betaFeedback

WorksPhysics and Astronomy

Disorder-blind Mpemba acceleration of quantum state preparation

Abstract

Using the strong Mpemba effect to accelerate relaxation by initial-state design requires suppressing the slowest mode, which typically needs prior knowledge of the system. However, in disordered systems, this mode is sample-dependent, so a fixed preparation generally retains a slow-mode overlap. We show that a ``blind'' preparation, designed from the clean chain without knowledge of the disorder realisation or its distribution, can start farther from stationarity yet overtake a reference in typical samples of a dephased disordered XX chain. It requires only the chain geometry and the particle number. At strong dephasing, the density dynamics maps to a resistor chain, and an exact inverse-generator identity links the residual overlap to the spatial resistance profile. We find two distinct thresholds: a finite mean resistance suffices for diffusion, but the residual overlap decays as L−1/2L^{-1/2} with chain length LL only when the resistance variance is finite; rare weak links slow this decay. We predict the fraction of samples in which the blind preparation wins, which is an ensemble-level measure of its reliability. Finite-size calculations support both scaling regimes with interactions at strong dephasing. We give preparation circuits and a density readout for the residual overlap.

Preparing many-body quantum states is an important but challenging task in quantum computation. In open-system approaches, a target is encoded as a stationary state, and the system is prepared by relaxing toward it [1], including dissipative algorithms for thermal and ground-state preparation [2, 3]. At late times, relaxation is controlled by the slowest decaying mode, but its amplitude depends on the initial state. A preparation that begins farther from stationarity can thereby overtake one that begins closer—the Mpemba effect [4, 5, 6, 7, 8, 9, 10, 11]. If its overlap with the slowest mode vanishes, the long-time decay is governed by a faster mode, giving the strong Mpemba effect. This offers a way to shorten equilibration by changing the preparation while keeping the dynamics fixed. The strong effect has been observed in colloidal dynamics [12], a trapped ion [13–15], and studied in open quantum systems [16–18].

Exact mode cancellation generally requires knowledge of the generator or symmetry [16, 19–24] of the quantum system. In the dephased XX chain, for example, symmetry suppresses the slow-mode contribution to Néel-state imbalance in the clean system [25]. Such state preparation, however, becomes difficult to implement in the presence of disorder, since disorder deforms the slow mode from sample to sample. Previous work has found Mpemba crossings in quasiperiodic systems [22] and disordered spin networks [26], acceleration driven by rare loss-free regions in lossy chains [27], and symmetry-restoration Mpemba dynamics robust against on-site disorder on a superconducting processor [28]. On the other hand, hydrodynamic relaxation produces a robust effect in clean chaotic chains [29], while a temporary dissipation quench [30] or coupling to a reset channel [31] can suppress slow modes by changing the dynamics. What remains unknown is whether a fixed preparation designed from the clean system can robustly suppress the slowest relaxation mode across typical disordered realisations, without sample-specific knowledge or modification of the dynamics, and how it scales with system size.

We answer this for a dephased XX chain with random fields or hoppings. We introduce a “blind” preparation which is a smooth density modulation orthogonal to the clean slow mode, and compare it with a reference modulation proportional to that mode. At strong dephasing, the density dynamics maps to a one-dimensional (1D) chain of random resistors [32, 33]: the relaxation rate depends on total resistance, while the scaling of the residual slow-mode overlap depends on its spatial distribution. We find that a finite mean resistance suffices for diffusion, while the typical residual decreases as L−1/2L^{-1/2} only for finite resistance variance [34, 35]. Rare weak links slow this decrease even when relaxation remains diffusive. Gaussian and stable limit laws give the limiting residual distributions and predict the fraction of samples in which the blind preparation overtakes a reference. Exact quantum calculations support the scaling at finite dephasing, while finite-size calculations with interactions support both regimes at strong dephasing. Circuits that prepare the blind and reference states exactly at fixed particle number, and a density-based readout protocol are also given.

**Blind state preparation.—**We consider relaxation under a time-independent Markovian generator Lρ=Lρ\mathcal{L}\rho=\mathcal{L}\rho toward a unique stationary state ρss\rho_{\mathrm{ss}}. Here uniqueness holds within each conserved particle-number sector. Assume a complete biorthogonal mode expansion of L\mathcal{L} [36] and a simple, isolated slowest nonstationary mode with real λ1<0\lambda_1 < 0. Its right and left eigenmodes satisfy LR1=λ1R1\mathcal{L}R_1 = \lambda_1R_1 and L†L1=λ1L1\mathcal{L}^{\dagger}L_1 = \lambda_1L_1, respectively, with ⟨L1∣R1⟩=1\langle L_1|R_1\rangle= 1, where ⟨A∣B⟩=Tr⁡(A†B)\langle A|B\rangle= \operatorname{Tr}(A^\dagger B). A deviation δ=ρ(0)−ρss\delta= \rho(0)-\rho_{\mathrm{ss}} then evolves at late times as

δ(t)=⟨L1∣δ⟩R1eλ1t+O(eRe⁡λ2t),(1)\delta(t) = \langle L_1|\delta\rangle R_1e^{\lambda_1t} + O(e^{\operatorname{Re}\lambda_2t}), \tag*{(1)}

where λ2\lambda_2 is the next-slowest eigenvalue. The amplitude ⟨L1∣δ⟩\langle L_1|\delta\rangle controls how much the slow mode contributes to the relaxation. Let L1(0)L_1^{(0)} denote the left slow mode of the clean reference generator. We call a preparation “blind” if its deviation δblind\delta_{\mathrm{blind}} satisfies ⟨L1(0)∣δblind⟩=0\langle L_1^{(0)}|\delta_{\mathrm{blind}}\rangle= 0 and is chosen without information about the disorder realisation or its distribution. Disorder generally changes L1L_1, leaving a nonzero overlap Figure 1(a). Relative to a reference preparation δref\delta_{\mathrm{ref}} with the same initial distance from stationarity and nonzero slow-mode overlap, we define the residual

Figure 1. (a) In a disordered resistor chain with L=512L = 512 and α=1.25\alpha= 1.25, the slow mode ϕ1\phi_1 is deformed while the blind profile is unchanged; α\alpha is the resistance-tail exponent in Eq. (11). The mode is normalised to unit cos⁡(πx)\cos(\pi x) coefficient. The mode deformation has a small projection onto the blind profile, giving 1/∣r∣≈151/|r| \approx15 for this sample. (b) Projecting the deformation ϕ1−cos⁡πx\phi_1-\cos\pi x onto the blind profile gives r=−0.065r=-0.065, normalised by ∫01cos⁡2(πx)dx=1/2\int_0^1 \cos^2(\pi x)dx = 1/2.

r=⟨L1∣δblind⟩⟨L1∣δref⟩=⟨L1−L1(0)∣δblind⟩⟨L1∣δref⟩,(2)r = \frac{\langle L_1|\delta_{\mathrm{blind}}\rangle}{\langle L_1|\delta_{\mathrm{ref}}\rangle} = \frac{\langle L_1-L_1^{(0)}|\delta_{\mathrm{blind}}\rangle}{\langle L_1|\delta_{\mathrm{ref}}\rangle}, \tag*{(2)}

where rr measures the projection of the slow-mode deformation onto the fixed blind preparation Figure 1(b).

We quantify the preparation’s effectiveness using a distance DD, defined as a norm of the measured deviation from stationarity. Starting from the equally distant deviations in Eq. (2), we scale the blind deviation by κ≥1\kappa\ge1. Provided the slow mode contributes to the measured deviation, their late-time distance ratio approaches Dblind(t)/Dref(t)→κ∣r∣D_{\mathrm{blind}}(t)/D_{\mathrm{ref}}(t) \to\kappa|r|. For κ>1\kappa> 1, a late-time ratio below one guarantees a crossing and the blind preparation is closer to stationarity. This happens when

1<κ<R≡1/∣r∣.(3)1 < \kappa< R \equiv1/|r|. \tag*{(3)}

We call RR the reach—the threshold ratio of initial distances. For equal starting distances and 0<∣r∣<10 < |r| < 1, the time advantage approaches Δt=ln⁡R/∣λ1∣\Delta t = \ln R/|\lambda_1|, corresponding to ln⁡R\ln R slow-mode lifetimes [37]. If r=0r = 0, the slow mode is absent and relaxation is governed by a faster mode. We next calculate how rr scales with size and varies across samples.

Model and the resistance profile.—We study the dephased disordered XX chain [25, 38–40] with LL sites, N=L−1N=L-1 bonds, and open boundaries

H=∑i=1L−1Ji(σixσi+1x+σiyσi+1y)+∑i=1Lhiσiz,(4)H = \sum_{i=1}^{L-1} J_i(\sigma_i^x\sigma_{i+1}^x+\sigma_i^y\sigma_{i+1}^y) + \sum_{i=1}^{L} h_i\sigma_i^z, \tag*{(4)}

where the hoppings JiJ_i or fields hih_i are random, and σiz\sqrt{\sigma_i^z} is local dephasing on each site. After a Jordan-Wigner transformation, the L×LL \times L correlation matrix Mij=⟨ci†cj⟩M_{ij}=\langle c_i^\dagger c_j\rangle obeys an exact closed linear equation [41], which we use for our calculations (SM Sec. II). We apply the preceding framework to this closed evolution and measure density relaxation through D(t)=[L−1∑iδni(t)2]1/2D(t) = [L^{-1}\sum_i \delta n_i(t)^2]^{1/2}, where δni=ni−ni,ss\delta n_i=n_i-n_{i,\mathrm{ss}} and ni=Miin_i=M_{ii}. From here on, λ1\lambda_1 is the eigenvalue of the slowest mode contributing to D(t)D(t).

Dephasing does not directly change the densities, but damps off-diagonal correlations at rate 4γ4\gamma. In the diffusive regime, the density-relaxation rates vanish as L−2L^{-2}, separating the fast coherence transient from the slow density dynamics. At strong dephasing, eliminating the coherences gives classical nearest-neighbor hopping, with current Ji=Gi(ni−ni+1)\mathcal{J}_i=G_i(n_i-n_{i+1}) and conductance [25, 42]

Gi=8Ji2γ4γ2+(hi−hi+1)2.(5)G_i = \frac{8J_i^2\gamma}{4\gamma^2+(h_i-h_{i+1})^2}. \tag*{(5)}

A bond ii is hence a weak link when its hopping is small or the fields on its two sides differ strongly. Its resistance is ρi=1/Gi\rho_i=1/G_i: under steady flow, the density drop across bond ii is Jiρi\mathcal{J}_i\rho_i, and resistances add. At finite dephasing, longer-range coherences modify this local resistor. For equal bonds, the density modes are discrete cosines sampled at xi=(i−12)/Lx_i=(i-\frac{1}{2})/L, and the slowest nonstationary profile is ϕ1(0)(x)=cos⁡(πx)\phi_1^{(0)}(x)=\cos(\pi x). We compare occupation-diagonal states with fixed particle number QQ. Their initial deviations from the uniform stationary density ni,ss=ν=Q/Ln_{i,\mathrm{ss}}=\nu=Q/L are δniref(0)=acos⁡(πxi)\delta n_i^{\mathrm{ref}}(0)=a\cos(\pi x_i) and δniblind(0)=acos⁡(2πxi)\delta n_i^{\mathrm{blind}}(0)=a\cos(2\pi x_i). The reference follows the clean slow mode, while the blind profile is orthogonal to it by design. Here aa is the density amplitude; Appendix B prepares these profiles on a quantum circuit. The two profiles have equal root-mean-square norms; scaling the blind profile by κ\kappa makes it start κ\kappa times farther from stationarity. A Néel state has only an O(1/L)O(1/L) normalised overlap with smooth modes (SM Sec. IX), so we use smooth density modulations and density-profile readout.

In a disordered sample, calculating rr requires the spatial deformation of the slow density mode. Here the bond resistances enter via the normalised cumulative profile

uN(x)=∑i=1⌊xN⌋ρi∑i=1Nρi,0≤x≤1.(6)u_N(x) = \frac{\sum_{i=1}^{\lfloor xN\rfloor}\rho_i}{\sum_{i=1}^{N}\rho_i}, \qquad0 \le x \le1. \tag*{(6)}

This is the fraction of the total resistance to the left of xx. We write the sample-mean resistance as ρˉN=N−1∑i=1Nρi\bar{\rho}_N=N^{-1}\sum_{i=1}^{N}\rho_i. In the continuum limit, a clean chain has u(x)=xu(x)=x; disorder produces deviations from this line while preserving u(0)=0u(0)=0 and u(1)=1u(1)=1. Their magnitude is measured by ω=sup⁡0≤x≤1∣u(x)−x∣\omega=\sup_{0\leq x\leq1}|u(x)-x|.

