator-based error analysis [55].

ArchitectureCoupling Topology2Q Gates / Step
Naive All-to-AllIdealized (unconstrained)≈9.6×103\approx9.6 \times10^{3}
Square lattice (e.g. IBM Nighthawk)degree-4 planar grid≈1.9×104\approx1.9 \times10^{4}
IBM HeronHeavy-hex planar≈3.0×104\approx3.0 \times10^{4}

Table 2. Two-qubit gate requirements for a single second-order Trotter step at N=28N = 28 across physical hardware topologies compared to naive all-to-all synthesis.

NN (E0E_0)δt=0.25\delta t=0.250.50.751.0
16 (28)0.980.690.120.03
20 (34)0.920.260.010.01
24 (44)0.960.160.000.00

Table 3. Ratio of the window-averaged ⟨P<⟩\langle P_{<}\rangle at Trotter step δt\delta t to its value at δt=0.05\delta t=0.05 (second-order formula, g=0.35g=0.35).

Using δt=0.25\delta t=0.25 for the resource estimate, evolution to t∈[10,30]t \in[10,30] requires 40–120 steps. The resulting volume is of order 10610^{6} two-qubit gates, approximately 7×1057 \times10^{5}–2×1062 \times10^{6} on a square lattice and 1×1061 \times10^{6}–4×1064 \times10^{6} on heavy-hex at N=28N=28.

Escape dynamics

With the sampling protocol fixed, we can ask which features of the dynamics account for the enhanced escape. We first establish the gain over classical search and relate it to the distance and energies of the accessible states. We then examine how the transverse field and the duration of evolution control the transfer into these states.

Escape from deep minima

We begin with the central comparison in Figure 1 and Table 4. They show the escape ratio R=⟨P<⟩/rclassical∗R=\langle P_{<}\rangle/r_{\mathrm{classical}}^{*} for N∈{20,22,24,26,28,30}N \in\{20,22,24,26,28,30\}, using Level 1 traps at E0∗E_0^{*} and Level 2 traps at E0>E0∗E_0 > E_0^{*}. The energy tiers, orbit counts and barrier distances are those in Table 1. The same field g=0.35g=0.35 and sampling window [10,38.5][10,38.5] are used at every size, including N=28N=28 and 30, so the comparison follows one common escape protocol.

Energy TierN=20N = 20N=22N = 22N=24N = 24N=26N = 26N=28N = 28N=30N = 30
Level 1 (Deepest Traps)
Trap Energy E0E_0344744535867
Distinct Orbits (nn)4291614
Barrier Distance dmin⁡d_{\min} Range6–84–986–91010–13
Sample Median RR30.2215.8420.2026.7535.2411.87
68% Bootstrap CI[24.28, 66.17][14.87, 16.93]n=1n = 1[21.13, 32.79]n=1n = 1[7.93, 16.27]
Level 2 (Excited Traps)
Trap Energy E0E_0385548616275
Distinct Orbits (nn)7101120219
Barrier Distance dmin⁡d_{\min} Range5–72–694–9114–11
Sample Median RR27.2216.1025.5323.8713.3623.18
68% Bootstrap CI[26.26, 27.76][15.41, 16.83]n=1n = 1[19.23, 27.25][13.24, 13.48][21.21, 26.04]

Table 4. Quantum-to-classical escape ratio R=⟨P<⟩/rclassical∗R = \langle P_< \rangle/r^*_{\mathrm{classical}} at g=0.35g = 0.35, window [10,38.5][10, 38.5], where rclassical∗r^*_{\mathrm{classical}} is the best of steepest-descent kicks, restarts and tabu search (Eq. (14)). Medians over symmetry-distinct orbits with 68% bootstrap intervals. Entries marked ≥\ge use the one-hit resolution convention for zero-success classical runs. Levels follow Table 1. Orbits with shallow barriers dmin⁡≤2d_{\min} \le2 are marginal (a two-flip kick escapes them with ≈90%\approx90\% probability).

Figure 1. Quantum-to-classical escape ratio R=⟨P<⟩/rclassical∗R = \langle P_< \rangle/r^*_{\mathrm{classical}} versus system size NN for Level 1 (deepest traps) and Level 2 (second-hardest traps) at g=0.35g = 0.35, against the best of steepest-descent kicks, restarts and tabu search. Points denote medians over symmetry-distinct orbits with 68% bootstrap intervals (B=2000B = 2000). Upward arrows mark one-hit resolution estimates for zero-success classical runs. The dotted line denotes parity (R=1R = 1).

The median ratio exceeds unity at every size in both tiers. More specifically, Level 1 medians lie between 11.9×\times and 35.2×\times and Level 2 medians between 13.4×\times and 27.2×\times over N=20N=20–30 (Table 4). The ratio varies with the orbits present at each size and stays within a constant-factor range over N=20N=20–30. The same field and sampling window give an enhanced yield also at N=28N=28 and 30.

To understand this improvement, we turn to the distributions generated by the escape moves. An isotropic kick assigns equal probability to configurations at the chosen distance, after which descent concentrates the samples in nearby minima. Quantum evolution spreads amplitude over large Hamming distances (⟨dH⟩≈N/2\langle d_{\mathrm{H}}\rangle\approx N/2, Sec. 4.2) while enhancing the weight of low-energy configurations relative to blind sampling (Figure 4). The coexistence of broad spreading and enhanced low-energy weight leads us to examine the target geometry and the phases accumulated along the paths between configurations.

Figure 4. Energy distribution of the quantum state ∣ψ(t)⟩|\psi(t)\rangle initialized in the deepest N=28N = 28 trap (E0=58E_0 = 58, g=0.35g = 0.35). (a) Full distribution at t=0,10,30t = 0, 10, 30 compared to the blind density of states Pblind(E)=N(E)/2NP_{\mathrm{blind}}(E) = N(E)/2^N, where N(E)N(E) counts configurations of energy EE (grey dashed). While the mean energy ⟨E⟩≈154\langle E\rangle\approx154–156156 remains far below the random-string expectation N(N−1)/2=378N(N - 1)/2 = 378, the wavepacket broadens into low-lying states. (b) Detail of the low-energy region E≤76E \le76. The ground-state level E∗=50E^* = 50 (8 energy units below the trap) accumulates probability 2.60×10−42.60 \times10^{-4} by t=30t = 30, an enrichment factor of ∼104\sim10^4 over blind sampling (Pblind(E∗)=2.98×10−8P_{\mathrm{blind}}(E^*) = 2.98 \times10^{-8}).

Figure 2 makes this connection through the initial rise rate dP</dt∣risedP_</dt|_{\mathrm{rise}} and the window-averaged probability ⟨P<⟩\langle P_< \rangle across the evaluated traps, whose minimal barrier distances span dmin⁡∈[2,13]d_{\min} \in[2, 13]. Comparing these two observables shows how the distance to a target and the number of available targets enter at different stages of escape.

Figure 2. Dependence of quantum escape dynamics on minimal barrier distance dmin⁡d_{\min}. (a) Initial rise rate dPrisedt∣rise\left.\frac{dP_{\mathrm{rise}}}{dt}\right|_{\mathrm{rise}} versus dmin⁡d_{\min}, showing the rapid decrease in the first-peak growth indicator as the distance increases. (b) Window-averaged escape probability ⟨P⟩\langle P\rangle versus dmin⁡d_{\min}, showing the range of later-time yields across target distances and energy tiers. Data points are colored by system size N∈{20,22,24,26,28,30}N \in\{20,22,24,26,28,30\} and shaped by coupling strength g∈{0.30,0.35,0.45}g \in\{0.30,0.35,0.45\}.