**Rate and mode shape.—**For i.i.d. positive resistances with finite mean μρ\mu_\rho, the slowest rate of a long chain is

λ1≃−π2N2ρˉN,ρˉN→μρ.(7)\lambda_1\simeq-\frac{\pi^2}{N^2\bar{\rho}_N},\qquad\bar{\rho}_N\to\mu_\rho. \tag*{(7)}

It also applies to the short-range-correlated disorder here [32, 43, 44]. Thus ρˉN\bar{\rho}_N sets the leading rate; the spatial arrangement enters via relative corrections O(ω)O(\omega) [Fig. 2(b)]. For bounded uniform random fields, the leading expression recovers the disorder-averaged gap in [25]. On large scales, the density deviation evolves as ∂tδn=N−2KWδn\partial_t\delta n=N^{-2}K_W\delta n, with no flux at the ends and

KW=∂x∂W,W(x)=ρˉNu(x),(8)K_W=\partial_x\partial_W,\qquad W(x)=\bar{\rho}_Nu(x), \tag*{(8)}

where W(x)=N−1∑i=1⌊xN⌋ρiW(x)=N^{-1}\sum_{i=1}^{\lfloor xN\rfloor}\rho_i is the cumulative resistance and ∂W\partial_W is differentiation w.r.t. accumulated resistance. For a smooth profile with W′>0W'>0, it reduces to ∂W=(W′)−1∂x\partial_W=(W')^{-1}\partial_x. On the lattice, the corresponding relation gives the bondwise density drop (SM Sec. III). The eigenfunctions of KWK_W are the density relaxation modes. Weak links can make KWK_W far from the clean generator, but its inverse has an exact resistance-weighted form. For real profiles, define ⟨g∣f⟩=∫01g(x)f(x) dx\langle g|f\rangle=\int_0^1g(x)f(x)\,dx and Ff(x)=∫0xf(y) dyF_f(x)=\int_0^x f(y)\,dy. For mean-zero ff and gg

⟨g∣KW−1f⟩=−∫01Ff(x)Fg(x) dW(x).(9)\langle g|K_W^{-1}f\rangle=-\int_0^1F_f(x)F_g(x)\,dW(x). \tag*{(9)}

This inverse is linear in the resistance profile and the same identity holds bond by bond on the lattice (SM Sec. IV). After removing the overall resistance scale, its difference from the clean inverse is O(ω)O(\omega), even for very weak links. Since KWK_W and its inverse share the nonuniform eigenfunctions, this controls the slow-mode deformation. For each disorder realization, let ϕ1(x)\phi_1(x) be the slowest nonuniform eigenfunction of KWK_W, normalised so that its cos⁡(πx)\cos(\pi x) is one. Writing ϕ1=cos⁡(πx)+∑k≥2ckcos⁡(kπx)\phi_1=\cos(\pi x)+\sum_{k\geq2}c_k\cos(k\pi x), the blind preparation cos⁡(2πx)\cos(2\pi x) selects the coefficient c2c_2, so r=c2r=c_2. For ω≪1\omega\ll1, first-order perturbation theory gives

r=∫01g(x) [u(x)−x]+O(ω2),(10)r=\int_0^1g(x)\,[u(x)-x]+O(\omega^2), \tag*{(10)}
g(x)=43sin⁡(πx)sin⁡(2πx).g(x)=\frac{4}{3}\sin(\pi x)\sin(2\pi x).

Eq. (10) expresses the residual as a weighted measure of spatial resistance fluctuations, with g(x)g(x) setting the sensitivity to their position. The other coefficients ckc_k are given in SM Sec. V. As an initial example, consider a weak bond at x0x_0 with large resistance ρw\rho_w but small resistance share sw=ρw/∑iρi≪1s_w=\rho_w/\sum_i\rho_i\ll1, while the rest is uniform. Then r≃swg(x0)r\simeq s_wg(x_0): its first-order contribution vanishes at the ends and is largest near x0≃0.30x_0\simeq0.30 or 0.700.70, as follows analytically from gg. Multiplying all resistances by a common factor changes the relaxation time but leaves rr unchanged. The residual fixes the late-time distance ratio and reach in each sample, while its distribution gives the success probability across samples. In Figure 3(a,Figure 3b), a blind state starting twice as far away crosses the reference and then decays with the predicted ratio 2∣r∣2|r|.

Figure 3. Top: Exact quantum dynamics at γ=1\gamma= 1, L=64L = 64: root-mean-square density distance versus time in slow-mode lifetimes. The blind preparation starts κ=2\kappa= 2 times farther away, crosses the reference, and approaches the late-time ratio κ∣r∣\kappa|r|. Dashed curves show the same preparation with its slow-mode component removed. (a) Bounded fields; (b) random hoppings with α=1.25\alpha= 1.25. Bottom: Resistor-chain ensembles, with 1500 samples per size. (c) Median reach 1/∣r∣1/|r|. Growth is slower for heavy tails and the α=2.5\alpha= 2.5 data approach slope 1/21/2 slowly. (d) Fraction of samples that win for κ=20\kappa= 20 (solid) and 50 (dashed). Dotted curves are parameter-free predictions: the Gaussian result Eq. (12) and the stable-law result Eq. (14) with the Appendix C correction for α=1.25\alpha= 1.25.

**Resistance tails and residual scaling.—**We now determine how the resistance distribution controls the typical residual as the chain grows. For i.i.d. resistances with μρ\mu_\rho, write Sm=∑i=1m(ρi−μρ)S_m=\sum_{i=1}^{m}(\rho_i-\mu_\rho) for the accumulated excess resistance. The normalised profile obeys u(x)−x≃[SxN−xSN]/(Nμρ)u(x)-x\simeq[S_{xN}-xS_N]/(N\mu_\rho). Subtracting xSNxS_N pins these fluctuations to zero at both ends. Their size depends on how frequently resistive bonds occur. For power-law disorder, let

Pr⁡(ρi>t)∼Aρt−α,t→∞.(11)\Pr(\rho_i>t)\sim A_\rho t^{-\alpha},\qquad t\to\infty. \tag*{(11)}

Here Aρ>0A_\rho>0 is the tail amplitude and α\alpha controls the frequency of large resistances. We consider α>1\alpha>1, for which the mean is finite. Bounded resistances belong to the finite-variance regime, which we denote α=∞\alpha=\infty. For α>2\alpha>2, many bonds contribute to the profile fluctuations, and the central limit theorem gives a typical residual proportional to L−1/2L^{-1/2}. For 1<α<21<\alpha<2, the mean remains finite but the variance diverges. The largest of NN resistances is of order N1/αN^{1/\alpha} and carries a fraction of order N1/α−1N^{1/\alpha-1} of the total resistance. These rare weak links dominate the profile fluctuations, giving ∣r∣∼L1/α−1|r| \sim L^{1/\alpha-1}. At α=2\alpha= 2, the marginal divergence of the variance gives ∣r∣∼ln⁡L/L|r| \sim\sqrt{\ln L}/L. These scalings follow by applying the Gaussian and stable limit laws [45, 46] to Eq. (10) (SM Sec. VI). Figure 2(a) tests the corresponding slow-mode distortion exponents (Table I in Appendix A).

Figure 2. (a) Fitted exponent pdp_d of d:=∥ϕ1−cos⁡(πx)∥/∥cos⁡(πx)∥d := \lVert\phi_1-\cos(\pi x)\rVert/\lVert\cos(\pi x)\rVert versus 1/α1/\alpha (i.e., d∝Lpdd \propto L^{p_d}) for bounded fields h0=1h_0=1–16, random hopping Ji=vi2J_i=v_i^2, and Pareto resistances. The line gives max⁡(−1/2,1/α−1)\max(-1/2,1/\alpha-1), the predicted typical exponent of both dd and ∣r∣|r|. At α=2\alpha=2, there is an additional logarithmic correction. Diamonds extend Pareto (equivalent to hopping at zero field) to L=216L=2^{16}, as heavy tails slow convergence near α=2\alpha=2. (b) Fitted exponent pλp_\lambda of the relaxation rate ∣λ1∣∝Lpλ|\lambda_1|\propto L^{p_\lambda}. For random hoppings over L=64L=64–4096, ∣λ1∣|\lambda_1| remains close to L−2L^{-2}, while dd decays more slowly for α<2\alpha<2.

The residual introduces a second threshold in the diffusive regime: a finite mean resistance suffices for diffusion, while finite variance gives the L−1/2L^{-1/2} improvement of the preparation. For every α>1\alpha> 1, rr vanishes in probability, so a blind preparation beats a reference with probability tending to one for any fixed starting-distance ratio κ>1\kappa> 1. From the asymptotic exponents, halving the typical residual requires a 32-fold increase in LL at α=1.25\alpha= 1.25, versus a fourfold increase for finite-variance disorder. The regime can be determined from the microscopic disorder through Eq. (5) (SM Sec. VI). Independent fields uniform in [−h0,h0][-h_0,h_0] at fixed hopping give bounded resistances. For random hoppings Ji=vizJ_i = v_i^z, with independent viv_i uniform on (0,1](0,1] and hi=0h_i = 0, α=1/(2z)\alpha= 1/(2z). For independent exponentially distributed spacings ℓi\ell_i of mean ℓ\ell, tunneling Ji=J0e−ℓi/ξJ_i = J_0e^{-\ell_i/\xi} at uniform fields gives α=ξ/(2ℓ)\alpha= \xi/(2\ell). In this example, relaxation is diffusive for ξ>2ℓ\xi> 2\ell, whereas the residual follows L−1/2L^{-1/2} only for ξ>4ℓ\xi> 4\ell.

Success probabilities.—The residual distribution determines how often the blind preparation wins. For an initial distance ratio κ>1\kappa> 1, we define Pwin(κ)≡Pr⁡(κ∣r∣<1)P_{\mathrm{win}}(\kappa) \equiv\Pr(\kappa|r| < 1) as the fraction of samples in which it has a late-time advantage over the reference. When the resistance variance is finite, Nr\sqrt{N}r, viewed across disorder realisations, converges to a zero-mean Gaussian with standard deviation s=2σ/(3μρ)s = 2\sigma/(3\mu_\rho). Here σ2\sigma^2 is the asymptotic variance per bond of the total resistance, including correlations (Appendix A). The Gaussian approximation hence gives

Pwin(κ)≃erf⁡(N2 κs).(12)P_{\mathrm{win}}(\kappa) \simeq\operatorname{erf}\left(\frac{\sqrt{N}}{\sqrt{2}\,\kappa s}\right). \tag*{(12)}

For independent resistances with 1<α<21 < \alpha< 2, rare weak links instead give a symmetric stable limit [46–48]

N1−1/αr→dsαZα,(13)N^{1-1/\alpha}r \xrightarrow{d} s_\alpha Z_\alpha, \tag*{(13)}

where ZαZ_\alpha is a standard symmetric α\alpha-stable random variable describing the spread of rr over disorder realisations, and the scale sαs_\alpha depends only on α\alpha, the tail amplitude AρA_\rho and the mean μρ\mu_\rho (Appendix A). The limit is symmetric because g(1−x)=−g(x)g(1-x)=-g(x): weak links at mirror positions contribute with opposite signs. Writing FαF_\alpha for the cumulative distribution function of ZαZ_\alpha, the corresponding prediction is

Pwin(κ)≃2Fα(N1−1/α/(κsα))−1.(14)P_{\mathrm{win}}(\kappa) \simeq2F_\alpha\left(N^{1-1/\alpha}/(\kappa s_\alpha)\right)-1. \tag*{(14)}

For a specified resistance distribution and ratio κ\kappa, this gives the fraction of samples in which the blind preparation wins. Figure 3(c,Figure 3d) compares the resulting reach and success probabilities with resistor-chain ensembles.

Interactions.—Adding V∑iσizσi+1zV\sum_i \sigma_i^z\sigma_{i+1}^z of strength VV to HH breaks the closed evolution of MM. At strong dephasing, the dynamics becomes a classical lattice gas whose hopping rates depend on the neighboring spins [49–51]

Gi(σ)=8Ji2γ4γ2+[hi+1−hi+V(σi+2z−σi−1z)]2.(15)G_i(\sigma) = \frac{8J_i^2\gamma}{4\gamma^2 + [h_{i+1}-h_i + V(\sigma_{i+2}^z-\sigma_{i-1}^z)]^2}. \tag*{(15)}

Forward and reverse rates are equal, so the stationary distribution remains uniform within each particle-number sector. The slow mode is now a many-body function, but its density profile still determines rr for weak density modulations. In a clean chain, the reflection-even blind preparation is orthogonal to a reflection-odd slow mode; this cancellation holds for all interaction strengths studied here. At half filling, particle-hole symmetry gives r(V)=r(−V)r(V)=r(-V) in each sample, so the leading correction near

V=0V = 0 is quadratic (SM Sec. VII). For bounded fields and finite VV, the rates remain proportional to Ji2J_i^2 up to bounded positive factors, preserving the resistance-tail exponent generated by weak hoppings.

Numerical results.—We test the scaling predictions using resistor chains up to L=216L = 2^{16}. Figure 2 shows that the size dependence of the slow-mode deformation changes at α=2\alpha= 2, while the rate retains its diffusive L−2L^{-2} scaling. Exact quantum calculations support this at finite dephasing: for bounded fields at γ=1\gamma= 1, the mean ∣r∣|r| has fitted exponent −0.49±0.02-0.49 \pm0.02 (SM Figure S2), while heavy-tailed hoppings give a more slowly decreasing slow-mode distortion (SM Figure S5). Figure 3(a,b) shows the resulting Mpemba effect: a blind preparation starting twice as far from stationarity overtakes the reference, with a late-time distance ratio 2∣r∣2|r|. Across resistor-chain ensembles, the median reach and success probability follow the predicted size dependence [Fig. 3(c,d)]. The heavy-tail comparison includes the finite-size approximation of Appendix C.

Figure S2. Full quantum dynamics, random fields h0=2h_0=2. (a) Mean slow-mode distortion dd and residuals at γ=1\gamma=1, with fitted exponents, for the blind profile cos⁡2πx\cos2\pi x and for a random blind state orthogonal to cos⁡πx\cos\pi x. The latter spreads over all modes and its residual falls as L−1L^{-1} (SM Sec. V). (b) Exponents over L=24L=24–9696 across dephasing rates.

Figure S5. Slow-mode distortion in the full quantum dynamics (points) and in the classical chain (line), random hoppings with (a) α=2.5\alpha=2.5 and (b) α=1.25\alpha=1.25. Legends give exponents over L=48L=48–256256.

The interacting extension in Figure 4 supports persistence of both residual-scaling regimes at strong dephasing (SM Sec. VII). For bounded fields, L rms(r)\sqrt{L}\,\mathrm{rms}(r) is approximately independent of size for the interaction strengths studied: interactions lower the prefactor while the data remain consistent with L−1/2L^{-1/2} residual scaling [Fig. 4(a)]. At V=γ=4V = \gamma= 4, sample-by-sample residuals track the configuration-averaged resistor chain [Fig. 4(b)]. For heavy-tailed hoppings, the estimated median is close to unity at sizes 16≤L≤2416 \le L \le24, with and without bounded fields [Fig. 4(c)]. The averaged-rate prediction tracks these residuals closely [Fig. 4(d)]. The blind and reference states can be prepared at fixed particle number without sample-specific information, and the residual rr can be measured from site-resolved densities (Appendix B).