For a fixed barrier distance (for instance, dmin⁡=6d_{\min} = 6), window escape probabilities vary by nearly two orders of magnitude (ranging from ⟨P<⟩≈3.2×10−4\langle P_< \rangle\approx3.2 \times10^{-4} up to 1.4×10−21.4 \times10^{-2}) depending on the trap’s energy tier E0E_0. The deepest classical traps (Level 1, E0∗E_0^*, with barriers dmin⁡=4d_{\min} = 4–13) systematically exhibit lower window escape probabilities (⟨P<⟩∼2.7×10−5\langle P_< \rangle\sim2.7 \times10^{-5}–1.7×10−31.7 \times10^{-3} across N∈{20,22,24,26,28,30}N \in\{20, 22, 24, 26, 28, 30\}), while excited local minima (Level 2, with dmin⁡=2d_{\min} = 2–11) achieve higher escape probabilities (up to ∼10−2\sim10^{-2}). This separation reflects the volume of the target set S<\mathcal{S}_{<} defined before Eq. (7). For excited traps, S<\mathcal{S}_{<} encompasses extensive lower-lying basins, whereas for Level 1 traps it is restricted to rare states near the global optimum (Pblind=1.49×10−8P_{\mathrm{blind}} = 1.49 \times10^{-8} at N=30N = 30).

To understand the early part of this evolution, we first use the distance in Eq. (6) to determine how escape begins. For a continuous-time quantum walk, the shortest paths determine the leading power of the transition probability [59]. Here the hypercube structure also fixes its coefficient. Consider a target s∈S<s \in\mathcal{S}_{<} at distance d=dH(s,s0)d = d_H(s,s_0), and let mm be a nonnegative integer power of the Hamiltonian. Each off-diagonal factor in Eq. (4) flips one spin, so reaching ss requires at least dd such factors. At that order, the required flips can occur in d!d! different sequences, giving

⟨s∣H^(g)m∣s0⟩=0(m<d),\langle s|\hat{H}(g)^m|s_0\rangle= 0 \quad(m < d),
⟨s∣H^(g)d∣s0⟩=(−g)dd!.(17)\langle s|\hat{H}(g)^d|s_0\rangle= (-g)^d d!. \tag*{(17)}

The factorial cancels the Taylor coefficient of the evolution operator. Since the Hamiltonian is real in the computational basis, transition probabilities are even functions of time. Summing over the nearest targets in Eq. (10) gives the exact onset law

P<(t)=Mdmin⁡(gt)2dmin⁡+O(t2dmin⁡+2).(18)P_<(t) = M_{d_{\min}}(gt)^{2d_{\min}} + \mathcal{O}(t^{2d_{\min}+2}). \tag*{(18)}

Distance fixes the power of time, while the number of nearest targets fixes its coefficient. This provides a direct explanation of the initial suppression in Figure 6.

Figure 6. Continuous-time escape probability P<(t)P_{<}(t) for the hardest local minimum at each system size N∈{20,22,24,26,28,30}N \in\{20, 22, 24, 26, 28, 30\} at g=0.35g = 0.35. Horizontal dashed lines indicate blind search probability PblindP_{\mathrm{blind}}.

The leading onset depends only on the shortest paths and their number. To see where the landscape energies enter, we next examine the first correction. For a fixed target at distance dd, let vjv_j be the configuration after jj flips along a shortest sequence, with v0=s0v_0 = s_0 and vd=sv_d = s. Denote the full sequence by π=(v0,v1,…,vd)\pi= (v_0, v_1, \ldots, v_d). Choose these d!d! sequences with equal weight. Define the energy offset ϵj\epsilon_j at vjv_j using the normalization of Eq. (4), and denote their sum along the sequence by SπS_\pi,

ϵj=E(vj)−E0N,Sπ=∑j=0dϵj.(19)\epsilon_j = \frac{E(v_j)-E_0}{N}, \qquad S_\pi= \sum_{j=0}^{d}\epsilon_j. \tag*{(19)}

To keep track of the phase accumulated during a sequence, let uju_j be the fraction of the total time spent at vjv_j. Integrating over the ordered flip times gives a uniform distribution over uj≥0u_j \ge0 with ∑j=0duj=1\sum_{j=0}^{d}u_j=1. With these fractions, the energy averaged along the path is Ys=∑j=0dujϵjY_s=\sum_{j=0}^{d}u_j\epsilon_j. We characterize its variation over both the paths and the time fractions by the variance VsV_s. Writing EπE_\pi for the uniform average over paths gives

Vs=Eπ[Sπ2+∑j=0dϵj2](d+1)(d+2)−(EπSπd+1)2≥0.(20)V_s=\frac{E_\pi\left[S_\pi^2+\sum_{j=0}^{d}\epsilon_j^2\right]}{(d+1)(d+2)}-\left(\frac{E_\pi S_\pi}{d+1}\right)^2\ge0. \tag*{(20)}

To see how this variance enters the transition probability, write Eπ,uE_{\pi,u} for the average over paths and time fractions (u0,…,ud)(u_0,\ldots,u_d). After removing the common phase e−iE0t/Ne^{-iE_0t/N}, the contribution with dd flips becomes (igt)dEπ,u[e−itYs](igt)^dE_{\pi,u}[e^{-itY_s}]. Its squared modulus supplies the correction −Vst2-V_st^2. In addition, paths with two further flips contribute the correction already present for the free transverse mixer. Combining the two contributions yields

∣⟨s∣e−itH^(g)∣s0⟩∣2=(gt)2d[1−t2{g2(N−2d/3)+Vs}+O(t4)].(21)\left|\langle s|e^{-it\hat{H}(g)}|s_0\rangle\right|^2 =(gt)^{2d}\left[1-t^2\{g^2(N-2d/3)+V_s\}+O(t^4)\right]. \tag*{(21)}

The coefficient proportional to g2g^2 follows by expanding the free mixer probability sin⁡2d(gt)cos⁡2(N−d)(gt)\sin^{2d}(gt)\cos^{2(N-d)}(gt). Thus the geometry determines the leading onset, and the spread of energies along the paths determines its first landscape-dependent correction. The nonnegative variance VsV_s makes this first landscape correction suppressive. The enhanced escape measured in the sampling window therefore develops beyond the earliest shortest-path contribution, as further spin flips and relative phases enter the evolution. This later evolution connects the window probability in Eq. (11) and the first-peak growth indicator to the field dependence observed in Figure 2.

We can now compare this early-time structure with the transfer observed inside the sampling window. Broad Hamming-space spreading (⟨dH⟩≈N/2\langle d_H\rangle\approx N/2, § 4.2) occurs together with enhanced weight on lower-energy configurations. This is the combination expected in energy-selective population transfer [37, 38]. The transverse field mixes configurations across Hamming shells, while the phases generated by H^B/N\hat{H}_B/N give the transfer its energy dependence. Window probabilities reach order 10−310^{-3} in the deepest tier and 10−210^{-2} in the second tier, even among traps with dmin⁡=6d_{\min}=6–1313, where the number of possible kicks (Ndmin⁡)\binom{N}{d_{\min}} reaches order 10710^7.

Figure 4 shows this directly for the deepest N=28N=28 trap. At t=30t=30 the weight on the ground-state energy E∗=50E^*=50 is 2.6×10−42.6\times10^{-4}, about 10410^4 times the blind fraction of 3.0×10−83.0\times10^{-8}, while the mean energy ⟨E⟩≈156\langle E\rangle\approx156 stays far below the random-string mean N(N−1)/2=378N(N-1)/2=378. The state therefore combines broad Hamming displacement with a strong preference for low energies. The enhanced ground-state weight occurs while the mean cost energy rises from the trap value 58 to about 156. Escape is therefore accompanied by a redistribution of probability towards both lower and higher classical energies.

To quantify the preference for low energies within the distance shells, we use Eq. (16). For the hardest orbits in Table 5, the target counts are M=8M=8 at N=28N=28 and M=16M=16 at N=30N=30, with dmin⁡=10d_{\min}=10 and 13, respectively. The corresponding isotropic-kick bounds are Biso=B_{\mathrm{iso}}=

6.10×10−76.10 \times10^{-7} and 1.34×10−71.34 \times10^{-7}. The window probabilities exceed these bounds by approximately 293 and 203, respectively. Since the bound holds for every choice of kick radius and for their mixtures, the enhancement requires a preference for successful configurations within the Hamming shells. The same bound applies to a time-averaged distribution, so the comparison applies directly to the sampling window used in the simulations.

This change in the mean cost energy has a neat interpretation through energy conservation. The relation between the cost and driver energies [35] connects the mean energy in Figure 4 to the transverse spin polarization. Write ⟨⋅⟩t\langle\cdot\rangle_t for expectation in the state evolved from ∣s0⟩|s_0\rangle. Initially ⟨X^i⟩0=0\langle\hat{X}_i\rangle_0=0, so continuous-time evolution under Eq. (4) conserves the total energy ⟨H^(g)⟩t=E0/N\langle\hat{H}(g)\rangle_t=E_0/N. Consequently,

⟨H^B⟩t=E0+gN⟨∑i=1NX^i⟩t,∣⟨H^B⟩t−E0∣≤∣g∣N2(22)\langle\hat{H}_B\rangle_t=E_0+gN\left\langle\sum_{i=1}^{N}\hat{X}_i\right\rangle_t, \qquad \left|\langle\hat{H}_B\rangle_t-E_0\right|\leq|g|N^2 \tag*{(22)}

where the inequality follows because the sum of NN Pauli-XX operators has eigenvalues between −N-N and NN. Define the variance of the full Hamiltonian by (ΔH)t2=⟨H^(g)2⟩t−⟨H^(g)⟩t2(\Delta H)_t^2=\langle\hat{H}(g)^2\rangle_t-\langle\hat{H}(g)\rangle_t^2. Acting once on ∣s0⟩|s_0\rangle, the driver produces NN orthogonal one-flip states, each with amplitude −g-g. The conserved variance is therefore

(ΔH)t2=g2N.(23)(\Delta H)_t^2=g^2N. \tag*{(23)}

At fixed NN and gg, every initial basis trap has the same standard deviation ∣g∣N|g|\sqrt{N} of the total energy, even though their escape probabilities differ. Their different yields probe the geometry of the lower-energy targets and the energies encountered on the way to them.

The low mean cost energy suggests a further question. How much of the enhanced ground-state weight follows from this mean alone, and how much depends on the way probability moves through the landscape? To make this comparison, write P(E,t)P(E,t) for the probability of measuring classical energy EE at time tt, and Eˉ(t)=∑EEP(E,t)\bar{E}(t)=\sum_E EP(E,t) for its mean. Let N(E)\mathcal{N}(E) count all configurations of energy EE, including their symmetry partners. Among distributions over these configurations with a prescribed mean, maximizing the Shannon entropy gives a canonical cost distribution [65]. Denote its dimensionless inverse temperature by β\beta and its partition function by Zcan(β)Z_{\mathrm{can}}(\beta). With the energy normalization of Eq. (4), they enter as

Pcan(E;β)=N(E)e−βE/NZcan(β),Zcan(β)=∑EN(E)e−βE/N.(24)P_{\mathrm{can}}(E;\beta)=\frac{\mathcal{N}(E)e^{-\beta E/N}}{Z_{\mathrm{can}}(\beta)}, \qquad Z_{\mathrm{can}}(\beta)=\sum_E\mathcal{N}(E)e^{-\beta E/N}. \tag*{(24)}

Thus the density of states fixes the relative abundance of each energy level, while the Boltzmann factor fixes the preference for low energies. For each snapshot, define the effective inverse temperature βeff(t)\beta_{\mathrm{eff}}(t) by matching its mean,

∑EEPcan(E;βeff(t))=Eˉ(t).(25)\sum_E E P_{\mathrm{can}}(E;\beta_{\mathrm{eff}}(t))=\bar{E}(t). \tag*{(25)}

The canonical mean decreases monotonically with β\beta, since its derivative is minus the canonical energy variance divided by NN. Consequently, this condition determines a unique inverse temperature, after which every energy-level probability is fixed.

For the N=28N=28 run in Figure 4, the saved density of states contains 803 distinct energies and sums to exactly 2282^{28} configurations. We use the recorded first moments 154.0269 at t=10t=10 and 155.8906 at t=30t=30 in Eq. (25).1

The corresponding inverse temperatures are 1.4351 and 1.3968. There are eight ground-state configurations at E∗=50E^* = 50, so Eq. (24) predicts ground-state probabilities of 1.61×10−41.61 \times10^{-4} and 1.40×10−41.40 \times10^{-4}, respectively. The measured values are 2.09×10−52.09 \times10^{-5} and 2.60×10−42.60 \times10^{-4}. To compare the distributions at every energy, define the enrichment η(E,t)\eta(E,t) as the measured probability relative to this canonical prediction,

η(E,t)=P(E,t)Pcan(E;βeff(t)).(26)\eta(E,t) = \frac{P(E,t)}{P_{\mathrm{can}}(E;\beta_{\mathrm{eff}}(t))}. \tag*{(26)}

Figure 5 shows both distributions and this ratio. At the ground-state energy, η(E∗,10)=0.129\eta(E^*,10) = 0.129 and η(E∗,30)=1.86\eta(E^*,30) = 1.86. In addition, the probability at the initial energy E0=58E_0 = 58 remains 0.0826 at t=30t = 30, about 881 times its canonical value. Substantial excess weight also persists in the high-energy tail. Between these two snapshots, the ground-state probability grows by a factor of 12.5 while the mean cost energy changes by only about 1.2%. The transfer therefore carries information beyond the mean, simultaneously retaining weight at the initial trap energy and increasing the population of the lower ground-state level.

Figure 5. Canonical cost comparison for the deepest N=28N = 28 trap (E0=58E_0 = 58, g=0.35g = 0.35), using the archived snapshots in Figure 4. (a) Recorded probabilities at t=10t = 10 and 3030 (solid lines with markers) and canonical distributions with the same recorded first moments (dashed lines in matching colours). The grey dotted curve is N(E)/2N\mathcal{N}(E)/2^N. Each canonical curve is normalized using the complete density of states, including energies beyond the displayed range. (b) Enrichment η(E,t)\eta(E,t) from Eq. (26) in the low-energy region. The horizontal line denotes agreement with the canonical prediction. The ground-state enrichment rises from 0.129 to 1.86, while the initial trap energy retains a much larger excess. Lines join discrete energy levels to guide the eye.

The distinction between cost energy and total energy is useful when interpreting this test. For a canonical reference based on the full quantum Hamiltonian, let βq\beta_q be its inverse temperature and let Π^E=∑s:E(s)=E∣s⟩⟨s∣\hat{\Pi}_E = \sum_{s:E(s)=E} |s\rangle\langle s| project onto the configurations of classical energy EE. Taking the trace over the full spin Hilbert space gives its prediction for a cost measurement,

Pq(E;βq)=Tr⁡[Π^Ee−βqH^(g)]Tr⁡e−βqH^(g).(27)P_q(E;\beta_q) = \frac{\operatorname{Tr}\left[\hat{\Pi}_E e^{-\beta_q \hat{H}(g)}\right]}{\operatorname{Tr}e^{-\beta_q \hat{H}(g)}}. \tag*{(27)}

Here βq\beta_q is fixed by the conserved mean total energy E0/NE_0/N, and the transverse field remains inside the Gibbs operator. Equilibration within conserved symmetry sectors also retains their initial weights [66]. Equation (26) quantifies the additional energy selection beyond a mean-matched canonical cost distribution. Field and time enter that reference through the measured mean, while η\eta records the further structure generated by the dynamics.

Energy conservation also constrains how broadly the classical energy can be sampled. A uniform distribution over bitstrings has mean energy N(N−1)/2N(N - 1)/2, as follows by averaging Eq. (3). For a seed below this mean, define gu(N,E0)g_u(N,E_0) as the lower bound on the field magnitude required to reach the uniform mean. Equation (22) gives

∣g∣≥gu(N,E0)≡N(N−1)/2−E0N2.(28)|g| \ge g_u(N,E_0) \equiv\frac{N(N - 1)/2 - E_0}{N^2}. \tag*{(28)}

Table 5 lists gug_u for the six hardest orbits. Its values lie between 0.3802 and 0.4089. For the N=28N = 28, E0=58E_0 = 58 seed, gu=0.40816g_u = 0.40816. At the adopted field g=0.35g = 0.35, Eq. (22) gives ⟨H^B⟩t≲332.4\langle\hat{H}_B\rangle_t \lesssim332.4, below the uniform mean 378. The continuous-time dynamics thus retains a mean energy below the uniform-string value even as the state spreads in Hamming distance. At g=0.45g = 0.45, the field exceeds gug_u for all six seeds, so this energy constraint allows the mean to reach the uniform value. The actual distribution is then determined by the transfer dynamics. These conservation relations refer to the continuous-time Hamiltonian used as the reference for the Trotter evolution in § 3.