Figure 4. Interacting strong-dephasing lattice gas at half filling and γ=4\gamma= 4. Exact diagonalisation is used for L≤20L \le20; DMRG [52, 53] is used beyond the dotted line. (a) Bounded fields hi∈[−4,4]h_i \in[-4,4]: L rms(r)\sqrt{L}\,\mathrm{rms}(r) is approximately independent of LL for the interaction strengths shown, consistent with ∣r∣∼L−1/2|r| \sim L^{-1/2}. (b) At V=4V = 4, the same-sample residuals are compared with a resistor chain using configuration-averaged rates; the solid fit has slope 1.61 and the dashed line has unit slope. (c) For heavy-tailed hoppings with α=1.25\alpha= 1.25, the median ∣r(V=4)/r(V=0)∣|r(V = 4)/r(V = 0)| is shown with and without fields. (d) At V=4V = 4 and L=24L = 24, the heavy-tailed residuals are compared with the averaged-rate chain; the fit has slope 0.98.

Discussion.—Our results make initial-state design a control resource for relaxation-based many-body state preparation. A profile chosen from the clean system can accelerate equilibration in typical disordered realisations while the generator remains fixed. This offers a route to improve equilibration when disorder varies between experimental runs or when characterising each sample is impractical. Diffusive transport alone is hence insufficient to assess preparation performance: fluctuations in the spatial distribution of resistance also matter. Specifically, rare weak links can leave macroscopic diffusion intact while substantially slowing the convergence of disorder-blind mode suppression depending on the heavy tail. The circuit preparation and density readout of Appendix B make the predicted overlap suppression testable without reconstructing the slow mode. The predicted success probability quantifies the resulting reliability across an ensemble. This addresses a source of uncertainty distinct from state-preparation errors, whose impact on Mpemba acceleration is investigated [54]. Combining these two sources of uncertainty can provide further assessment of experimental performance in state preparation. Establishing how broadly this strategy extends to various interacting models, higher dimensions, and weaker dephasing is an interesting next step.

Acknowledgements.—A.Y. and A.I. are supported by UKRI Future Leaders Fellowship (Grant No. 10128920).

Data availability.—The code and data supporting the findings of this work are available at https://github.com/NotNish7/mpemba_blind_disorder.

References

End Matter

Scaling and distribution

Appendix A: Scaling and distribution.—Table I summarizes the typical size dependence of the slowest rate and blind residual. We next give the disorder-dependent scale factors for the Gaussian and stable limits.

For stationary resistances with summable covariances, the finite-variance scale is determined by

σ2=lim⁡N→∞1NVar⁡(∑i=1Nρi)=Var⁡(ρ1)+2∑m≥1Cov⁡(ρ1,ρ1+m).\begin{aligned} \sigma^2=\lim_{N\to\infty}\frac{1}{N}\operatorname{Var}\left(\sum_{i=1}^{N}\rho_i\right) \\ &=\operatorname{Var}(\rho_1)+2\sum_{m\geq1}\operatorname{Cov}(\rho_1,\rho_{1+m}). \end{aligned}

For independent resistances this reduces to the single-bond variance. Since the weight in Eq. (10) satisfies ∫01g(x)2dx=4/9\int_0^1 g(x)^2\mathrm{d}x=4/9, the Gaussian scale is s=2σ/(3μρ)s=2\sigma/(3\mu_\rho). For independent resistances with tail (11) and 1<α<21<\alpha<2, the stable scale is

sα=σαμρ(∫01∣g(x)∣αdx)1/α,σαα=AρΓ(1−α)cos⁡(πα/2)s_\alpha=\frac{\sigma_\alpha}{\mu_\rho}\left(\int_0^1|g(x)|^\alpha\mathrm{d}x\right)^{1/\alpha},\quad\sigma_\alpha^\alpha=A_\rho\Gamma(1-\alpha)\cos(\pi\alpha/2)

Here Γ\Gamma is the gamma function, σα\sigma_\alpha sets the scale of centred resistance sums, whereas sαs_\alpha sets the dimensionless residual scale. We write Zα,βZ_{\alpha,\beta} for a standard α\alpha-stable random variable with skewness β∈[−1,1]\beta\in[-1,1], unit scale and zero location

E eiθZα,β=exp⁡{−∣θ∣α[1−iβsgn⁡(θ)tan⁡(πα/2)]}.\mathbb{E}\,e^{i\theta Z_{\alpha,\beta}}=\exp\{-|\theta|^\alpha[1-i\beta\operatorname{sgn}(\theta)\tan(\pi\alpha/2)]\}.

Thus Zα=Zα,0Z_\alpha=Z_{\alpha,0} in Eq. (13) is symmetric, whereas the centred resistance sum SN=∑i(ρi−μρ)S_N=\sum_i(\rho_i-\mu_\rho) in Appendix C has the maximally right-skewed limit Zα,1Z_{\alpha,1} after rescaling. The former describes the signed profile-weighted numerator of rr; the latter describes its fluctuating total-resistance denominator. Detailed derivations are given in SM Sec. VI.

Appendix B: Preparation with a quantum circuit and measurement.—To turn our disorder-blind initial-state design into an experimentally testable control strategy, we must connect the proposed spatial profiles to both a physical preparation and a measurable signature of slow-mode suppression. Two issues arise on the preparation and measurement sides: (i) a prescribed density profile does not uniquely specify a many-body state, and (ii) the residual slow-mode overlap is defined through a sample-dependent left eigenoperator that is generally unknown.

We address (i) by constructing diagonal states with exactly QQ particles, whose preparation weights depend only on the chosen profile, particle number, and modulation amplitude. For (ii), we compare the density responses of the blind and reference preparations in the same disorder realisation. In linear response, once a single slow mode dominates, their ratio eliminates the common decay and readout factors and yields the normalised residual overlap rr. This provides an experimental test of the proposed relaxation mechanism without prior characterisation of the microscopic disorder or reconstruction of its slow eigenmode.

For fi=cos⁡(kπxi)f_i=\cos(k\pi x_i), with k=1k=1 for the reference and k=2k=2 for the blind preparation, we prepare a diagonal state with occupation probabilities

pf(η)=exp⁡(ϵ∑ifiηi)ZQ(ϵ),∑iηi=Q.p_f(\eta)=\frac{\exp(\epsilon\sum_i f_i\eta_i)}{Z_Q(\epsilon)},\qquad\sum_i\eta_i=Q.

Here ηi∈{0,1}\eta_i\in\{0,1\}, ZQ(ϵ)Z_Q(\epsilon) normalises the distribution, and ϵ\epsilon controls the modulation. For small ϵ\epsilon, ⟨ni⟩−ν=ϵχLfi+O(ϵ2)\langle n_i\rangle-\nu=\epsilon\chi_L f_i+O(\epsilon^2), where ν=Q/L\nu=Q/L and χL=ν(1−ν)L/(L−1)\chi_L=\nu(1-\nu)L/(L-1). Thus the density amplitude of the main text is a=ϵχLa=\epsilon\chi_L to first order. A sequential circuit first prepares ∑ηpf(η)∣η⟩\sum_\eta\sqrt{p_f(\eta)}\lvert\eta\rangle using rotations controlled by a register that tracks the remaining particle number. Applying independent random ZZ gates and discarding their classical record removes the coherences, yielding the desired diagonal state exactly. Figure 5 shows a four-qubit implementation. The circuit fixes QQ at its output; an alternative using classical sampling and SWAP gates conserves it throughout (SM Sec. VIII). Background Hamiltonian evolution and dephasing must be negligible or controlled during the coherent gates.

Four-qubit preparation circuit for the blind state

Figure 5. Four-qubit preparation of the blind state with Q=2Q=2 and f=(1,−1,−1,1)/2f=(1,-1,-1,1)/\sqrt{2}. All qubits start in ∣0⟩\lvert0\rangle. Boxes denote RyR_y rotations, with angles specified in SM Sec. VIII. The CNOTs onto s4s_4 set its occupation to the parity of the first three sites. Independent fair bits ziz_i implement the final phase averaging when their record is discarded.

Having specified how to prepare the initial states, we next describe how to measure their slow-mode suppression. To measure rr, prepare each profile at amplitudes ±a\pm a (biases ±ϵ\pm\epsilon) in the same disorder realisation and measure A1=∑icos⁡(πxi)(ni−Q/L)A_1=\sum_i\cos(\pi x_i)(n_i-Q/L) from site-resolved densities. Once the slow mode dominates, its common decay and readout factors cancel

⟨A1⟩2,+a(t)−⟨A1⟩2,−a(t)⟨A1⟩1,+a(t)−⟨A1⟩1,−a(t)→t→∞r+O(a2).\frac{\langle A_1\rangle_{2,+a}(t)-\langle A_1\rangle_{2,-a}(t)}{\langle A_1\rangle_{1,+a}(t)-\langle A_1\rangle_{1,-a}(t)} \xrightarrow[t\to\infty]{}r+O(a^2).

Using both signs of the amplitude cancels the even-order nonlinear response, and the ratio of the two profiles cancels the unknown decay and readout factors; a single sign would leave an O(a)O(a) error. The symmetric difference leaves an O(a2)O(a^2) bias for this state family. The affine preparations in SM Sec. VIII remove this bias at every allowed amplitude. That section also gives the general circuit construction and explicit rotation angles. Appendix C: Finite-size approximation for heavy tails. — The first-order residual retains the fluctuating total resistance in its denominator

r≃∑i=1Ng(i/N)(ρi−μρ)Nμρ+SN,SN=∑i=1N(ρi−μρ).r\simeq\frac{\sum_{i=1}^{N}g(i/N)(\rho_i-\mu_\rho)}{N\mu_\rho+S_N},\qquad S_N=\sum_{i=1}^{N}(\rho_i-\mu_\rho).

Here SN/(σαN1/α)S_N/(\sigma_\alpha N^{1/\alpha}) converges in distribution to Zα,1Z_{\alpha,1}. Let mα+m_\alpha^+ denote the median of this centred, maximally right-skewed stable variable, Pr⁡(Zα,1≤mα+)=1/2\Pr(Z_{\alpha,1}\le m_\alpha^+)=1/2. Replacing the denominator by its limiting median estimate gives

Pwin(κ)≃2Fα(N1−1/α+(σα/μρ)mα+κsα)−1.P_{\mathrm{win}}(\kappa)\simeq2F_\alpha\left(\frac{N^{1-1/\alpha}+(\sigma_\alpha/\mu_\rho)m_\alpha^+}{\kappa s_\alpha}\right)-1.

This approximation accounts for the typical shift of the total resistance without fitted parameters. It omits fluctuations about the median, correlations between numerator and denominator, and higher-order mode deformation. Its accuracy is assessed against the joint stable approximation and resistor-chain ensembles in SM Figure S4.

Stable-law statistics of the residual for heavy-tailed random hoppings

Figure S4. Stable-law statistics of the residual for heavy-tailed random hoppings α=1.25\alpha= 1.25, compared with exact slow modes of the resistor chain. (a) Distribution of N1−1/αrN^{1-1/\alpha}r at L=1024L = 1024 (6000 samples) against the symmetric stable law (S23) (dashed) and with the finite-size correction (S24) (solid). (b) Fraction of samples in which a blind state starting κ=20\kappa= 20 (filled) or 50 (open) times farther away wins, from the 1500 samples per size of Figure 3. (c) Median reach, same data. No parameter is fitted.

Supplemental Material for

“Disorder-blind Mpemba acceleration of quantum state preparation”

Here we derive the results of the main text: first the model-independent framework (SM § 2.1), then the reduction of the spin chain to a chain of resistors (SM Sec. II), the rate (SM Sec. III), the resistance profile (SM Sec. IV), the deformed slow mode (SM § 2.5), the statistics of the resistance profile (SM Sec. VI), and interacting chains (SM Sec. VII). The preparation and measurement of the blind states are in SM Sec. VIII. The numerical checks can be found in SM Sec. IX.

What is known, derived, and observed. The rate formula (7) and the limit theorems behind Table I are standard [32, 43, 45–47]. The profile deviation u(x)−xu(x)-x is the classical corrector of 1D stochastic homogenisation; for finite variance its Brownian-bridge limit is known [55, 56], as are Gaussian corrections to spectral problems [57]. Our contributions are the reduction of the disorder-blind Mpemba effect to the residual and its reach, Eqs. (2) and (3); the exact linear inverse generator Eq. (9), which controls the slow-mode deformation even when links are arbitrarily weak; its projection onto a fixed blind preparation, Eq. (10); the heavy-tailed regime, whose non-Gaussian corrector separates the threshold α=2\alpha=2 for the preparation from α=1\alpha=1 for the rate and which we have not found treated in the homogenisation literature; and parameter-free success probabilities in both regimes, Eqs. (12) and (13). We also give exact preparation circuits for the blind and reference states, together with a protocol that measures rr (SM Sec. VIII). For interacting chains (SM Sec. VII), the reflection protection and the evenness of rr in VV at half filling are derived; the persistence of both scaling regimes is a numerical observation for L≤32L \le32, extended to L=4096L=4096 for heavy tails through a resistor chain with configuration-averaged rates.

TailResistancesRate ∣λ1∣|\lambda_1|Residual ∣r∣|r|
α>2\alpha>2finite varianceL−2L^{-2}L−1/2L^{-1/2}
α=2\alpha=2marginalL−2L^{-2}ln⁡L/L\sqrt{\ln L/L}
1<α<21<\alpha<2finite mean onlyL−2L^{-2}L1/α−1L^{1/\alpha-1}
α<1\alpha<1infinite meananomalousno claim

Table I. Typical scaling of the slowest rate and the blind residual for independent resistances with tail as in Eq. (11). The residual scalings establish a second threshold at α=2\alpha=2 within the diffusive regime. At α=2\alpha=2, the resistance variance diverges logarithmically. The finite-variance row also applies to bounded resistances and the short-range correlations generated by independent random fields.