NNE0E_0gug_udmin⁡d_{\min}PblindP_{\mathrm{blind}}Window ⟨P<⟩\langle P_{<}\rangleBest Classical Methodrclassical∗r_{\mathrm{classical}}^*Ratio RR
20340.390087.63×10−67.63 \times10^{-6}1.67×10−31.67 \times10^{-3}Tabu Direct4.79×10−54.79 \times10^{-5}34.9×
22470.380295.72×10−65.72 \times10^{-6}1.33×10−31.33 \times10^{-3}Kick-SD (d=12d = 12)8.52×10−58.52 \times10^{-5}15.6×
24440.402884.77×10−74.77 \times10^{-7}4.13×10−44.13 \times10^{-4}Tabu Restart2.05×10−52.05 \times10^{-5}20.2×
26530.402493.58×10−73.58 \times10^{-7}5.02×10−45.02 \times10^{-4}Tabu Kick (d=10d = 10)2.38×10−52.38 \times10^{-5}21.1×
28580.4082102.98×10−82.98 \times10^{-8}1.79×10−41.79 \times10^{-4}Tabu Kick (d=4d = 4)5.07×10−65.07 \times10^{-6}35.2×
30670.4089131.49×10−81.49 \times10^{-8}2.71×10−52.71 \times10^{-5}Tabu Kick (d=12d = 12)3.34×10−63.34 \times10^{-6}8.1×

Table 5. Hardest single orbit at each system size (g=0.35g = 0.35), defined as the Level 1 orbit with the largest dmin⁡d_{\min} (first evaluated orbit on ties). PblindP_{\mathrm{blind}} is the fraction of all configurations below E0E_0. The ratio R=⟨P<⟩/rclassical∗R = \langle P_{<}\rangle/r_{\mathrm{classical}}^* uses the best classical method listed. gug_u is the necessary field bound for reaching the uniform mean cost energy in Eq. (28). These are single-orbit values. Table 4 and the abstract quote medians over all orbits in a tier.

Effect of the transverse field

The preceding analysis shows how the landscape shapes the transfer. We now ask how this transfer changes when the transverse field becomes stronger. The field gg controls the balance between the diagonal Bernasconi potential H^B/N\hat{H}_B/N and the spin flips generated by ∑iX^i\sum_i \hat{X}_i. Figure 3 follows this balance through the window-averaged escape probability ⟨Pesc⟩\langle P_{\mathrm{esc}}\rangle and mean Hamming displacement ⟨dH⟩\langle d_H\rangle, for g∈{0.30,0.35,0.45}g \in\{0.30, 0.35, 0.45\} and N∈{22,24,26,28,30}N \in\{22, 24, 26, 28, 30\}.

Figure 3. Effect of transverse field gg on quantum escape and Hamming-space spreading for system sizes N∈{22,24,26,28,30}N \in\{22,24,26,28,30\}. (a) Window-averaged escape probability ⟨P<⟩\langle P_{<}\rangle versus gg, showing the field dependence of the low-energy yield within the chosen window. (b) Mean Hamming displacement ⟨dH⟩\langle d_H\rangle versus gg, showing increasing displacement towards N/2N/2. Error bars indicate standard error of the mean (SEM).

The transverse field alone gives the continuous-time walk on the hypercube [32, 33], providing a useful reference for this spreading. With the diagonal term removed, Eq. (9) makes each spin flip independently with probability q(t)=sin⁡2(gt)q(t) = \sin^2(gt) upon measurement. Denote the resulting escape probability by Pfree(t)P_{\mathrm{free}}(t) and the mean Hamming displacement by ⟨dH(t)⟩free\langle d_H(t)\rangle_{\mathrm{free}}. Then

Pfree(t)=∑d=0NMdq(t)d[1−q(t)]N−d,P_{\mathrm{free}}(t) = \sum_{d=0}^{N} M_d q(t)^d[1-q(t)]^{N-d},
⟨dH(t)⟩free=Nq(t).(29)\langle d_H(t)\rangle_{\mathrm{free}} = Nq(t). \tag*{(29)}

For the free mixer, the field enters only through gtgt, so changing gg rescales the time along the same evolution. Once the Bernasconi term is present, changing gg also changes the balance between spin flips and the phases accumulated from the fixed energy landscape.

To bound the escape produced by this free spreading, consider a flip probability qq and a target at distance dd. The probability qd(1−q)N−dq^d(1-q)^{N-d} is largest at q=d/Nq=d/N. Over the allowed target distances dmin⁡≤d≤N−dmin⁡d_{\min} \le d \le N-d_{\min}, this maximum is largest at the endpoints. Summing the bound over the MM targets gives

Pfree(t)≤M(dmin⁡N)dmin⁡(1−dmin⁡N)N−dmin⁡.(30)P_{\mathrm{free}}(t) \le M\left(\frac{d_{\min}}{N}\right)^{d_{\min}}\left(1-\frac{d_{\min}}{N}\right)^{N-d_{\min}}. \tag*{(30)}

The bound holds for every time and field, and hence for every sampling window. For the hardest N=28N=28 and N=30N=30 seeds it gives 9.50×10−89.50 \times10^{-8} and 1.95×10−81.95 \times10^{-8}, respectively. The observed window probabilities are about 1.88×1031.88 \times10^3 and 1.39×1031.39 \times10^3 times larger. The diagonal Bernasconi term is therefore essential to the observed transfer into lower-energy configurations, while

the mixer alone already produces a large Hamming displacement.

With this reference in place, the field dependence in Figure 3 traces a progression from partial spreading to broad exploration. At g=0.30g = 0.30, the mean Hamming displacement is approximately 5.965.96–9.239.23 for N∈{22,24,26,28,30}N \in\{22,24,26,28,30\}, or 43%43\%–84%84\% of N/2N/2. For example, ⟨dH⟩=6.60±0.15\langle d_H\rangle= 6.60 \pm0.15 at N=30N = 30, whose hypercube midpoint is 15. At N=20N = 20 the displacement is 7.15, or 71%71\% of N/2N/2. The potential retains substantial weight near the initial trap. This slower spreading can nevertheless produce a larger late-time escape probability for excited traps, as discussed in § 4.3.

Increasing the field to g=0.35g = 0.35 produces a pronounced expansion. For N∈{22,24,26,28,30}N \in\{22,24,26,28,30\}, the mean displacement rises to 10.8410.84–12.9012.90, or 86%86\%–99%99\% of N/2N/2, reaching 12.90±0.2012.90 \pm0.20 at N=30N = 30. The wavepacket now explores configurations well beyond the initial basin while retaining elevated escape probabilities, with the averages in Figure 3 ranging from about 1.38×10−41.38 \times10^{-4} to 9.79×10−39.79 \times10^{-3}.

Increasing the field further to g=0.45g = 0.45 brings the mean displacement to approximately N/2N/2. Its values are 11.0111.01, 12.0312.03, 13.0513.05, 14.0814.08 and 15.0415.04 for N=22N = 22, 2424, 2626, 2828 and 3030, respectively. However, the escape probability decreases at the same time, reaching about 8.63×10−58.63 \times10^{-5} at N=30N = 30. Stronger driving thus gives a larger displacement but a smaller yield of lower-energy configurations. This decrease occurs above the energy-conservation bound gug_u listed in Table 5, where a wider range of mean cost energies becomes accessible. The comparison with Eq. (30) shows why both observables are useful. Efficient escape requires the diagonal potential to shape the distribution as the transverse field spreads it. A superposition with equal weight on a configuration and its globally inverted partner has mean Hamming displacement N/2N/2 from the initial trap. Together with this displacement, the energy distribution identifies which parts of the landscape carry the probability.

The rapid change from partial spreading at g=0.30g = 0.30 to displacement above 85% of N/2N/2 at g=0.35g = 0.35 marks a finite-time crossover to broad Hamming-space spreading at the sizes studied. The adopted field, g=0.35g = 0.35, combines this broad spreading with a strong preference for low-energy configurations within the chosen window. A similar competition between spreading and energy selectivity appears in population transfer in random-energy and impurity-band models [37, 38, 42], whose extended-state regimes have also been studied analytically [41]. Here the deterministic Bernasconi potential retains the energetic preference while the field supplies enough mixing to traverse wide Hamming barriers. At weaker fields the transfer builds more slowly, and at stronger fields the additional spreading reduces the low-energy yield within the same window.

Growth of escape over time

The field scan compares averages over one interval. To see how those averages develop, we now follow the full escape trajectories and then vary the time at which they are compared. Figure 6 shows P<(t)P_{<}(t) for the hardest local minimum at each N∈{20,22,24,26,28,30}N \in\{20, 22, 24, 26, 28, 30\}, using g=0.35g = 0.35.

As expected from the short-time expansion, the trajectories begin with a delay of order t≲3t \lesssim3–55. All these seeds have dmin⁡≥6d_{\min} \ge6, so reaching the target requires several spin flips and P<(1)≲10−7P_{<}(1) \lesssim10^{-7} across the sizes shown. This suppression follows from the power of time in Eq. (18). During t≈5t \approx5–3030, the escape probability grows by two to four orders of magnitude as the state develops weight in lower-energy basins, reaching order 10−310^{-3} for the larger-probability trajectories.

For the deepest Level 1 traps at g=0.35g = 0.35 the late-time behaviour differs by instance. At N=28N = 28 (E0=58E_0 = 58, dmin⁡=10d_{\min} = 10) the probability has essentially plateaued by the end of the window (P<(60)=1.96×10−4P_{<}(60) = 1.96 \times10^{-4} against a window average of 1.79×10−41.79 \times10^{-4}), whereas at N=30N = 30 (E0=67E_0 = 67) it is still rising at t=60t = 60 (P<(60)=1.43×10−4P_{<}(60) = 1.43 \times10^{-4} for the dmin⁡=11d_{\min} = 11 orbit, 2.5×2.5 \times its window average). The late rise is more pronounced for Level 2 traps and for g=0.30g = 0.30, as the comparisons in Table 7 will show. For these rising trajectories, the probability measured at t=60t = 60 exceeds the fixed-window average.

NNTrap E0E_0dmin⁡d_{\min}Field gg⟨P<⟩[10,38.5]\langle P_{<}\rangle_{[10,38.5]}P<(t=60)P_{<}(t = 60)Factor
244480.307.59×10−57.59 \times10^{-5}7.09×10−57.09 \times10^{-5}0.93×
244890.309.28×10−49.28 \times10^{-4}3.99×10−33.99 \times10^{-3}4.30×
2858100.351.79×10−41.79 \times10^{-4}1.96×10−41.96 \times10^{-4}1.10×
2862110.301.37×10−51.37 \times10^{-5}3.69×10−53.69 \times10^{-5}2.69×
3067110.355.65×10−55.65 \times10^{-5}1.43×10−41.43 \times10^{-4}2.54×
3067130.352.71×10−52.71 \times10^{-5}6.34×10−56.34 \times10^{-5}2.34×
3067130.301.89×10−61.89 \times10^{-6}3.60×10−63.60 \times10^{-6}1.90×

Table 7. Comparison of window-averaged escape probability ⟨P<⟩[10,38.5]\langle P_{<}\rangle_{[10,38.5]} versus late-time escape probability P<(t=60)P_{<}(t = 60) for deep traps across N∈{24,28,30}N \in\{24, 28, 30\}.

To appreciate the size of this transfer, we compare these trajectories with the fraction of configurations that qualify as targets. For the hardest orbit at each size, Table 5 lists the blind success probability Pblind=M/2NP_{\mathrm{blind}} = M/2^N, using the target count in Eq. (7). It falls by more than two orders of magnitude, from 7.63×10−67.63 \times10^{-6} at N=20N = 20 to 1.49×10−81.49 \times10^{-8} at N=30N = 30. The quantum window probabilities range from 1.67×10−31.67 \times10^{-3} to 2.71×10−52.71 \times10^{-5}, showing a strong concentration of weight on the sparse lower-energy set.

Classical local search also uses the landscape to improve its proposals, so it gives a more demanding comparison than the blind fraction. Their best rate for the hardest orbit is 4.79×10−54.79 \times10^{-5} at N=20N = 20 and 3.34×10−63.34 \times10^{-6} at N=30N = 30, where the barrier distances in the table grow from 8 to 13. Relative to this best baseline, quantum sampling gives R=8.1R = 8.1–35.235.2 for the hardest individual orbits within the fixed window [10.0,38.5][10.0, 38.5]. Table 4 gives the corresponding medians over all orbits in each tier.

The same trajectories also let us revisit the choice of field. Denote by g∗g^* the transverse field giving the largest escape probability at the observation time. Although g=0.35g = 0.35 produces faster early transfer, the initially slower g=0.30g = 0.30 trajectory can continue to grow until it matches or exceeds the stronger-field result. Thus the preferred field depends on how long the move is allowed to develop.

Table 6 connects the preferred field to the time available for the move. At g=0.35g = 0.35, the state spreads rapidly beyond the initial basin and develops a larger early escape probability. At g=0.30g = 0.30, the initial growth is slower, but the probability continues to rise beyond t=40t = 40 and can overtake the stronger-field trajectory. The same competition between exploration and energetic preference therefore appears both in the field scan and in the ordering of the late-time probabilities. The dependence on observation time is also natural in quantum-walk optimization and population transfer [34, 35, 42]. Under unitary evolution, probability can flow both into and out of the lower-energy set. The crossing of trajectories at different fields therefore makes the measurement time part of the escape move, alongside the field strength.

NNTime ttP<(0.30)P_{<}(0.30)P<(0.35)P_{<}(0.35)P<(0.30)P<(0.35)\frac{P_{<}(0.30)}{P_{<}(0.35)}Dynamic Regime
N=22N = 22 (E0=47E_0 = 47, dmin⁡=6d_{\min} = 6)5.03.22×10−43.22 \times10^{-4}7.45×10−47.45 \times10^{-4}0.43g=0.35g = 0.35 faster rise
10.05.76×10−45.76 \times10^{-4}1.34×10−31.34 \times10^{-3}0.43g=0.35g = 0.35 faster rise
38.51.36×10−31.36 \times10^{-3}1.66×10−31.66 \times10^{-3}0.82g=0.30g = 0.30 catching up
50.01.59×10−31.59 \times10^{-3}8.07×10−48.07 \times10^{-4}1.96g=0.30g = 0.30 superior
60.01.56×10−31.56 \times10^{-3}1.14×10−31.14 \times10^{-3}1.37g=0.30g = 0.30 superior
N=26N = 26 (E0=61E_0 = 61, dmin⁡=9d_{\min} = 9)10.01.07×10−41.07 \times10^{-4}5.11×10−45.11 \times10^{-4}0.21g=0.35g = 0.35 faster rise
38.53.45×10−33.45 \times10^{-3}4.94×10−34.94 \times10^{-3}0.70g=0.30g = 0.30 catching up
50.06.95×10−36.95 \times10^{-3}4.49×10−34.49 \times10^{-3}1.55g=0.30g = 0.30 superior
60.09.19×10−39.19 \times10^{-3}6.11×10−36.11 \times10^{-3}1.50g=0.30g = 0.30 superior
N=30N = 30 (E0=75E_0 = 75, dmin⁡=4d_{\min} = 4)10.09.16×10−59.16 \times10^{-5}3.10×10−43.10 \times10^{-4}0.30g=0.35g = 0.35 faster rise
38.58.87×10−48.87 \times10^{-4}9.99×10−49.99 \times10^{-4}0.89g=0.30g = 0.30 catching up
50.01.41×10−31.41 \times10^{-3}1.07×10−31.07 \times10^{-3}1.31g=0.30g = 0.30 superior
60.01.72×10−31.72 \times10^{-3}1.26×10−31.26 \times10^{-3}1.37g=0.30g = 0.30 superior

Table 6. Escape probability P<(t)P_{<}(t) comparing couplings g=0.30g = 0.30 and g=0.35g = 0.35 for evaluated traps across N∈{22,26,30}N \in\{22, 26, 30\}. Early on (t≤10t \le10), g=0.35g = 0.35 leads. Later (t≥45−50t \ge45 - 50), g=0.30g = 0.30 maintains amplitude in low-energy basins, overtaking g=0.35g = 0.35. In the probability columns, the argument denotes the field strength at the listed time.