Blind preparation, crossing and reach

Consider any Markovian dynamics ρ˙=Lρ\dot{\rho}=\mathcal{L}\rho with a unique stationary state; in our model this holds within each sector of fixed particle number, which the dynamics conserves. We assume that the generator is diagonalisable and has a simple, isolated slowest nonstationary eigenvalue λ1<0\lambda_1<0, with right and left modes satisfying LR1=λ1R1\mathcal{L}R_1=\lambda_1R_1 and L†L1=λ1L1\mathcal{L}^{\dagger}L_1=\lambda_1L_1, respectively, normalised by ⟨L1∣R1⟩=1\langle L_1|R_1\rangle=1. Any deviation from stationarity decays as δ(t)=∑k⟨Lk,δ⟩Rkeλkt\delta(t)=\sum_k\langle L_k,\delta\rangle R_k e^{\lambda_k t}, so at late times only ⟨L1,δ⟩R1eλ1t\langle L_1,\delta\rangle R_1e^{\lambda_1t} survives: the late-time amplitude of a state is its overlap with the left slow mode.

A blind preparation is chosen orthogonal to the slow mode of the clean system ⟨L1(0),δblind⟩=0\langle L_1^{(0)},\delta_{\mathrm{blind}}\rangle=0. In a disordered sample the true slow mode differs from L1(0)L_1^{(0)}, and the blind state keeps the overlap

r=⟨L1,δblind⟩⟨L1,δref⟩=⟨L1−L1(0),δblind⟩⟨L1,δref⟩.(16)r=\frac{\langle L_1,\delta_{\mathrm{blind}}\rangle}{\langle L_1,\delta_{\mathrm{ref}}\rangle}=\frac{\langle L_1-L_1^{(0)},\delta_{\mathrm{blind}}\rangle}{\langle L_1,\delta_{\mathrm{ref}}\rangle}. \tag*{(16)}

Here δref\delta_{\mathrm{ref}} and δblind\delta_{\mathrm{blind}} have the same initial measured distance from stationarity. We amplify the blind deviation by preparing

ρref(0)=ρss+δref,ρblind(κ)(0)=ρss+κδblind,\rho_{\mathrm{ref}}^{(0)}=\rho_{\mathrm{ss}}+\delta_{\mathrm{ref}},\qquad\rho_{\mathrm{blind}}^{(\kappa)}(0)=\rho_{\mathrm{ss}}+\kappa\delta_{\mathrm{blind}},

with κ≥1\kappa\ge1 chosen so that the states remain positive. κ>1\kappa>1 is chosen such that crossing indicates that the Mpemba effect takes place, though the Mpemba effect itself does not rely on κ\kappa.

Since DD is a norm of the measured deviation, κ=Dblind(0)/Dref(0)\kappa=D_{\mathrm{blind}}(0)/D_{\mathrm{ref}}(0). Explicit positive preparations are given in Sec. VIII. When the slow mode contributes to the readout

Dblind(t)Dref(t)⟶κ∣r∣.(17)\frac{D_{\mathrm{blind}}(t)}{D_{\mathrm{ref}}(t)}\longrightarrow\kappa|r|. \tag*{(17)}

Thus 1<κ<1/∣r∣1<\kappa<1/|r| guarantees a crossing: the blind preparation starts farther away but becomes closer at late times. For r≠0r\ne0, the late-time distances are Dref(t)≃Ae−∣λ1∣tD_{\mathrm{ref}}(t)\simeq Ae^{-|\lambda_1|t} and Dblind(t)≃κ∣r∣Ae−∣λ1∣tD_{\mathrm{blind}}(t)\simeq\kappa|r|Ae^{-|\lambda_1|t}, with A>0A>0. Let treft_{\mathrm{ref}} and tblindt_{\mathrm{blind}} be the times after which these distances remain below a common tolerance DtolD_{\mathrm{tol}}. Solving the two exponential decays gives

tref−tblind⟶ln⁡[1/(κ∣r∣)]∣λ1∣(Dtol→0).(18)t_{\mathrm{ref}}-t_{\mathrm{blind}}\longrightarrow\frac{\ln[1/(\kappa|r|)]}{|\lambda_1|}\qquad(D_{\mathrm{tol}}\to0). \tag*{(18)}

The logarithm converts the reduction in slow-mode amplitude into a time advantage. For equal initial distances κ=1\kappa=1, this reduces to Δt=ln⁡(1/∣r∣)/∣λ1∣\Delta t=\ln(1/|r|)/|\lambda_1|. If r=0r=0, the blind preparation instead decays through a faster mode, and the two decay rates must be treated separately. In our model the left and right slow density profiles coincide (SM Sec. II), so we simply write ϕ1\phi_1 for the slow mode, and use the distance D=[L−1∑iδni2]1/2D=[L^{-1}\sum_i\delta n_i^2]^{1/2}.

From the spin chain to a chain of resistors

Exact equation for the densities. The chain is H=∑iJi(σixσi+1x+σiyσi+1y)+∑ihiσizH=\sum_i J_i(\sigma_i^x\sigma_{i+1}^x+\sigma_i^y\sigma_{i+1}^y)+\sum_i h_i\sigma_i^z with dephasing γ σiz\sqrt{\gamma}\,\sigma_i^z on every site. The Jordan-Wigner transformation turns the hopping term into free fermions σixσi+1x+σiyσi+1y=2(ci†ci+1+ci+1†ci)\sigma_i^x\sigma_{i+1}^x+\sigma_i^y\sigma_{i+1}^y=2(c_i^\dagger c_{i+1}+c_{i+1}^\dagger c_i), with σiz=2ni−1\sigma_i^z=2n_i-1. For this Hermitian jump operator, the real constant shift does not change the dissipator, so dephasing acts as 4γ ni\sqrt{4\gamma}\,n_i: it leaves the densities alone and damps every coherence ci†cjc_i^\dagger c_j (i≠ji\ne j) at rate 4γ4\gamma. Since commutators of fermion bilinears are again bilinears, the two-point matrix Mij=⟨cj†ci⟩M_{ij}=\langle c_j^\dagger c_i\rangle obeys a closed equation [1]

M˙=−i[h,M]−4γ(M−diag⁡M),hii=2hi,hi,i+1=hi+1,i=2Ji.(19)\dot{M}=-i[h,M]-4\gamma(M-\operatorname{diag}M),\qquad h_{ii}=2h_i,\quad h_{i,i+1}=h_{i+1,i}=2J_i. \tag*{(19)}

This is exact and evolves L2L^2 correlation-matrix elements rather than the 4L4^L entries of the full many-body density matrix. Because hh is real, transposing MM maps the generator onto its adjoint, so the left and right slow modes have the same density profile.

Late times: densities only. The diagonal of Eq. (19) gives a continuity equation, n˙i=Ji−1−Ji\dot{n}_i=\mathcal{J}_{i-1}-\mathcal{J}_i, with the current Ji=4JiIm⁡Mi,i+1\mathcal{J}_i=4J_i\operatorname{Im}M_{i,i+1} from site ii to i+1i+1. The coherence that carries this current obeys

M˙i,i+1=−2i(hi−hi+1)Mi,i+1−2iJi(ni+1−ni)−4γMi,i+1+…(20)\dot{M}_{i,i+1}=-2i(h_i-h_{i+1})M_{i,i+1}-2iJ_i(n_{i+1}-n_i)-4\gamma M_{i,i+1}+\ldots \tag*{(20)}

where the dots couple to coherences between more distant sites. In the diffusive regime, the slow density-relaxation rates vanish as L−2L^{-2} and become small compared with 4γ4\gamma. A local nearest-neighbor hopping equation follows at strong dephasing, where couplings to longer-range coherences can be neglected. Setting M˙i,i+1=0\dot{M}_{i,i+1}=0 then gives Mi,i+1=−2iJi(ni+1−ni)/[4γ+2i(hi−hi+1)]M_{i,i+1}=-2iJ_i(n_{i+1}-n_i)/[4\gamma+2i(h_i-h_{i+1})], and taking the imaginary part

Ji=Gi(ni−ni+1),Gi=8Ji2γ4γ2+(hi−hi+1)2.(21)\mathcal{J}_i=G_i(n_i-n_{i+1}),\qquad G_i=\frac{8J_i^2\gamma}{4\gamma^2+(h_i-h_{i+1})^2}. \tag*{(21)}

The densities hop classically n˙i=Gi−1(ni−1−ni)+Gi(ni+1−ni)\dot{n}_i=G_{i-1}(n_{i-1}-n_i)+G_i(n_{i+1}-n_i), with no flux through the ends. A bond is a weak link when its hopping is small or when the energies on its two sides differ strongly. These are exactly the bond rates of the effective late-time description used in Ref. [2] (their Eq. (15) for the non-interacting chain). At finite γ\gamma, longer-range coherences modify this local approximation. The analytical results below apply to the resistor chain defined by these bond rates; their persistence at finite dephasing is assessed using the exact correlation matrix dynamics below.

Resistors in series. Write ρi=1/Gi\rho_i=1/G_i. In a steady flow the same current crosses every bond, so the density drops across bond ii by Jρi\mathcal{J}\rho_i, and the total drop is J∑iρi\mathcal{J}\sum_i\rho_i. Resistances add in series; a single nearly blocked bond can dominate the total.

The relaxation rate

For a uniform chain, the decaying modes are cosines. With x=(i−12)/Lx=(i-\frac{1}{2})/L one checks directly that ni=cos⁡kπxn_i=\cos k\pi x solves the hopping equation with no-flux ends, with λk=−4Gsin⁡2(kπ/2L)≃−π2k2G/L2\lambda_k=-4G\sin^2(k\pi/2L)\simeq-\pi^2k^2G/L^2. The slowest is the single arc cos⁡πx\cos\pi x.

For a disordered chain, label the N=L−1N=L-1 bonds yi=i/Ny_i=i/N and define W(x)=N−1∑yi≤xρiW(x)=N^{-1}\sum_{y_i\le x}\rho_i. The lattice equations are, exactly, n˙i=Ji−1−Ji\dot{n}_i=\mathcal{J}_{i-1}-\mathcal{J}_i and ni+1−ni=−ρiJin_{i+1}-n_i=-\rho_i\mathcal{J}_i: the density drops across a bond by current times resistance. Across a large resistance the density therefore jumps, so in a slow mode it is the current that varies smoothly. Since bond ii adds ρi/N\rho_i/N to WW, the two equations read ∂xW n=−NJ\partial_xW\,n=-N\mathcal{J} and ∂tn=−N−1∂xJ\partial_tn=-N^{-1}\partial_x\mathcal{J}, so that n˙=N−2KWn\dot{n}=N^{-2}K_Wn with

KW=∂x∂W,∂W=1W′(x)∂x.(22)K_W=\partial_x\partial_W,\qquad\partial_W=\frac{1}{W'(x)}\partial_x. \tag*{(22)}

and no flux at the ends. A uniform chain with resistance ρˉN=N−1∑iρi\bar{\rho}_N=N^{-1}\sum_i\rho_i per bond has W=ρˉNxW=\bar{\rho}_Nx and slowest rate λ1=−π2/(N2ρˉN)\lambda_1=-\pi^2/(N^2\bar{\rho}_N). Disorder changes this only at relative order ω\omega (SM Sec. IV), which vanishes for long chains as long as the mean resistance μρ\mu_\rho is finite [3, 4]. The rate thereby depends only on the total resistance

λ1≃−π2N2ρˉN,ρˉN→μρ.(23)\lambda_1\simeq-\frac{\pi^2}{N^2\bar{\rho}_N},\qquad\bar{\rho}_N\to\mu_\rho. \tag*{(23)}

Equivalently, KWK_W is self-adjoint with −⟨f,KWf⟩=∫01(f′)2/W′ dx-\langle f,K_Wf\rangle=\int_0^1(f')^2/W'\,dx, so N2∣λ1∣N^2|\lambda_1| is the minimum of −⟨f,KWf⟩/⟨f,f⟩-\langle f,K_Wf\rangle/\langle f,f\rangle over mean-zero ff. On large scales, 1/W′1/W' is replaced by the harmonic mean of the local conductances 1/μρ1/\mu_\rho, which is why only the total resistance enters.

The resistance profile and the inverse generator

Define the resistance profile u(x)=W(x)/W(1)u(x)=W(x)/W(1), which is the fraction of the total resistance to the left of xx. It is a straight line u(x)=xu(x)=x in a clean chain, and u(1)=1u(1)=1 in every sample. Its deviation from a straight line u(x)−xu(x)-x vanishes at both ends, and we denote its largest value ω=sup⁡x∣u(x)−x∣\omega=\sup_x|u(x)-x|.

Why we work with the inverse generator. When a few resistances are huge, the generator KWK_W is not close to the clean one, so ordinary perturbation theory in KWK_W fails. Its inverse, however, is simple. To solve KWn=fK_W n=f for a mean-zero source ff, integrate once: with no flux at x=0x=0, ∂Wn=Ff(x)≡∫0xf\partial_W n=F_f(x)\equiv\int_0^x f, i.e., dn=Ff dWdn=F_f\,dW. Physically, the current through each bond is fixed by the source to its left, and the density drops by current times resistance. For a second mean-zero function gg with Fg(x)=∫0xgF_g(x)=\int_0^x g, an integration by parts then gives

⟨g,KW−1f⟩=−∫01Ff(x)Fg(x) dW(x).(24)\langle g,K_W^{-1}f\rangle=-\int_0^1 F_f(x)F_g(x)\,dW(x). \tag*{(24)}

The inverse generator is linear in the cumulative resistance. The same identity holds on the lattice, bond by bond ⟨g,K−1f⟩=−∑iFf,iFg,iρi\langle g,K^{-1}f\rangle=-\sum_i F_{f,i}F_{g,i}\rho_i. Writing W=ρˉNx+ΔWW=\bar{\rho}_N x+\Delta W, the change of the inverse generator caused by disorder is ∫01ΔW(fFg+Ffg) dx\int_0^1\Delta W(fF_g+F_f g)\,dx, which is bounded by 2πρˉNω∥f∥∥g∥\frac{2}{\pi}\bar{\rho}_N\omega\lVert f\rVert\lVert g\rVert (using ∥Ff∥≤∥f∥/π\lVert F_f\rVert\le\lVert f\rVert/\pi). So it is small whenever the profile is close to a straight line, however large individual resistances are.