This time dependence also explains the role of the measurement window. Fixing the window gives a common protocol for every instance, while following the trajectories out to t=60t = 60 reveals how its choice affects the measured yield. Several Level 2 orbits and g=0.30g = 0.30 runs continue to rise, as does the dmin⁡=11d_{\min} = 11 orbit at N=30N = 30 under g=0.35g = 0.35. Other trajectories, including the deepest N=24N = 24 and N=28N = 28 traps at g=0.35g = 0.35, have largely plateaued within the observation interval. Table 7 compares the fixed-window average with the observed endpoint probability for representative traps. For the rising trajectories in Table 7, P<(60)P_{<}(60) exceeds the window average by factors up to 4.30. At N=30N = 30, E0=67E_0 = 67 and dmin⁡=11d_{\min} = 11, for example, the g=0.35g = 0.35 trajectory has ⟨P<⟩[10,38.5]=5.65×10−5\langle P_{<}\rangle_{[10,38.5]} = 5.65 \times10^{-5} and P<(60)=1.43×10−4P_{<}(60) = 1.43 \times10^{-4}, a factor of 2.54. For the hardest dmin⁡=13d_{\min} = 13 orbit at the same field, the fixed-window ratio is R=8.1R = 8.1 and the endpoint probability is 2.34 times the window average. These observations identify a further gain in sample yield available at the measured later time. They complement the R>1R > 1 medians at every size and tier under the same field and sampling window.

The same effect becomes more pronounced at the weaker field. At N=30N = 30, E0=67E_0 = 67 and dmin⁡=13d_{\min} = 13, g=0.30g = 0.30 gives ⟨P<⟩=1.89×10−6\langle P_{<}\rangle= 1.89 \times10^{-6} within [10,38.5][10,38.5], or R=0.57R = 0.57 relative to rclassical∗=3.34×10−6r_{\mathrm{classical}}^* = 3.34 \times10^{-6}. By t=60t = 60, the escape probability reaches 3.60×10−63.60 \times10^{-6}, a factor of 1.90 higher and slightly above the classical rate. The additional evolution time is therefore a physical resource that can change the ranking of the escape modes.

Conclusions

The numerical comparison and analytical results together identify the physical ingredients of successful escape. We bring these observations together here and use them to outline the next steps for understanding and implementing the dynamics.

Real-time transverse-field dynamics provides an effective escape move from the deep local minima of the Bernasconi model. The GPU statevector calculations cover N=20N = 20–30, up to approximately 1.07×1091.07 \times10^9 basis states, and begin from the complete set of symmetry-distinct traps in the two lowest excited tiers. With the same field and sampling window across all sizes, the median yield of lower-energy configurations exceeds the best classical perturbation rate by factors of 11.9–35.2 for Level 1 and 13.4–27.2 for Level 2. The improvement remains a constant factor across this range of sizes, including the larger instances at N=28N = 28 and 30.

The analytical results explain how this gain develops from the structure of the dynamics. The minimum target distance fixes the power of the short-time onset, the number of nearest targets fixes its coefficient, and the intermediate energies enter through the first correction. By the sampling times, this initially constrained motion has developed into broad Hamming-space spreading together with a strong preference for low-energy configurations. For the hardest N=28N = 28 and N=30N = 30 seeds, the window probabilities exceed the bound for every isotropic kick distribution by factors of approximately 293 and 203. The free-mixer bounds give an additional direct measure of the role played by the Bernasconi potential. Energy conservation explains how the mean cost energy remains below the uniform-string value during this broad spreading. The canonical cost comparison further separates the effect of this low mean energy from the selection generated by the dynamics. At N=28N = 28, the ground-state weight rises from below its mean-matched canonical value at t=10t = 10 to an enhancement of about 1.9 at t=30t = 30, while substantial weight persists at the initial trap energy.

The field scan brings the geometry and energy dependence together. At g=0.30g = 0.30, the mean displacement remains at 43%–84% of N/2N/2, falling to at most 54% for N≥24N \ge24. At g=0.35g = 0.35 it exceeds 85% of N/2N/2, while the low-energy yield remains enhanced. At g=0.45g = 0.45, the displacement approaches N/2N/2 and the escape probability falls. The operating field thus lies just beyond the rapid increase in spreading, where spatial exploration and energetic preference work together. The later overtaking by g=0.30g = 0.30 shows that the optimum also depends on the duration of the move. Even as the blind target fraction falls to 1.49×10−81.49 \times10^{-8} at N=30N = 30, the measured window probabilities remain of order 10−510^{-5}–10−310^{-3} for the hardest traps.

Turning from the escape mechanism to its implementation, the ratio RR characterizes sample efficiency, while state preparation and evolution determine the cost of generating each sample. Full-statevector simulation requires resources exponential in NN and reaches about N=30N = 30 on a single GPU in this study. The circuit estimate identifies the two- and four-body phases as the main hardware cost. These resource requirements give a concrete target for improved simulation and compilation while the present unitary dynamics establishes the escape mechanism.

These observations also suggest a natural next analytical step, namely to connect the Bernasconi spectrum near E0E_0 to the transfer rate and its field dependence, building on the population-transfer treatments of Refs. [37, 38]. The short-time expansion and the bounds in Eqs. (16) and (30) already constrain this development using the existing trajectories. In addition, time-dependent fields g(t)g(t) offer a way to combine rapid initial spreading with the larger late-time yield seen at weaker coupling. Matrix-product-state evolution and variational time-evolution methods provide possible routes to larger instances, with their cost set by the entanglement generated during the escape [60, 61, 62]. Finally, sample-based diagonalization and Krylov methods help clarify the role of the sampled states [63, 64]. For the diagonal cost Hamiltonian used here, its restriction to a span of sampled computational-basis states has those sampled energies as its eigenvalues. Its minimum is therefore the best energy already sampled, which also remained unchanged when descent was applied to the lowest-energy samples in the present runs. The quantum contribution lies in the distribution that makes these low-energy configurations available to the classical search.

Acknowledgements

The authors thank the Sharanga HPC cluster for providing the computational facility.

RB thanks support from an intramural grant CDRF C2/24/282 from BITS Pilani, and from ANRF India through ANRF/ARG/2025/009241/PS. RB acknowledges stimulating discussions with Rajatj Majumdar at IBM India and Indrakshi Raychowdhury on quantum algorithms, which motivated the present work.

ChatGPT Astra (OpenAI) was used to assist in framing the logical flow of the manuscript, and Gemini 3.8 Flash (Google) was used to assist in writing and maintaining the simulation and analysis code. The authors take responsibility for the final results and conclusions.

The code and numerical data are hosted on GitHub at github.com/ssk1001/bernasconi-krylov, with access provided upon request to the authors.