The deformed slow mode

For the uniform chain, the inverse generator has eigenfunctions 2cos⁡kπx\sqrt{2}\cos k\pi x with eigenvalues μk=−ρˉN/(kπ)2\mu_k=-\bar{\rho}_N/(k\pi)^2. The slow mode corresponds to the most negative one μ1\mu_1, well separated from the rest. Standard first-order perturbation theory [5] gives the deformed slow mode as

ϕ1=cos⁡πx+∑k≥2ckcos⁡kπx+O(ω2),ck=⟨ek,δK−1e1⟩μ1−μk.(25)\phi_1=\cos\pi x+\sum_{k\ge2}c_k\cos k\pi x+O(\omega^2),\qquad c_k=\frac{\langle e_k,\delta K^{-1}e_1\rangle}{\mu_1-\mu_k}. \tag*{(25)}

The numerator follows from Eq. (24), applied to WW and to ρˉNx\bar{\rho}_N x and subtracted, with Fe1=∫0xe1=2sin⁡πx/πF_{e_1}=\int_0^x e_1=\sqrt{2}\sin\pi x/\pi and Fek=∫0xek=2sin⁡kπx/(kπ)F_{e_k}=\int_0^x e_k=\sqrt{2}\sin k\pi x/(k\pi), which gives −2kπ2∫01sin⁡πxsin⁡kπx dΔW-\frac{2}{k\pi^2}\int_0^1\sin\pi x\sin k\pi x\,d\Delta W. The denominator is −ρˉNπ2k2−1k2-\frac{\bar{\rho}_N}{\pi^2}\frac{k^2-1}{k^2}. Their ratio, with dΔW/ρˉN=d[u−x]d\Delta W/\bar{\rho}_N=d[u-x], is

ck=2kk2−1∫01sin⁡kπxsin⁡πx d[u(x)−x].(26)c_k=\frac{2k}{k^2-1}\int_0^1\sin k\pi x\sin\pi x\,d[u(x)-x]. \tag*{(26)}

The residual. For a blind profile cos⁡(2πx)\cos(2\pi x) with the same norm as the reference cos⁡(πx)\cos(\pi x), the residual selects the k=2k=2 coefficient. With the normalisation used here, the reference overlap is fixed, so r=c2r=c_2. Hence

r=43∫01sin⁡πxsin⁡2πx d[u(x)−x]+O(ω2).(27)r=\frac{4}{3}\int_0^1\sin\pi x\sin2\pi x\,d[u(x)-x]+O(\omega^2). \tag*{(27)}

Integrating by parts, r=−∫01g′(x)[u(x)−x] dx+O(ω2)r=-\int_0^1g'(x)[u(x)-x]\,dx+O(\omega^2) with g=43sin⁡πxsin⁡2πxg=\frac{4}{3}\sin\pi x\sin2\pi x: the residual is a fixed weighted average of the profile deviation. The size of the whole deformation d=∥ϕ1−cos⁡πx∥/∥cos⁡πx∥d=\lVert\phi_1-\cos\pi x\rVert/\lVert\cos\pi x\rVert is likewise of order ω\omega. Figure S1 shows the profile and the deformation in real space.

Figure S1. What the blind state sees. Top: deviation of the resistance profile from a straight line u(x)−xu(x)-x. Bottom: deformation of the slow mode ϕ1−cos⁡πx\phi_1-\cos\pi x; three samples each, L=2048L=2048. (a) Bounded disorder: many bonds contribute and the profile wiggles like a random walk pinned at both ends. (b) Heavy-tailed disorder (α=1.25\alpha=1.25): the profile jumps at the weakest links, and the slow mode steps at the same places.

A single weak link. Suppose one bond at x0x_0 carries a share sws_w of the total resistance and the rest is uniform. Then u(x)−x=sw[Θ(x−x0)−x]u(x)-x=s_w[\Theta(x-x_0)-x], so d[u−x]=sw[δ(x−x0)−1] dxd[u-x]=s_w[\delta(x-x_0)-1]\,dx, and since ∫01g=0\int_0^1g=0, then

r≃swg(x0)=43swsin⁡πx0sin⁡2πx0.(28)r\simeq s_w g(x_0)=\frac{4}{3}s_w\sin\pi x_0\sin2\pi x_0. \tag*{(28)}

The weight vanishes at the centre and at both ends and is largest where tan⁡2πx0=2\tan^2\pi x_0=2, i.e., x0≃0.304x_0\simeq0.304 or 0.6960.696, where ∣g∣=16/(93)≃1.03|g|=16/(9\sqrt{3})\simeq1.03. This is first order in sws_w and thus quantitative only for small shares. It shows that in the heavy-tailed regime the residual of a sample is set by the share and position of its few largest resistances.

Other blind profiles. Any smooth blind profile ∑k≥2bkcos⁡kπx\sum_{k\ge2}b_k\cos k\pi x with ∑kbk2=1\sum_k b_k^2=1 gives r=∑kbkckr=\sum_k b_kc_k instead. The prefactor 2k/(k2−1)=43,34,815,…2k/(k^2-1)=\frac{4}{3},\frac{3}{4},\frac{8}{15},\ldots makes cos⁡2πx\cos2\pi x the most strongly coupled single mode. For finite variance (SM Sec. VI), Nr\sqrt{N}r is Gaussian with standard deviation s[b]=(σ/μρ)bTMbs[b]=(\sigma/\mu_\rho)\sqrt{b^\mathsf{T}Mb}, where Mkl=βkβl∫01sin⁡2πxsin⁡kπxsin⁡lπx dxM_{kl}=\beta_k\beta_l\int_0^1\sin^2\pi x\sin k\pi x\sin l\pi x\,dx and βk=2k/(k2−1)\beta_k=2k/(k^2-1). The disorder enters only through σ/μρ\sigma/\mu_\rho, so all finite-variance distributions rank blind profiles identically. The largest eigenvalue of MM gives the least favourable smooth profile 0.975cos⁡2πx−0.221cos⁡4πx0.975\cos2\pi x-0.221\cos4\pi x, with s[b]=0.682 σ/μρs[b]=0.682\,\sigma/\mu_{\rho}, only 2% above cos⁡2πx\cos2\pi x. Rough profiles behave differently: a random blind state orthogonal to cos⁡πx\cos\pi x, spread over all ∼N\sim N lattice modes, catches only a 1/N1/\sqrt{N} share of the deformation, and its residual falls as L−1L^{-1} [Figure S2]. The L−1/2L^{-1/2} statements in the main text therefore refer to smooth blind profiles, which are also the experimentally natural ones.

How uneven is the profile?

Since rr is a fixed weighted average of u(x)−xu(x)-x, its size is set by how much the profile fluctuates. Defining ΔW(x)=W(x)−xW(1)\Delta W(x)=W(x)-xW(1) and writing Sm=∑i≤m(ρi−μρ)S_m=\sum_{i\le m}(\rho_i-\mu_\rho) for the accumulated excess resistance, the profile deviation is the walk minus its straight-line interpolation

ΔW(x)=S⌊xN⌋−xSNN+O(N−1).(29)\Delta W(x)=\frac{S_{\lfloor xN\rfloor}-xS_N}{N}+O(N^{-1}). \tag*{(29)}

Finite variance: many bonds. Suppose the resistances have a finite variance and correlations of finite range, and let σ2=Var⁡ρi+2∑m≥1Cov⁡(ρi,ρi+m)\sigma^2=\operatorname{Var}\rho_i+2\sum_{m\ge1}\operatorname{Cov}(\rho_i,\rho_{i+m}) be the variance of their sum per bond. Then SmS_m is a random walk, which on large scales becomes Brownian motion with SxN≃σNB(x)S_{xN}\simeq\sigma\sqrt{N}B(x) [6]. By Eq. (29), u(x)−x≃(σ/μρN)B∘(x)u(x)-x\simeq(\sigma/\mu_\rho\sqrt{N})B^\circ(x), where B∘(x)=B(x)−xB(1)B^\circ(x)=B(x)-xB(1) is a Brownian bridge (this is the classical 1D homogenisation corrector [7, 8]). So ω∼L−1/2\omega\sim L^{-1/2}, and the rescaled ω\omega follows Kolmogorov’s distribution Figure S3,Figure S3). Inserting into Eq. (27)

Figure S3. The two regimes. (a) The slow-mode distortion dd tracks the largest profile deviation ω\omega sample by sample (random positions with bounded fields, ξ/ℓ=6,3,2.5\xi/\ell=6, 3, 2.5, i.e., α=3,1.5,1.25\alpha=3, 1.5, 1.25; circles L=256L=256, squares L=8192L=8192; line d=ωd=\omega). (b) For independent resistances uniform in [1,2][1,2] at L=4096L=4096, the rescaled ω\omega follows Kolmogorov’s distribution. (c) At α=1.25\alpha=1.25, ω\omega is set by the share of the largest resistance; at α=3\alpha=3 it is larger, because many bonds contribute.

N r≃σμρ∫01g dB∘=σμρ∫01(g−gˉ) dB,(30)\sqrt{N}\,r\simeq\frac{\sigma}{\mu_\rho}\int_0^1g\,\mathrm{d}B^\circ=\frac{\sigma}{\mu_\rho}\int_0^1(g-\bar{g})\,\mathrm{d}B, \tag*{(30)}

a Gaussian with variance (σ/μρ)2[∫01g2−gˉ2](\sigma/\mu_\rho)^2[\int_0^1g^2-\bar{g}^2], where gˉ=∫01g\bar{g}=\int_0^1g. For g=43sin⁡πxsin⁡2πxg=\frac{4}{3}\sin\pi x\sin2\pi x, ∫01g2=169⋅14=49\int_0^1g^2=\frac{16}{9}\cdot\frac{1}{4}=\frac{4}{9} and gˉ=0\bar{g}=0, so

N r→dN(0,s2),s=23σμρ.(31)\sqrt{N}\,r\xrightarrow{\mathrm{d}}\mathcal{N}(0,s^2),\qquad s=\frac{2}{3}\frac{\sigma}{\mu_\rho}. \tag*{(31)}

Two predictions follow without any free parameters: the median residual is 0.674 s/N0.674\,s/\sqrt{N}, and a blind state starting κ\kappa times farther away wins when ∣r∣<1/κ|r|<1/\kappa (SM Sec. I), which happens with probability

Pwin≃erf⁡(N2 κs).(32)P_{\mathrm{win}}\simeq\operatorname{erf}\left(\frac{\sqrt{N}}{\sqrt{2}\,\kappa s}\right). \tag*{(32)}

Infinite variance: a few weak links. Now suppose the resistances are independent with a tail P(ρ>t)∼t−αP(\rho>t)\sim t^{-\alpha}, where 1<α<21<\alpha<2; the mean is finite but the variance is not. Among NN bonds the largest resistance is of order tt with NP(ρ>t)∼1NP(\rho>t)\sim1, i.e., N1/αN^{1/\alpha}, while the total is μρN\mu_\rho N. So each of the weakest links carries a share of order N1/α−1N^{1/\alpha-1} of the total resistance, much more than the N−1/2N^{-1/2} wiggle of many bonds Figure S3. The accumulated excess SmS_m is then dominated by these jumps and grows as N1/αN^{1/\alpha} (the generalised central limit theorem [9]), so by Eq. (29)

ω, ∣r∣∼L1/α−1.(33)\omega,\ |r|\sim L^{1/\alpha-1}. \tag*{(33)}

The profile jumps at the weakest links, and since the density drops across a bond by current times resistance, the slow mode steps at the same places Figure S1. The residual still vanishes for every α>1\alpha> 1, so blind preparation works whenever the mean resistance is finite, but reaches its full L−1/2L^{-1/2} strength only when the variance is finite too. At α=2\alpha= 2, for a pure t−2t^{-2} tail, the variance of the truncated sum grows as Nlog⁡NN \log N, so ∣r∣∼log⁡L/L|r| \sim\sqrt{\log L/L}. For α<1\alpha< 1 the mean resistance itself diverges, the chain no longer homogenises. This assumes independent resistances; correlated heavy-tailed disorder is tested numerically.

Infinite variance: the distribution of the residual. Let the resistances be independent with P(ρi>t)≃Aρt−αP(\rho_i > t) \simeq A_\rho t^{-\alpha}, where 1<α<21 < \alpha< 2. On the lattice, Eq. (27) becomes a ratio of two sums. The profile uu jumps by ρi/∑jρj\rho_i/\sum_j \rho_j across bond ii, so ∫01g du=∑igiρi/∑iρi\int_0^1 g\,\mathrm{d}u = \sum_i g_i\rho_i/\sum_i \rho_i, while ∫01g dx\int_0^1 g\,\mathrm{d}x becomes N−1∑igiN^{-1}\sum_i g_i. Hence

r=∑igiρi∑iρi−1N∑igi,gi=g(xi).(34)r = \frac{\sum_i g_i\rho_i}{\sum_i \rho_i} - \frac{1}{N}\sum_i g_i,\qquad g_i = g(x_i). \tag*{(34)}

The last term is O(N−2)O(N^{-2}), because ∫01g=0\int_0^1 g = 0 and gg vanishes at both ends, so that ∑igi=O(N−1)\sum_i g_i = O(N^{-1}). To expose the fluctuations, write each resistance as its mean plus an excess ρi=μρ+(ρi−μρ)\rho_i = \mu_\rho+ (\rho_i - \mu_\rho), and define

TN=∑igi(ρi−μρ),SN=∑i(ρi−μρ).(35)T_N = \sum_i g_i(\rho_i - \mu_\rho),\qquad S_N = \sum_i(\rho_i - \mu_\rho). \tag*{(35)}

TNT_N is the excess resistance weighted by the blind kernel, and SNS_N is the excess of the total resistance. Then ∑igiρi=TN+μρ∑igi\sum_i g_i\rho_i = T_N + \mu_\rho\sum_i g_i and ∑iρi=Nμρ+SN\sum_i \rho_i = N\mu_\rho+ S_N. Dropping terms smaller by a factor N−1N^{-1}

r≃TNNμρ+SN=TN/Nμρ+SN/N.(36)r \simeq\frac{T_N}{N\mu_\rho+ S_N} = \frac{T_N/N}{\mu_\rho+ S_N/N}. \tag*{(36)}

For heavy tails both sums are dominated by the largest resistances and grow as N1/αN^{1/\alpha} rather than N1/2N^{1/2}. Writing each on this natural scale TN/N=N1/α−1(TN/N1/α)T_N/N = N^{1/\alpha-1}(T_N/N^{1/\alpha}) and likewise for SNS_N, gives

r≃N1/α−1TN/N1/αμρ+N1/α−1SN/N1/α.(37)r \simeq\frac{N^{1/\alpha-1}T_N/N^{1/\alpha}}{\mu_\rho+ N^{1/\alpha-1}S_N/N^{1/\alpha}}. \tag*{(37)}

Here TN/N1/αT_N/N^{1/\alpha} and SN/N1/αS_N/N^{1/\alpha} are random variables of order one, whose stable limits are derived next. The prefactor N1/α−1N^{1/\alpha-1} sets the size of rr. Because the same prefactor multiplies SNS_N, the denominator fluctuates at the same relative order as rr itself; this is the origin of the finite-size correction below. By the generalised central limit theorem for weighted sums [9, 10], (TN,SN)/N1/α(T_N,S_N)/N^{1/\alpha} converges jointly to (∫01g dZ,Z(1))(\int_0^1 g\,\mathrm{d}Z,Z(1)), where ZZ is a totally skewed α\alpha-stable Lévy process whose increments over unit length have scale σα=[AρΓ(1−α)cos⁡(πα/2)]1/α\sigma_\alpha= [A_\rho\Gamma(1-\alpha)\cos(\pi\alpha/2)]^{1/\alpha} (positive, since both factors are negative). For deterministic gg, ∫01g dZ\int_0^1 g\,\mathrm{d}Z is α\alpha-stable with scale σα∥g∥α\sigma_\alpha\lVert g\rVert_\alpha, where ∥g∥α=(∫01∣g∣α dx)1/α\lVert g\rVert_\alpha= (\int_0^1 |g|^\alpha\,\mathrm{d}x)^{1/\alpha}, and skewness βg=∫01∣g∣αsgn⁡g dx/∥g∥αα\beta_g = \int_0^1 |g|^\alpha\operatorname{sgn}g\,\mathrm{d}x/\lVert g\rVert_\alpha^\alpha [10]. Since g(1−x)=−g(x)g(1-x)=-g(x), βg=0\beta_g=0 exactly: although every ρi\rho_i is strongly skewed, the residual is symmetric. To leading order the denominator of Eq. (37) is μρ\mu_\rho, so

N1−1/αr→dsαZα,sα=σα∥g∥αμρ.(38)N^{1-1/\alpha}r \xrightarrow{\mathrm{d}} s_\alpha Z_\alpha,\qquad s_\alpha= \frac{\sigma_\alpha\lVert g\rVert_\alpha}{\mu_\rho}. \tag*{(38)}

with ZαZ_\alpha standard symmetric α\alpha-stable, EeiθZα=e−∣θ∣α\mathbb{E}e^{\mathrm{i}\theta Z_\alpha}=e^{-|\theta|^\alpha}. With FαF_\alpha its distribution function and mαm_\alpha the median of ∣Zα∣|Z_\alpha|, a blind state starting κ\kappa times farther away wins with probability Pwin=2Fα(zκ)−1P_{\mathrm{win}}=2F_\alpha(z_\kappa)-1, where zκ=N1−1/α/(κsα)z_\kappa=N^{1-1/\alpha}/(\kappa s_\alpha), and the median reach is asymptotic z1/mαz_1/m_\alpha. For random hoppings, ρi=(γ/2)νi−2\rho_i=(\gamma/2)\nu_i^{-2} is exactly Pareto distributed, with Aρ=(γ/2)αA_\rho=(\gamma/2)^\alpha and μρ=(γ/2)α/(α−1)\mu_\rho=(\gamma/2)\alpha/(\alpha-1), so sαs_\alpha depends on α\alpha alone. For α=1.25\alpha=1.25, σα/μρ=0.331\sigma_\alpha/\mu_\rho=0.331, ∥g∥α=0.597\lVert g\rVert_\alpha=0.597, sα=0.198s_\alpha=0.198, and mα=0.979m_\alpha=0.979.

Finite-size approximation. Eq. (38) is approached slowly. The relative fluctuation of the denominator in Eq. (37) is of order N1/α−1N^{1/\alpha-1}, the same order as rr, and it is totally skewed: in a typical sample the total resistance lies below NμρN\mu_\rho, because the mean is carried by rare weak links, so typical residuals are larger than the leading order suggests. Replacing SN/N1/αS_N/N^{1/\alpha} by its median sαmα+s_\alpha m_\alpha^+, where mα+m_\alpha^+ is the median of the standard totally skewed stable law (m1.25+=−1.996)(m_{1.25}^+=-1.996), gives

zκ=μρN1−1/α+σαmα+κσα∥g∥α.(39)z_\kappa= \frac{\mu_\rho N^{1-1/\alpha} + \sigma_\alpha m_\alpha^+}{\kappa\sigma_\alpha\lVert g\rVert_\alpha}. \tag*{(39)}

For the cases tested, sampling the joint stable law of the numerator and denominator changes PwinP_{\mathrm{win}} by at most 0.01 relative to the median approximation. This checks the treatment of the fluctuating denominator within the first-order residual formula. Figure S4 compares both forms with exact slow modes of the resistor chain (the 1500 samples per size of Figure 3): with the correction, PwinP_{\mathrm{win}} agrees within 0.022 (κ=20\kappa= 20) and 0.008 (κ=50\kappa= 50) for 256≤L≤4096256 \le L \le4096, and the median reach within 9% at L=256L = 256 and 3% for 512≤L≤4096512 \le L \le4096. The leading order alone is 0.02-0.08 too high over 64≤L≤409664 \le L \le4096.

Rare failures. The stable limit describes the distribution of typical rescaled residuals. At fixed κ\kappa, failure probes increasingly remote tails of this distribution as the chain grows, so convergence in distribution alone does not determine the asymptotic failure probability. Samples containing a weak link with an appreciable share of the total resistance can also require higher-order mode deformation. We therefore assess the finite-size success predictions against the resistor-chain ensembles in Figure S4.

Where the tails come from. Random fields hih_i uniform in [−h0,h0][-h_0,h_0] with Ji=JJ_i = J give ρi=[4γ2+Di2]/(8J2γ)\rho_i = [4\gamma^2 + D_i^2]/(8J^2\gamma), where Di=hi−hi+1D_i = h_i - h_{i+1}: bounded resistances, so α=∞\alpha= \infty. Neighbouring resistances share a field, so the covariance term in σ2\sigma^2 matters. Using Eh2=h02/3\mathbb{E}h^2 = h_0^2/3 and Eh4=h04/5\mathbb{E}h^4 = h_0^4/5, one finds Var⁡Di2=28h04/45\operatorname{Var}D_i^2 = 28h_0^4/45 and Cov⁡(Di2,Di+12)=4h04/45\operatorname{Cov}(D_i^2,D_{i+1}^2) = 4h_0^4/45, so that σ2=36h04/(8J2γ)2\sigma^2 = 36h_0^4/(8J^2\gamma)^2. For h0=2h_0 = 2, γ=J=1\gamma= J = 1 this gives μρ=5/6\mu_\rho= 5/6, σ2=0.2\sigma^2 = 0.2 and s=0.358s = 0.358. Random hoppings Ji=vizJ_i = v_i^z, with viv_i uniform on (0,1](0,1] and hi=0h_i = 0, give ρi∝vi−2z\rho_i \propto v_i^{-2z}, hence P(ρ>t)∝t−1/2zP(\rho> t) \propto t^{-1/2z} and α=1/2z\alpha= 1/2z. For independent exponentially distributed spacings ℓi\ell_i of mean ℓ\ell, tunnelling Ji=J0e−ℓi/ξJ_i = J_0e^{-\ell_i/\xi} at uniform fields gives ρi=ρ0e2ℓi/ξ\rho_i = \rho_0e^{2\ell_i/\xi}, where ρ0=γ/(2J02)\rho_0 = \gamma/(2J_0^2). Thus Pr⁡(ρi>t)=(ρ0/t)α\Pr(\rho_i > t) = (\rho_0/t)^\alpha for t≥ρ0t \ge\rho_0, with α=ξ/(2ℓ)\alpha= \xi/(2\ell). Independent bounded random fields multiply these resistances by a bounded positive factor and preserve the tail exponent, but introduce correlations between neighboring bonds. The distributional results above are derived for independent heavy-tailed resistances.

Interacting chains

Effective dynamics. Add V∑iσizσi+1zV\sum_i \sigma_i^z\sigma_{i+1}^z to HH. The density equation no longer closes, but the separation of timescales of Sec. II still applies to the many-body populations. A hop across bond ii connects two configurations that differ on two sites, so the coherence between them decays at rate 4γ4\gamma and oscillates at their energy difference ΔEi\Delta E_i, which now includes the change of the interaction energy with the two outer neighbours. Eliminating this coherence at strong dephasing gives a classical exclusion process with the rate of Eq. (15), Gi=32Ji2γ/(16γ2+ΔEi2)G_i = 32J_i^2\gamma/(16\gamma^2+\Delta E_i^2), where ΔEi=2(hi+1−hi)+2V(σi+2z−σi−1z)\Delta E_i = 2(h_{i+1}-h_i)+2V(\sigma_{i+2}^z-\sigma_{i-1}^z) and σz≡0\sigma^z \equiv0 beyond the ends. The rate depends on ΔEi2\Delta E_i^2 and is therefore the same for a hop and its reverse. The generator is then symmetric, and the uniform distribution π\pi on configurations with fixed particle number is stationary for every VV and every sample. This is the infinite-temperature state of the dephased chain, and it carries no information about the disorder. At V=0V = 0 the rates do not depend on the configuration and the densities obey the resistor chain of Sec. II.

The residual. The left slow mode is a function f1(η)f_1(\eta) (the left eigenvector of the lattice-gas generator with the slowest nonzero rate; with interactions the density equation does not close, so the slow mode is a function of the whole configuration rather than a density profile) on configurations. A density modulation bib_i prepared in local equilibrium, P=π[1+ϵ∑ibi(ηi−ν)]P = \pi[1+\epsilon\sum_i b_i(\eta_i-\nu)] with filling ν\nu, has slow-mode amplitude ϵ∑ibimi\epsilon\sum_i b_im_i with mi=Eπ[f1(η)ηi]m_i = \mathbb{E}_\pi[f_1(\eta)\eta_i]. The density profile mim_i of the slow mode therefore plays the role of ϕ1\phi_1, and

r=∑icos⁡(2πxi)mi∑icos⁡(πxi)mi.(40)r = \frac{\sum_i \cos(2\pi x_i)m_i}{\sum_i \cos(\pi x_i)m_i}. \tag*{(40)}

The overlap is ∑ηf1(η)P(η)=ϵ∑ibiEπ[f1(ηi−ν)]=ϵ∑ibimi\sum_\eta f_1(\eta)P(\eta)=\epsilon\sum_i b_i\mathbb{E}_\pi[f_1(\eta_i-\nu)]=\epsilon\sum_i b_im_i, because Eπf1=0\mathbb{E}_\pi f_1=0 (the slow mode is orthogonal to the stationary state). The residual is the ratio of these amplitudes for bi=cos⁡(2πxi)b_i=\cos(2\pi x_i) and cos⁡(πxi)\cos(\pi x_i) at equal ϵ\epsilon. Such a state is the linear-response member of the maximum-entropy family P∝πe∑ibiηiP\propto\pi e^{\sum_i b_i\eta_i} at fixed particle number, which can also be prepared exactly by quantum circuits. SM Sec. VIII gives these circuits, and its affine states make the linear-response form exact.

Symmetries and reflection. If the generator and π\pi are invariant under i→L+1−ii\to L+1-i and f1f_1 is odd, then Eπ[f1P]=0\mathbb{E}_\pi[f_1P]=0 for every reflection-even PP, at any amplitude and any VV. In clean chains we find ∣r∣<10−13|r|<10^{-13} for all VV studied. Interaction sign. At half filling, particle-hole conjugation ηi→1−ηi\eta_i\to1-\eta_i reverses every σz\sigma^z and every hop, so it maps the rates at VV onto those at −V-V. It also reverses every zero-sum density deviation, and with it the slow-mode profile, mi→−mim_i\to-m_i. Hence r(V)=r(−V)r(V)=r(-V) in every sample, which we confirm to 10−1410^{-14} for L=10L=10-1414. The same holds for the full Lindblad generator at any γ\gamma: the map O↦SCOTCSO\mapsto SC O^T CS, with C=∏iXiC=\prod_i X_i and S=∏i oddZiS=\prod_{i\ {\rm odd}}Z_i, sends the generator at VV to the one at −V-V for real hoppings, and odd density deviations to minus themselves. The leading interaction correction is therefore of order V2V^2. Consistently, the slope of rr against the averaged-rate chain below exceeds one by 0.06 at V=1V=1 and 0.25 at V=2V=2.

Methods. For L≤20L\le20 we diagonalise the generator in the half-filled sector exactly (sparse Lanczos; 184 756 configurations at L=20L=20). For larger LL we write minus the generator as a matrix-product operator with four-site terms. It is real symmetric, so the slowest mode is the first excited state of a Hermitian problem. We obtain it with two-site DMRG [11], orthogonalising against the uniform ground state and shifting the operator by a constant so that the excited state has negative energy. The ground state is a fixed-particle-number superposition of all configurations, with entanglement growing only logarithmically, and bond dimension 64 suffices: at L=24L=24 raising it to 96 changes rr in the sixth digit. DMRG reproduces exact diagonalisation to all printed digits for L=10L=10-1616, including samples with a link of Ji2=10−4J_i^2=10^{-4}. Unless stated otherwise J=1J=1, γ=4\gamma=4.