References

  1. [1]M. Golay, “Sieves for low autocorrelation binary sequences,” IEEE Transactions on Information Theory, vol. 23, pp. 43–51, 1977. doi:10.1109/tit.1977.1055653DOI
  2. [2]M. Golay, “The merit factor of long low autocorrelation binary sequences (Corresp.),” IEEE Transactions on Information Theory, vol. 28, pp. 543–549, 1982. doi:10.1109/tit.1982.1056505DOI
  3. [3]J. Jedwab, “A Survey of the Merit Factor Problem for Binary Sequences,” in Sequences and Their Applications—SETA 2004, Lecture Notes in Computer Science, vol. 3486, Springer, pp. 30–55, 2005. doi:10.1007/11423461_2DOI
  4. [4]J. Jedwab, D. J. Katz, and K.-U. Schmidt, “Advances in the merit factor problem for binary sequences,” Journal of Combinatorial Theory, Series A, vol. 120, pp. 882–906, 2013. doi:10.1016/j.jcta.2013.01.010DOI
  5. [5]K.-U. Schmidt, “Sequences with small correlation,” Designs, Codes and Cryptography, vol. 78, pp. 237–267, 2016. doi:10.1007/s10623-015-0154-7DOI
  6. [6]J. Bernasconi, “Low autocorrelation binary sequences: statistical mechanics and configuration space analysis,” Journal of Physique, vol. 48, pp. 559–567, 1987. doi:10.1051/jphys:01987004804055900
  7. [7]J. P. Bouchaud and M. Mézard, “Self induced quenched disorder: a model for the glass transition,” Journal of Statistical Physics I, vol. 4, pp. 1109–1114, 1994. doi:10.1051/jp1:1994240DOI
  8. [8]E. Marinari, G. Parisi, and F. Ritort, “Replica field theory for deterministic models. II. A non-random spin glass with glassy behaviour,” Journal of Physics A: Mathematical and General, vol. 27, pp. 7647–7668, 1994. doi:10.1088/0305-4470/27/23/011DOI
  9. [9]S. Mertens and C. Bessenrodt, “On the ground states of the Bernasconi model,” Journal of Physics A: Mathematical and General, vol. 31, pp. 3731–3749, 1998. doi:10.1088/0305-4470/31/16/004DOI
  10. [10]F. F. Ferreira, J. F. Fontanari, and P. F. Stadler, “Landscape statistics of the low-autocorrelation binary string problem,” Journal of Physics A: Mathematical and General, vol. 33, pp. 8635–8647, 2000. doi:10.1088/0305-4470/33/48/304DOI
  11. [11]S. Mertens, “Exhaustive search for low-autocorrelation binary sequences,” Journal of Physics A: Mathematical and General, vol. 29, pp. L473–L481, 1996. doi:10.1088/0305-4470/29/18/005DOI
  12. [12]T. Packebusch and S. Mertens, “Low autocorrelation binary sequences,” Journal of Physics A: Mathematical and Theoretical, vol. 49, p. 165001, 2016. doi:10.1088/1751-8113/49/16/165001DOI
  13. [13]S. Prestwich, “Exploiting relaxation in local search for LABS,” Annals of Operations Research, vol. 156, pp. 129–141, 2007. doi:10.1007/s10479-007-0226-9DOI
  14. [14]J. E. Gallardo, C. Cotta, and A. J. Fernández, “Finding low autocorrelation binary sequences with memetic algorithms,” Applied Soft Computing, vol. 9, pp. 1252–1262, 2009. doi:10.1016/j.asoc.2009.03.005DOI
  15. [15]B. Bošković, F. Brglez, and J. Brest, “Low-autocorrelation binary sequences: On improved merit factors and runtime predictions to achieve them,” Applied Soft Computing, vol. 56, pp. 262–285, 2017. doi:10.1016/j.asoc.2017.02.024DOI
  16. [16]B. Bošković, J. Herzog, and J. Brest, “Parallel self-avoiding walks for a low-autocorrelation binary sequences problem,” Journal of Computational Science, vol. 77, p. 102260, 2024. doi:10.1016/j.jocs.2024.102260
  17. [17]T. Koch, D. E. Bernal Neira, Y. Chen et al., “The Quantum Optimization Benchmarking Library,” Nature Computational Science, vol. 6, pp. 653–671, 2026. doi:10.1038/s43588-026-00991-1DOI
  18. [18]H. R. Lourenço, O. C. Martin, and T. Stützle, “Iterated Local Search,” in Handbook of Metaheuristics, vol. 57, Kluwer Academic Publishers, pp. 320–353, 2003. doi:10.1007/0-306-48056-5_11
  19. [19]D. J. Wales and J. P. K. Doye, “Global Optimization by Basin-Hopping and the Lowest Energy Structures of Lennard-Jones Clusters Containing up to 110 Atoms,” The Journal of Physical Chemistry A, vol. 101, pp. 5111–5116, 1997. doi:10.1021/jp970984nDOI
  20. [20]R. Martí, “Multi-Start Methods,” in Handbook of Metaheuristics, vol. 57, Kluwer, pp. 355–368, 2003. doi:10.1007/0-306-48056-5_12
  21. [21]S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by Simulated Annealing,” Science, vol. 220, pp. 671–680, 1983. doi:10.1126/science.220.4598.671DOI
  22. [22]A. B. Finnila, M. A. Gomez, C. Sebenik, C. Stenson, and J. D. Doll, “Quantum annealing: A new method for minimizing multidimensional functions,” Chemical Physics Letters, vol. 219, pp. 343–348, 1994. doi:10.1016/0009-2614(94)00117-0DOI
  23. [23]T. Kadowaki and H. Nishimori, “Quantum annealing in the transverse Ising model,” Physical Review E, vol. 58, pp. 5355–5363, 1998. doi:10.1103/physreve.58.5355DOI
  24. [24]E. Farhi, J. Goldstone, S. Gutmann, J. Lapan, A. Lundgren, and D. Preda, “A Quantum Adiabatic Evolution Algorithm Applied to Random Instances of an NP-Complete Problem,” Science, vol. 292, pp. 472–475, 2001. doi:10.1126/science.1057726DOI
  25. [25]G. E. Santoro, R. Martoňák, E. Tosatti, and R. Car, “Theory of Quantum Annealing of an Ising Spin Glass,” Science, vol. 295, pp. 2427–2430, 2002. doi:10.1126/science.1068774DOI
  26. [26]V. S. Denchev, S. Boixo, S. V. Isakov, N. Ding, R. Bab-bush, V. Smelyanskiy, J. Martinis, and H. Neven, “What is the Computational Value of Finite-Range Tunneling?,” Physical Review X, vol. 6, p. 031015, 2016. doi:10.1103/physrevx.6.031015DOI
  27. [27]T. Albash and D. A. Lidar, “Adiabatic quantum computation,” Reviews of Modern Physics, vol. 90, p. 015002, 2018. doi:10.1103/revmodphys.90.015002DOI
  28. [28]P. Hauke, H. G. Katzgraber, W. Lechner, H. Nishimori, and W. D. Oliver, “Perspectives of quantum annealing: methods and implementations,” Reports on Progress in Physics, vol. 83, p. 054401, 2020. doi:10.1088/1361-6633/ab85b8DOI
  29. [29]B. Altshuler, H. Krovi, and J. Roland, “Anderson localization makes adiabatic quantum optimization fail,” Proceedings of the National Academy of Sciences, vol. 107, pp. 12446–12450, 2010. doi:10.1073/pnas.1002116107DOI
  30. [30]E. Farhi and S. Gutmann, “Quantum computation and decision trees,” Physical Review A, vol. 58, pp. 915–928, 1998. doi:10.1103/physreva.58.915DOI
  31. [31]A. M. Childs and J. Goldstone, “Spatial search by quantum walk,” Physical Review A, vol. 70, p. 022314, 2004. doi:10.1103/physreva.70.022314DOI
  32. [32]O. Mülken and A. Blumen, “Continuous-time quantum walks: Models for coherent transport on complex networks,” Physics Reports, vol. 502, pp. 37–87, 2011. doi:10.1016/j.physrep.2011.01.002
  33. [33]C. Moore and A. Russell, “Quantum Walks on the Hypercube,” in Randomization and Approximation Techniques in Computer Science, Lecture Notes in Computer Science, vol. 2483, Springer, pp. 164–178, 2002. doi:10.1007/3-540-45726-7_14DOI
  34. [34]A. Callison, N. Chancellor, F. Mintert, and V. Kendon, “Finding spin glass ground states using quantum walks,” New Journal of Physics, vol. 21, p. 123022, 2019. doi: 10.1088/1367-2630/ab5ca2DOI
  35. [35]A. Callison, M. Fenstentein, J. Chen, L. Nita, V. Kendon, and N. Chancellor, “Energetic Perspective on Rapid Quenches in Quantum Annealing,” PRX Quantum, vol. 2, p. 010338, 2021. doi: 10.1103/prxquantum.2.010338DOI
  36. [36]K. Kechedzhi, V. Smelyanskiy, J. R. McClean, V. Denchev, M. Mohseni, S. Isakov, S. Boixo, B. Altshuler, and H. Neven, “Efficient population transfer via non-ergodic extended states in quantum spin glass,” arXiv:1807.04792, 2018.arxiv.org/abs/1807.04792
  37. [37]V. N. Smelyanskiy, K. Kechedzhi, S. Boixo, S. V. Isakov, H. Neven, and B. Altshuler, “Nonergodic Delocalized States for Efficient Population Transfer within a Narrow Band of the Energy Landscape,” Physical Review X, vol. 10, p. 011017, 2020. doi:10.1103/physrevx.10.011017DOI
  38. [38]C. L. Baldwin and C. R. Laumann, “Quantum algorithm for energy matching in hard optimization problems,” Physical Review B, vol. 97, p. 224201, 2018. doi:10.1103/physrevb.97.224201DOI
  39. [39]C. R. Laumann, A. Pal, and A. Scardicchio, “Many-Body Mobility Edge in a Mean-Field Quantum Spin Glass,” Physical Review Letters, vol. 113, p. 200405, 2014. doi:10.1103/physrevlett.113.200405DOI
  40. [40]C. L. Baldwin, C. R. Laumann, A. Pal, and A. Scardicchio, “The many-body localized phase of the quantum random energy model,” Physical Review B, vol. 93, p. 024202, 2016. doi:10.1103/physrevb.93.024202DOI
  41. [41]L. Faoro, M. V. Feigel’man, and L. Ioffe, “Non-ergodic extended phase of the Quantum Random Energy model,” Annals of Physics, vol. 409, p. 167916, 2019. doi:10.1016/j.aop.2019.167916DOI
  42. [42]T. Parolini and G. Mossi, “Multifractal Dynamics of the QREM,” arXiv:2007.00315, 2020.arxiv.org/abs/2007.00315
  43. [43]N. Chancellor, “Modernizing quantum annealing using local searches,” New Journal of Physics, vol. 19, p. 023024, 2017. doi:10.1088/1367-2630/aa59c4DOI
  44. [44]M. Ohkuwa, H. Nishimori, and D. A. Lidar, “Reverse annealing for the fully connected p-spin model,” Physical Review A, vol. 98, p. 022314, 2018. doi: 10.1103/physreva.98.022314DOI
  45. [45]D. Layden, G. Mazzola, R. V. Mishmash, M. Motta, P. Wojcian, J.-S. Kim, and S. Sheldon, “Quantum-enhanced Markov chain Monte Carlo,” Nature, vol. 619, pp. 282–287, 2023. doi:10.1038/s41586-023-06095-4
  46. [46]D. J. Egger, J. Mareček, and S. Woerner, “Warm-starting quantum optimization,” Quantum, vol. 5, p. 479, 2021. doi:10.22331/q-2021-06-17-479DOI
  47. [47]E. Farhi, J. Goldstone, and S. Gutmann, “A Quantum Approximate Optimization Algorithm,” arXiv:1411.4028, 2014.arxiv.org/abs/1411.4028
  48. [48]L. Zhou, S.-T. Wang, S. Choi, H. Pichler, and M. D. Lukin, “Quantum Approximate Optimization Algorithm: Performance, Mechanism, and Implementation on Near-Term Devices,” Physical Review X, vol. 10, p. 021067, 2020. doi:10.1103/physrevx.10.021067DOI
  49. [49]R. Shaydu​​lin, C. Li, S. Chakrabarti et al., “Evidence of scaling advantage for the quantum approximate optimization algorithm on a classically intractable problem,” Science Advances, vol. 10, p. eadm6761, 2024. doi:10.1126/sciadv.adm6761DOI
  50. [50]M. Sciorilli, G. Camilo, T. O. Maciel, A. Canabarro, L. Borges, and L. Aolita, “A competitive NISQ and qubit-efficient solver for the LABS problem,” arXiv:2506.17391, 2025.arxiv.org/abs/2506.17391
  51. [51]NVIDIA, “cuStateVec: A High-Performance Library for State Vector Quantum Simulators,” cuQuantum SDK documentation, accessed November 10, 2026.
  52. [52]H. F. Trotter, “On the product of semi-groups of operators,” Proceedings of the American Mathematical Society, vol. 10, pp. 545–551, 1959. doi:10.1090/S0002-9939-1959-0108732-6DOI
  53. [53]G. Strang, “On Construction and Comparison of Difference Schemes,” SIAM Journal on Numerical Analysis, vol. 5, pp. 506–517, 1968. doi:10.1137/0705041DOI
  54. [54]M. Suzuki, “Generalized Trotter’s formula and systematic approximants of exponential operators and inner derivations with applications to many-body problems,” Communications in Mathematical Physics, vol. 51, pp. 183–190, 1976. doi:10.1007/bf01609348DOI
  55. [55]A. M. Childs, Y. Su, M. C. Tran, N. Wiebe, and S. Zhu, “Theory of Trotter Error with Commutator Scaling,” Physical Review X, vol. 11, p. 011020, 2021. doi:10.1103/physrevx.11.011020DOI
  56. [56]F. Glover and M. Laguna, Tabu Search, Kluwer Academic Publishers, 1997. doi:10.1007/978-1-4615-6089-0DOI
  57. [57]B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap, Chapman & Hall/CRC, 1994. doi:10.1201/9780429246593DOI
  58. [59]B. E. Szigeti, G. Homa, Z. Zimborás, and N. Barankai, “Short-time behavior of continuous-time quantum walks on graphs,” Physical Review A, vol. 100, p. 062320, 2019. doi:10.1103/physreva.100.062320DOI
  59. [60]U. Schollwöck, “The density-matrix renormalization group in the age of matrix product states,” Annals of Physics, vol. 326, pp. 96–192, 2011. doi:10.1016/j.aop.2010.09.012DOI
  60. [61]J. Haegeman, J. I. Cirac, T. J. Osborne, I. Pižorn, H. Verschelde, and F. Verstraete, “Time-Dependent Variational Principle for Quantum Lattices,” Physical Review Letters, vol. 107, p. 070601, 2011. doi:10.1103/physrevlett.107.070601DOI
  61. [62]S. Paeckel, T. Köhler, A. Swobodа, S. R. Manmana, U. Schlöckow, and C. Hubig, “Time-evolution methods for matrix-product states,” Annals of Physics, vol. 411, p. 167998, 2019. doi:10.1016/j.aop.2019.167998DOI
  62. [63]J. Robledo-Moreno, M. Motta, H. Haas et al., “Chemistry beyond the scale of exact diagonalization on a quantum-centric supercomputer,” Science Advances, vol. 11, p. eadu9991, 2025. doi:10.1126/sciadv.adu9991DOI
  63. [64]J. Yu, J. Robledo Moreno, J. T. Iosue et al., “Quantum-Centric Algorithm for Sample-Based Krylov Diagonalization,” arXiv:2501.09702, 2025.arxiv.org/abs/2501.09702
  64. [65]E. T. Jaynes, “Information Theory and Statistical Mechanics,” Phys. Rev. 106, 620–630 (1957). doi:10.1103/PhysRev.106.620.DOI
  65. [66]L. D’Alessio, Y. Kafri, A. Polkovnikov, and M. Rigol, “From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics,” Adv. Phys. 65, 239–362 (2016). doi:10.1080/00018732.2016.1198134.DOI

Paper details

vlnc:W_6dz46crb
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

ChatGPT Astra, Gemini 3.8 Flash (bare)

View statement
Looking for related papers…
Contents