Results. Table S1 collects the residual statistics. For bounded fields, hi∈[−4,4]h_i\in[-4,4], the rms of Lr\sqrt{L}r is independent of LL for every VV. Interactions lower it from 0.12 to 0.08 at V=4V=4 while leaving the exponent at −1/2-1/2. Sample by sample, rr is 98-100% correlated with the residual of a resistor chain whose conductances are the configuration-averaged rates, Gˉi=Eπ[Gi1ηi≠ηi+1]\bar G_i=\mathbb{E}_\pi[G_i\mathbf{1}_{\eta_i\ne\eta_{i+1}}], which differ from bond to bond because the interaction energy adds to the local field difference. The ratio between the two settles at 1.6 for L=24L=24-3232. This enhancement comes from correlations between neighbouring occupations, which the averaged rates neglect. For heavy-tailed hoppings, Ji=viαJ_i=v_i^\alpha (α=1.25\alpha=1.25), interactions change rr by less than 2% for L≥16L\ge16, with and without fields, and rr is 99.7-100% correlated with the averaged-rate chain. In this regime the residual is dominated by the few weakest links, and interactions keep each rate between 8Ji2γ/[4γ2+(∣hi+1−hi∣+2V)2]8J_i^2\gamma/[4\gamma^2+(|h_{i+1}-h_i|+2V)^2] and 2Ji2/γ2J_i^2/\gamma. The resistance of a weak link therefore still scales as Ji−2J_i^{-2}, and its tail exponent is unchanged. The averaged-rate chain can be evaluated up to L=4096L=4096: over L=64L=64-4096 its median residual has exponent −0.230-0.230 at V=0V=0 and at V=4V=4 (−0.233-0.233 and −0.229-0.229 with fields) for α=1.25\alpha=1.25, and −0.446-0.446 for α=2.5\alpha=2.5. These are the same slowly converging exponents as for non-interacting resistor chains (Table S2 predicted −0.2-0.2 and −0.5-0.5). The threshold at α=2\alpha=2 thus survives interactions.

LL81012141618202432
bounded, V=0V=00.1260.1150.1230.1250.1320.1230.123
bounded, V=1V=10.1220.1110.1180.1210.1270.1180.118
bounded, V=2V=20.1110.1000.1060.1100.1140.1060.106
bounded, V=4V=40.0840.0760.0800.0850.0870.0800.0810.0770.081
heavy, no fields0.9600.9750.9790.9810.9820.9830.983
heavy, with fields0.9630.9820.9800.9850.9880.9910.998

Table S1. Interacting chains. Bounded fields: rms of Lr\sqrt{L}r (300 samples for L≤18L\le18, 100 at L=20L=20, 40 and 30 DMRG samples at L=24L=24, 32; relative error about 4%, 7%, 11%, 13%). Heavy tails: median of ∣r(V=4)/r(V=0)∣|r(V=4)/r(V=0)| over the same samples (400 for L≤18L\le18, 40 at L=24L=24).

quantitypredictionresult
exponent of dd, α=4, 3, 2.5\alpha= 4,\ 3,\ 2.5−0.5-0.5−0.50, −0.51, −0.49 (±0.01)-0.50,\ -0.51,\ -0.49\ (\pm0.01)
exponent of dd, α=2\alpha= 2−0.448-0.448 (with log⁡L\sqrt{\log L})−0.448±0.014-0.448 \pm0.014
exponent of dd, α=1.67, 1.43, 1.25, 1.11\alpha= 1.67,\ 1.43,\ 1.25,\ 1.11−0.40, −0.30, −0.20, −0.10-0.40,\ -0.30,\ -0.20,\ -0.10−0.39, −0.29, −0.15, −0.12 (±0.02-0.39,\ -0.29,\ -0.15,\ -0.12\ (\pm0.02-0.03)0.03)
exponent of dd, random fields, h0=1h_0 = 1-1616−0.5-0.5−0.495-0.495 to −0.498-0.498
exponent of the rate−2-2−2.00-2.00 to −2.09-2.09
exponent of ∣r∣\lvert r\rvert, quantum, γ=1\gamma= 1−0.5-0.5−0.49±0.02-0.49 \pm0.02
exponent of ∣r∣\lvert r\rvert, quantum, γ=0.125\gamma= 0.125-1616−0.5-0.5−0.47-0.47 to −0.53-0.53
exponent of the reach, random fields and α=1.25\alpha= 1.250.50.5 and 0.20.20.510.51 and 0.210.21
success probability, random fieldsEq. (32), no free parameteragrees within statistical errors
residual distribution, α=1.25\alpha= 1.25symmetric stable, Eq. (38)Figure S4
success probability, α=1.25\alpha= 1.25, 256≤L≤4096256 \le L \le4096Eqs. (38), (39)within 0.022
median reach, α=1.25\alpha= 1.25Eqs. (38), (39)9% at L=256L = 256, 3% for 512≤L≤4096512 \le L \le4096

Table S2. Main predictions and their numerical tests. Exponents of the slow-mode distortion are fitted over the largest chains (up to L=65536L = 65536). Quantum results use the exact Eq. (19).

Preparing the blind states on a quantum circuit and measuring them

Appendix B summarises the preparation and readout protocols and shows the four-qubit circuit in Figure 5. Here we give the state definitions, general circuit construction, and derivations. The reference and blind profiles are fi(k)=cos⁡(kπxi)f_i^{(k)}=\cos(k\pi x_i), with k=1k=1 and k=2k=2, respectively. Both preparations require only the profile, the particle number QQ, and the modulation amplitude. We consider 0<Q<L0<Q<L and write PQP_Q for the projector onto the QQ-particle sector, dQ=(LQ)d_Q=\binom{L}{Q}, ν=Q/L\nu=Q/L, and ρ∗=PQ/dQ\rho_* = P_Q/d_Q. For a profile satisfying ∑ifi=0\sum_i f_i=0, define

Af=∑ifi(ni−ν).(41)A_f=\sum_i f_i(n_i-\nu). \tag*{(41)}

Two families of states. We use two preparation families for complementary purposes. The exponential family is positive for any finite bias and admits the sequential circuit below; its density modulation follows the chosen profile in linear response. The affine family has a restricted amplitude range but gives exactly linear density and slow-mode responses, allowing the residual to be measured without an amplitude extrapolation.

The exponential family consists of the diagonal states

ρf(ϵ)=PQeϵAfPQZQ(ϵ)=∑∣η∣=Qpf(η)∣η⟩⟨η∣,\rho_f(\epsilon)=\frac{P_Qe^{\epsilon A_f}P_Q}{Z_Q(\epsilon)}=\sum_{|\eta|=Q}p_f(\eta)|\eta\rangle\langle\eta|,
pf(η)=1ZQ(ϵ)∏iwiηi,wi=eϵfi.(42)p_f(\eta)=\frac{1}{Z_Q(\epsilon)}\prod_i w_i^{\eta_i},\qquad w_i=e^{\epsilon f_i}. \tag*{(42)}

Here ηi∈{0,1}\eta_i\in\{0,1\}, ∣η∣=∑iηi|\eta|=\sum_i\eta_i, and ZQ(ϵ)Z_Q(\epsilon) normalises the state. This is a member of the maximum-entropy family at fixed particle number and prescribed mean occupations. Indeed, maximising −∑ηp(η)ln⁡p(η)-\sum_\eta p(\eta)\ln p(\eta) under these constraints gives p(η)∝exp⁡(∑ibiηi)p(\eta)\propto\exp(\sum_i b_i\eta_i); Eq. (42) uses the particular biases bi=ϵfib_i=\epsilon f_i. Under ρ∗\rho_*, exchangeability and the fixed total particle number give

Cov⁡ρ∗(ni,nj)=χL(δij−1L),χL=ν(1−ν)LL−1.(43)\operatorname{Cov}_{\rho_*}(n_i,n_j)=\chi_L\left(\delta_{ij}-\frac{1}{L}\right),\qquad\chi_L=\nu(1-\nu)\frac{L}{L-1}. \tag*{(43)}

Consequently

⟨ni⟩ρf(ϵ)=ν+ϵχLfi+O(ϵ2),(44)\langle n_i\rangle_{\rho_f(\epsilon)}=\nu+\epsilon\chi_Lf_i+O(\epsilon^2), \tag*{(44)}

Expanding ln⁡ZQ\ln Z_Q in the biases, ∂⟨ni⟩/∂bj∣b=0=Cov⁡ρ∗(ni,nj)\left.\partial\langle n_i\rangle/\partial b_j\right|_{b=0}=\operatorname{Cov}_{\rho_*}(n_i,n_j), so ⟨ni⟩=ν+ϵ∑jCov⁡ρ∗(ni,nj)fj+O(ϵ2)=ν+ϵχLfi+O(ϵ2)\langle n_i\rangle=\nu+\epsilon\sum_j\operatorname{Cov}_{\rho_*}(n_i,n_j)f_j+O(\epsilon^2)=\nu+\epsilon\chi_Lf_i+O(\epsilon^2), using ∑jfj=0\sum_jf_j=0. Thus the circuit prepares the state in Eq. (42) exactly, while its density modulation is proportional to fif_i to leading order in ϵ\epsilon. Exact target densities can also be imposed at finite amplitude. For 0<nitar<10<n_i^{\mathrm{tar}}<1 and ∑initar=Q\sum_i n_i^{\mathrm{tar}}=Q, the required biases minimise

ln⁡ZQ(b)−∑ibinitar,ZQ(b)=∑∣η∣=Qexp⁡(∑ibiηi),(45)\ln Z_Q(b)-\sum_i b_i n_i^{\mathrm{tar}},\qquad Z_Q(b)=\sum_{|\eta|=Q}\exp\left(\sum_i b_i\eta_i\right), \tag*{(45)}

over zero-sum bias vectors. The Hessian is the occupation covariance matrix, which is positive definite on this subspace. The biases therefore depend only on the target densities, not on the disorder sample. Reflection-even targets give reflection-even biases.

The second family consists of the affine states

ρfaff(ϵ)=PQ(1+ϵAf)PQdQ,∣ϵ∣≤1Mf,Mf=max⁡∣η∣=Q∣Af(η)∣.(46)\rho_f^{\mathrm{aff}}(\epsilon)=\frac{P_Q(1+\epsilon A_f)P_Q}{d_Q},\qquad|\epsilon|\leq\frac{1}{M_f},\qquad M_f=\max_{|\eta|=Q}|A_f(\eta)|. \tag*{(46)}

These states are positive throughout the stated range. Their densities are exactly ν+ϵχLfi\nu+\epsilon\chi_Lf_i, and their slow-mode amplitude is exactly ϵtr⁡(L1ρ∗Af)\epsilon\operatorname{tr}(L_1\rho_*A_f). Thus the residual has no linear-response error at any allowed ϵ\epsilon. Since the blind and reference cosine profiles have equal norms, their initial root-mean-square density-distance ratio is exactly κ=∣ϵb/ϵr∣\kappa=|\epsilon_b/\epsilon_r|. The restriction is the available amplitude: for cosine profiles at fixed filling, MfM_f grows linearly with LL, so this family requires ϵ=O(L−1)\epsilon=O(L^{-1}).

A deterministic circuit. Let Ei,qE_{i,q} be the sum of ∏j≥iwjηj\prod_{j\ge i} w_j^{\eta_j} over occupations of sites i,…,Li,\ldots,L containing qq particles. Separating the cases ηi=0\eta_i=0 and ηi=1\eta_i=1 gives

Ei,q=Ei+1,q+wiEi+1,q−1,EL+1,q=δq0,E1,Q=ZQ(ϵ).(47)E_{i,q}=E_{i+1,q}+w_iE_{i+1,q-1},\qquad E_{L+1,q}=\delta_{q0},\qquad E_{1,Q}=Z_Q(\epsilon). \tag*{(47)}

We set Ei,q=0E_{i,q}=0 for infeasible particle numbers. Conditional on having qq particles still to place, site ii is occupied with probability

pi,q=wiEi+1,q−1Ei,q.(48)p_{i,q}=\frac{w_iE_{i+1,q-1}}{E_{i,q}}. \tag*{(48)}

Use LL site qubits and a counter of m=⌈log⁡2(Q+1)⌉m=\lceil\log_2(Q+1)\rceil qubits, initialised in ∣0L⟩∣Q⟩\lvert0^L\rangle\lvert Q\rangle. At site ii, apply the controlled rotation

Vi=∑q∣q⟩⟨q∣⊗Ry(θi,q)i,θi,q=2arcsin⁡pi,q,(49)V_i=\sum_q\lvert q\rangle\langle q\rvert\otimes R_y(\theta_{i,q})_i,\qquad\theta_{i,q}=2\arcsin\sqrt{p_{i,q}}, \tag*{(49)}

where

Ry(θ)∣0⟩=cos⁡θ2∣0⟩+sin⁡θ2∣1⟩.(50)R_y(\theta)\lvert0\rangle=\cos\frac{\theta}{2}\lvert0\rangle+\sin\frac{\theta}{2}\lvert1\rangle. \tag*{(50)}

The formula for θi,q\theta_{i,q} applies when Ei,q>0E_{i,q}>0; angles on unreachable counter states can be chosen arbitrarily. Follow the rotation by the reversible decrement

Wi:∣ηi⟩∣q⟩⟼∣ηi⟩∣(q−ηi) mod 2m⟩.(51)W_i:\lvert\eta_i\rangle\lvert q\rangle\longmapsto\lvert\eta_i\rangle\lvert(q-\eta_i)\bmod2^m\rangle. \tag*{(51)}

Along a branch with occupations η\eta, let qi=Q−∑j<iηjq_i=Q-\sum_{j<i}\eta_j. Its squared amplitude telescopes

∏i=1LwiηiEi+1,qi−ηiEi,qi=∏iwiηiZQ(ϵ)EL+1,Q−∣η∣.(52)\prod_{i=1}^{L}\frac{w_i^{\eta_i}E_{i+1,q_i-\eta_i}}{E_{i,q_i}}=\frac{\prod_i w_i^{\eta_i}}{Z_Q(\epsilon)}E_{L+1,Q-\lvert\eta\rvert}. \tag*{(52)}

This vanishes unless ∣η∣=Q\lvert\eta\rvert=Q. All amplitudes are nonnegative because the rotation angles lie in [0,π][0,\pi], and the counter ends in ∣0⟩\lvert0\rangle on every branch. The circuit WLVL⋯W1V1W_LV_L\cdots W_1V_1 therefore outputs

∣Ψf⟩∣0⟩,∣Ψf⟩=1ZQ(ϵ)∑∣η∣=Q∏iwiηi/2∣η⟩.(53)\lvert\Psi_f\rangle\lvert0\rangle,\qquad\lvert\Psi_f\rangle=\frac{1}{\sqrt{Z_Q(\epsilon)}}\sum_{\lvert\eta\rvert=Q}\prod_i w_i^{\eta_i/2}\lvert\eta\rangle. \tag*{(53)}

Next apply ZziZ^{z_i} to each site, with independent fair bits ziz_i, and discard their record. Since 2−L∑z∈{0,1}L(−1)z⋅(η−ξ)=δηξ2^{-L}\sum_{z\in\{0,1\}^L}(-1)^{z\cdot(\eta-\xi)}=\delta_{\eta\xi}, this removes all coherences and yields ρf(ϵ)\rho_f(\epsilon) in Eq. (42). Equivalently, copy each site onto an ancilla with a CNOT and trace out the ancillas. Computing the table Ei,qE_{i,q} costs O(LQ)O(LQ) operations, and the circuit has at most O(LQ)O(LQ) conditional rotation angles. If stage jj has operator-norm error ϵj\epsilon_j, the output error obeys 12∥ρ~−ρf(ϵ)∥1≤∑jϵj\frac{1}{2}\lVert\widetilde{\rho}-\rho_f(\epsilon)\rVert_1\le\sum_j\epsilon_j. Gate-level simulations reproduce ∣Ψf⟩\lvert\Psi_f\rangle to 3×10−163\times10^{-16} for (L,Q)(L,Q) up to (9,5)(9,5), using both cosine and random biases.

Four qubits. For L=4L=4, Q=2Q=2, and the blind profile f=(1,−1,−1,1)/2f=(1,-1,-1,1)/\sqrt{2}, set w=eϵ/2w=e^{\epsilon/\sqrt{2}}. The coherent state before phase averaging is

∣Ψf⟩=∣1100⟩+∣1010⟩+w∣1001⟩+w−1∣0110⟩+∣0101⟩+∣0011⟩w2+w−2+4.(54)\lvert\Psi_f\rangle=\frac{\lvert1100\rangle+\lvert1010\rangle+w\lvert1001\rangle+w^{-1}\lvert0110\rangle+\lvert0101\rangle+\lvert0011\rangle}{\sqrt{w^2+w^{-2}+4}}. \tag*{(54)}

Controls on the earlier site qubits replace the counter, as shown in Figure 5 in Appendix B. Site 1 is occupied with probability

p1=w2+2w2+w−2+4,θ1=2arcsin⁡p1.(55)p_1=\frac{w^2+2}{w^2+w^{-2}+4},\qquad\theta_1=2\arcsin\sqrt{p_1}. \tag*{(55)}

The rotation on site 2 depends on the occupation of site 1

θ2∣0=2arcsin⁡w−2+1w−2+2,θ2∣1=2arcsin⁡1w2+2.(56)\theta_{2\lvert0}=2\arcsin\sqrt{\frac{w^{-2}+1}{w^{-2}+2}},\qquad\theta_{2\lvert1}=2\arcsin\frac{1}{\sqrt{w^2+2}}. \tag*{(56)}

Define

θ+=θ2∣0+θ2∣12,θ−=θ2∣0−θ2∣12.(57)\theta_{+} = \frac{\theta_{2|0}+\theta_{2|1}}{2}, \qquad\theta_{-} = \frac{\theta_{2|0}-\theta_{2|1}}{2}. \tag*{(57)}

The conditional rotation is implemented by Ry(θ+)R_y(\theta_{+}), a CNOT controlled by site 1, Ry(θ−)R_y(\theta_{-}), and a second CNOT. The net angle is θ2∣0\theta_{2|0} when site 1 is empty and θ2∣1\theta_{2|1} when it is occupied.

Site 3 must be occupied after 00 and empty after 11. After 01 or 10, it requires the angle

θ3=2arcsin⁡11+w2.(58)\theta_3 = 2\arcsin\frac{1}{\sqrt{1+w^2}}. \tag*{(58)}

This is implemented by Ry(π)R_y(\pi), two singly controlled Ry(θ3−π)R_y(\theta_3-\pi) rotations, and a doubly controlled Ry(π−2θ3)R_y(\pi-2\theta_3) rotation. Their angles add to the required value on each branch. Finally, three CNOTs set the occupation of site 4 to the parity of the first three sites, completing the configuration with exactly two particles. Independent random ZZ gates then produce the diagonal state. At ϵ=0.2\epsilon=0.2, the angles are

(θ1,θ2∣0,θ2∣1,θ3)=(1.665214,1.848022,1.160546,1.429844),(59)(\theta_1,\theta_{2|0},\theta_{2|1},\theta_3)=(1.665214,1.848022,1.160546,1.429844), \tag*{(59)}

and the densities are (0.547139,0.452861,0.452861,0.547139)(0.547139,0.452861,0.452861,0.547139). The simulated circuit reproduces ∣Ψf⟩|\Psi_f\rangle to 2×10−162\times10^{-16} at ϵ=−1\epsilon=-1, 0, 0.2, and 1.

Conserving particle number throughout preparation. The rotation circuits end in the QQ-particle sector but pass through other particle numbers. If QQ must be conserved at every step, the diagonal target state can instead be prepared by classical sampling followed by loading. For the maximum-entropy family, sample a configuration sequentially using the probabilities pi,qp_{i,q}. For the affine family, draw a uniform configuration with QQ particles and accept it with probability

1+ϵAf(η)1+∣ϵ∣Mf.\frac{1+\epsilon A_f(\eta)}{1+|\epsilon|M_f}.

The mean acceptance probability is 1/(1+∣ϵ∣Mf)≥1/21/(1+|\epsilon|M_f)\ge1/2 throughout the positivity range in Eq. (46). For either family, prepare ∣1Q0L−Q⟩|1^Q0^{L-Q}\rangle and move the particles to the sampled sites using SWAP gates. These gates commute with total particle number. Pairing misplaced particles with empty target sites requires at most min⁡(Q,L−Q)\min(Q,L-Q) swaps between arbitrary sites, or O(LQ)O(LQ) nearest-neighbor swaps. Averaging over the sampled configurations gives the desired diagonal state.

A heralded alternative. Another construction fixes the particle number by postselection. Apply local rotations Ry(2arcsin⁡pi)R_y(2\arcsin\sqrt{p_i}), with

pi=ζwi1+ζwi,ζ>0,(60)p_i=\frac{\zeta w_i}{1+\zeta w_i}, \qquad\zeta>0, \tag*{(60)}

and measure the total particle number. The outcome QQ heralds ∣Ψf⟩|\Psi_f\rangle with probability

PQsucc=ζQZQ(ϵ)∏i(1+ζwi).(61)P_Q^{\mathrm{succ}}=\frac{\zeta^Q Z_Q(\epsilon)}{\prod_i(1+\zeta w_i)}. \tag*{(61)}

This probability is maximised when ∑ipi=Q\sum_i p_i=Q. When the particle-number distribution admits a Gaussian approximation

PQsucc≃[2π∑ipi(1−pi)]−1/2.(62)P_Q^{\mathrm{succ}}\simeq\left[2\pi\sum_i p_i(1-p_i)\right]^{-1/2}. \tag*{(62)}

For uniform weights, this becomes [2πLν(1−ν)]−1/2[2\pi L\nu(1-\nu)]^{-1/2}. The final phase averaging gives ρf(ϵ)\rho_f(\epsilon). This construction fixes QQ at its output but does not conserve particle number during the rotations.

Measuring the residual. Each profile is measured at ±ϵ\pm\epsilon: the symmetric difference removes the even-order nonlinear response, and the ratio of the two profiles removes the unknown decay and readout factors common to both. Write Ak=Af(k)A_k=A_{f^{(k)}}, with k=1k=1 for the reference and k=2k=2 for the blind profile. In the same disorder realisation, prepare each state at biases ±ϵ\pm\epsilon and measure A1A_1 from the site-resolved densities. Define

R1k(t)=⟨A1⟩k,+ϵ(t)−⟨A1⟩k,−ϵ(t)2ϵ.(63)R_{1k}(t)=\frac{\langle A_1\rangle_{k,+\epsilon}(t)-\langle A_1\rangle_{k,-\epsilon}(t)}{2\epsilon}. \tag*{(63)}

For the affine family, [Ak,PQ]=0[A_k,P_Q]=0 gives

ρkaff(ϵ)=ρ∗+ϵρ∗Ak,ρkaff(ϵ)−ρkaff(−ϵ)2ϵ=ρ∗Ak.\rho_k^{\mathrm{aff}}(\epsilon)=\rho_*+\epsilon\rho_*A_k,\qquad\frac{\rho_k^{\mathrm{aff}}(\epsilon)-\rho_k^{\mathrm{aff}}(-\epsilon)}{2\epsilon}=\rho_*A_k.

This identity is exact throughout the positivity range. For the exponential family, the same symmetric difference equals ρ∗Ak+O(ϵ2)\rho_*A_k+O(\epsilon^2), because the even powers of ϵ\epsilon cancel. Once the slow mode dominates, the response is

R1k(t)≃Tr⁡(A1R1)eλ1t[Tr⁡(L1†ρ∗Ak)+O(ϵ2)],(64)R_{1k}(t)\simeq\operatorname{Tr}(A_1R_1)e^{\lambda_1t}\left[\operatorname{Tr}(L_1^\dagger\rho_*A_k)+O(\epsilon^2)\right], \tag*{(64)}

where the O(ϵ2)O(\epsilon^2) term is absent for the affine family. Provided the reference overlap and the readout factor Tr⁡(A1R1)\operatorname{Tr}(A_1R_1) are nonzero, the common factors cancel

R12(t)R11(t)⟶t→∞{r,affine preparations,r+O(ϵ2),maximum-entropy preparations.(65)\frac{R_{12}(t)}{R_{11}(t)}\mathop{\longrightarrow}_{t\to\infty} \begin{cases} r, & \text{affine preparations},\\ r+O(\epsilon^2), & \text{maximum-entropy preparations}. \end{cases} \tag*{(65)}

Thus the affine family measures rr at every allowed amplitude, while the maximum-entropy family requires a small-amplitude extrapolation to remove the remaining bias. For the interacting lattice gas with L=10L=10, V=γ=4V=\gamma=4, and fields in [−4,4][-4,4], the numerical response ratio agrees with rr to within 10−610^{-6} once t>10/(λ1−λ2)t>10/(\lambda_1-\lambda_2), at the two modulation amplitudes tested.

Assumptions during preparation. The profiles, biases, and rotation angles depend only on the chosen target, QQ, and the modulation amplitude, never on the fields, hoppings, or their distribution. During the coherent gate sequence, unwanted Hamiltonian evolution and dephasing must be negligible or controlled well enough to implement the prescribed gates. Switching off hopping alone is not sufficient in general, because the intermediate states need not be diagonal. Once the diagonal state has been prepared, diagonal fields, density-density interactions, and local dephasing leave it unchanged. The prescribed relaxation dynamics, including hopping, can then be applied.

Numerical checks

We test these predictions in two ways: with the exact quantum dynamics Eq. (19), and with the classical resistor chain, which reaches L=216L=2^{16}. Table S2 compares the main predictions with the data. For random fields the quantum exponents match the resistor picture at all dephasing rates studied [Fig. S2]. For random hoppings over L=48L=48–256256 they agree with the classical chain within about two standard errors, except at α=2.5\alpha=2.5, γ=0.25\gamma=0.25, where the quantum exponent −0.47±0.04-0.47\pm0.04 lies closer to the asymptotic −1/2-1/2 than the classical one −0.37±0.02-0.37\pm0.02 at these sizes [Fig. S5]. The Néel state and the imbalance used for this model in Ref. [2] barely see the slow mode. For a smooth profile ff, the alternating sum ∑i(−1)if(i/L)\sum_i (-1)^i f(i/L) is only a boundary term, so the Néel state’s overlap with the slow mode and the slow mode’s weight in the imbalance are each of order 1/L1/L. In a clean periodic chain of even length this overlap vanishes exactly by translation symmetry; for odd length it is of order L−2L^{-2} [2]. We therefore compare the preparations using the root-mean-square density deviation defined in the main text. The cos⁡(πx)\cos(\pi x) component of a site-resolved density profile can monitor the late-time slow mode, but it vanishes initially for the cos⁡(2πx)\cos(2\pi x) blind preparation and cannot by itself define the starting-distance ratio κ\kappa.

References

References

  1. [1]M. Žnidarič, Phys. Rev. E 92, 042143 (2015).
  2. [2]R. Ghosh and M. Žnidarič, Phys. Rev. B 107, 184303 (2023).
  3. [3]S. Alexander, J. Bernasconi, W. R. Schneider, and R. Orbach, Rev. Mod. Phys. 53, 175 (1981).
  4. [4]A. Faggionato, Electron. J. Probab. 17, 15 (2012).
  5. [5]T. Kato, Perturbation Theory for Linear Operators (Springer, Berlin, 1995).
  6. [6]P. Billingsley, Convergence of Probability Measures, 2nd ed. (Wiley, New York, 1999).
  7. [7]A. Bourgeat and A. Piatnitski, Asymptot. Anal. 21, 303 (1999).
  8. [8]G. Bal, J. Garnier, S. Motsch, and V. Perrier, Asymptot. Anal. 59, 1 (2008).
  9. [9]W. Whitt, Stochastic-Process Limits (Springer, New York, 2002).DOI
  10. [10]G. Samorodnitsky and M. S. Taqqu, Stable Non-Gaussian Random Processes (Chapman & Hall, New York, 1994).
  11. [11]J. Hauschild and F. Pollmann, SciPost Phys. Lect. Notes 5 (2018).

Paper details

vlnc:W_g6rv26vs
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 12

downloads · 1

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents