betaFeedback

WorksPhysics and Astronomy

Precision Limits and Non-Gaussian Advantage in Optical Phase Estimation under Preparation Constraints

Abstract

In optical quantum metrology, the scaling of the quantum Fisher information (QFI) with the mean photon number is a standard benchmark for relating precision to probe resources. For phase estimation, Heisenberg scaling corresponds to a quadratic growth of the QFI with the mean photon number, which Gaussian probes can approach. Interestingly, certain non-Gaussian probes can exhibit arbitrarily large QFI even at fixed mean photon number. Here, we clarify the resource requirements underlying this behavior by introducing a constraint that separates physically admissible states from inadmissible ones: an envelope on the photon-number distribution that arises naturally from the physical limitations of state-generation methods. Building on this constraint and considering pure single-mode probes, we define a resource quantity, NRMSN_\mathrm{RMS}, fixed by the preparation constraints, and derive a finite bound on the QFI whose Heisenberg scaling is attainable. We identify a genuine metrological advantage of non-Gaussian states over Gaussian probes subject to the same constraints. Finally, we illustrate these results with an optical preparation model based on a universal gate set.

Introduction

Quantum metrology exploits quantum resources to improve parameter-estimation precision beyond classical limits [1, 2]. An unknown parameter θ\theta is encoded in a probe state ρθ\rho_\theta, from which it is inferred through measurements. For a locally unbiased estimator, the variance satisfies the quantum Cramér–Rao bound (QCRB) [3, 4, 5]

(Δθ)2≥1mFθ.(1)(\Delta\theta)^2 \ge\frac{1}{m F_\theta}. \tag*{(1)}

Here, mm is the number of independent experiment repetitions and FθF_\theta denotes the quantum Fisher information (QFI), which quantifies the response of the probe to a small variation in the unknown parameter.

Optical systems provide a prominent platform for quantum metrology [6, 7]. A paradigmatic task is the estimation of a phase shift encoded by eiθn^e^{i\theta\hat{n}}, where n^\hat{n} is the photon-number operator. Taking the mean photon number ⟨n^⟩\langle\hat{n}\rangle as the resource, classical probes, described by mixtures of coherent states, satisfy Fθ≤4⟨n^⟩F_\theta\le4\langle\hat{n}\rangle, corresponding to shot-noise scaling. Nonclassical probes can surpass this bound; for instance squeezed-vacuum states exhibit Fθ∼⟨n^⟩2F_\theta\sim\langle\hat{n}\rangle^2 at large mean photon number, achieving Heisenberg scaling [8].

While Gaussian states have been extensively studied and their QFI bounded at fixed mean photon number [8], two natural questions arise: whether non-Gaussian probes can offer greater sensitivity and what limits apply to these more general states. The first question has attracted considerable recent interest; see, for instance, [9, 10, 11]. However, the mean photon number alone imposes no finite universal bound on the QFI: non-Gaussian probes can exhibit arbitrarily large sensitivity even at fixed mean photon number [12, 13, 14]. As previously emphasized, this resource measure can obscure the resources associated with rare, highly excited components and therefore provide an incomplete account of metrological resources [15].

In this work, we propose a framework that resolves this issue by showing that the limitations of experimental setups impose a constraint on the photon-number distributions of admissible probes. This constraint reveals that states exhibiting divergent behavior are physically inaccessible. Within this framework, we introduce a counterpart of the mean photon number to quantify the required resources, which we call the root-mean-square photon budget, NRMSN_{\mathrm{RMS}}. This quantity incorporates the physical limitations of the setup and allows us to derive a finite bound on the QFI that recovers a Heisenberg limit in terms of this budget. We further compare Gaussian and non-Gaussian probes, showing that preparation constraints can make the latter necessary to achieve Heisenberg scaling. In particular, for families of exponentially decaying envelopes with fixed decay length, the Gaussian QFI grows at most linearly with NRMSN_{\mathrm{RMS}}, whereas non-Gaussian probes can attain quadratic scaling. Finally, we illustrate the framework with a setup built from a universal gate set, in which the limiting resources that determine both the envelope and the photon budget of the experiment are readily identified.

Preliminaries

We consider local phase estimation with pure single-mode optical probes, assuming lossless encoding and an ideal external phase reference whose resources are not counted [16]. We quantify sensitivity through the QFI and assume asymptotic attainability of the QCRB as the number of measurements, mm, grows; prior uncertainty and precision at fixed total resources are beyond this scope. An unknown phase shift θ\theta is encoded into the probe ∣ψ0⟩|\psi_0\rangle as,

∣ψθ⟩=eiθn^∣ψ0⟩.(2)|\psi_\theta\rangle= e^{i\theta\hat{n}}|\psi_0\rangle. \tag*{(2)}

For this encoding, the QFI can be written in the following simple form [5]

Fθ=4Var⁡∣ψ0⟩[n^].(3)\mathcal{F}_\theta= 4\operatorname{Var}_{|\psi_0\rangle}[\hat{n}]. \tag*{(3)}

A well-studied class of probes are Gaussian states [8, 17], which QFI is upper-bounded as,

Fθ≤8(⟨n^⟩ψ02+⟨n^⟩ψ0).(4)\mathcal{F}_\theta\le8\left(\langle\hat{n}\rangle_{\psi_0}^{2}+\langle\hat{n}\rangle_{\psi_0}\right). \tag*{(4)}

This bound is attained for squeezed-vacuum states, i.e., they are optimal within this family. This bound defines the standard Gaussian limit (SGL), which remains finite whenever the mean photon number ⟨n^⟩ψ0\langle\hat{n}\rangle_{\psi_0} is finite. Therefore, when the mean photon number is the sole resource, optimal Gaussian states reach the Heisenberg limit. For Gaussian probes, fixing this first moment guarantees a finite QFI, because constraining the mean photon number bounds all higher-order moments.

Heavy-tailed states

Outside the set of Gaussian states, we can always construct probes with a finite mean photon number but a divergent second moment and, hence, a divergent QFI. The simplest such family consists of what we call heavy-tailed states,

∣ψγ⟩=N∑n=1∞1nγ∣n⟩,γ∈(1,32].(5)|\psi_\gamma\rangle= \mathcal{N}\sum_{n=1}^{\infty}\frac{1}{n^\gamma}|n\rangle,\qquad\gamma\in\left(1,\frac{3}{2}\right]. \tag*{(5)}

In Appendix § 2.0.1, we show that their QFI diverges while their mean photon number remains finite.

Although these states are normalizable and hence belong to the Hilbert space, their physical realization is problematic because their energy variance diverges. Their unbounded support alone is not responsible for the divergence, as coherent states also have unbounded support but finite variance. Instead, this behavior arises from the slow algebraic decay of their photon-number distribution.

Asymmetrically superposed states

Along these lines, Refs. [11, 12] proposed probes based on asymmetric states. They superpose the vacuum ∣0⟩|0\rangle with a highly excited state ∣ϕ⟩|\phi\rangle (with ⟨ϕ∣n^∣ϕ⟩>0\langle\phi|\hat{n}|\phi\rangle> 0 and ⟨0∣ϕ⟩=0\langle0|\phi\rangle= 0) to build states with large photon-number variances. To keep the mean photon number low, the excitation is weighted with a probability amplitude 0<ϵ≤10 < \epsilon\le1,

∣ψϵ⟩=1−ϵ2∣0⟩+ϵ∣ϕ⟩.(6)|\psi_\epsilon\rangle= \sqrt{1-\epsilon^2}|0\rangle+\epsilon|\phi\rangle. \tag*{(6)}

For excitations that overlap the vacuum, the state requires an ϵ\epsilon-dependent normalization, which modifies the expressions below, although the physical conclusions are not altered. These states have finite energy fluctuations whenever the excitation has a finite second moment, ⟨ϕ∣n^2∣ϕ⟩<∞\langle\phi|\hat{n}^2|\phi\rangle< \infty. Since ⟨n^⟩ψϵ=ϵ2⟨n^⟩ϕ\langle\hat{n}\rangle_{\psi_\epsilon}=\epsilon^2\langle\hat{n}\rangle_\phi, the QFI of this state is

Fθ=4ϵ2(⟨n^2⟩ϕ−ϵ2⟨n^⟩ϕ2)=4⟨n^⟩ψϵ(⟨n^2⟩ϕ⟨n^⟩ϕ−⟨n^⟩ψϵ).(7)\begin{aligned} \mathcal{F}_\theta= 4\epsilon^2\left(\langle\hat{n}^2\rangle_\phi-\epsilon^2\langle\hat{n}\rangle_\phi^2\right) \\ &=4\langle\hat{n}\rangle_{\psi_\epsilon}\left(\frac{\langle\hat{n}^2\rangle_\phi}{\langle\hat{n}\rangle_\phi}-\langle\hat{n}\rangle_{\psi_\epsilon}\right). \tag*{(7)} \end{aligned}

Higher excitations can therefore enhance the QFI at fixed mean photon number. For instance, the Fock excitation ∣ϕ⟩=∣N⟩|\phi\rangle=|N\rangle with N=ν/ϵ2N=\nu/\epsilon^2, along values of ϵ\epsilon for which NN is an integer, keeps the probe mean fixed at ⟨n^⟩ψϵ=ν\langle\hat{n}\rangle_{\psi_\epsilon}=\nu and gives

Fθ=4ν2(ϵ−2−1),(8)\mathcal{F}_\theta=4\nu^2\left(\epsilon^{-2}-1\right), \tag*{(8)}

which diverges as ϵ→0\epsilon\to0 and surpasses the Heisenberg limit when the mean photon number is taken as the resource. Other excitations produce the same divergence with family-dependent prefactors.

The divergence in the preceding example requires increasingly large Fock excitations as ϵ→0\epsilon\to0. For a fixed excitation ∣ϕ⟩|\phi\rangle with finite second moment, the QFI instead vanishes in this limit. Its maximum over the weight ϵ\epsilon occurs at

ϵ∗2=min⁡{1,⟨n^2⟩ϕ2⟨n^⟩ϕ2}.(9)\epsilon_*^2=\min\left\{1,\frac{\langle\hat{n}^2\rangle_\phi}{2\langle\hat{n}\rangle_\phi^2}\right\}. \tag*{(9)}

Since ⟨n^2⟩ϕ≥⟨n^⟩ϕ2\langle\hat{n}^2\rangle_\phi\ge\langle\hat{n}\rangle_\phi^2, the optimal weight satisfies 1/2≤ϵ∗2≤11/2\le\epsilon_*^2\le1, with ϵ∗2=1/2\epsilon_*^2=1/2 for a Fock excitation (see Appendix § 2.0.2). Figure 1 illustrates analogous behavior for coherent and squeezed-vacuum excitations, with the maximum QFI attained away from ϵ→0\epsilon\to0. Thus, reducing the excitation weight does not indefinitely improve sensitivity when the excitation itself is fixed. The probe’s mean photon number alone does not capture this distinction, since a small weight can compensate for an increasingly energetic excitation.

QFI versus $\epsilon$ for Rivas-Luis and asymmetric superposition of coherent states

Figure 1. QFI versus ϵ\epsilon for ∣ψRL⟩=N(1−ϵ2∣0⟩+ϵ∣χ⟩)\lvert\psi_{\mathrm{RL}}\rangle= \mathcal{N}(\sqrt{1-\epsilon^2}\lvert0\rangle+\epsilon\lvert\chi\rangle), with ∣χ⟩=S^(r)∣0⟩\lvert\chi\rangle=\hat{S}(r)\lvert0\rangle at 21.4 dB [r=21.4ln⁡10/20r=21.4\ln10/20, zero relative phase] (blue), and for the asymmetric superposition of coherent states ∣ψaSCS⟩=N(ϵ∣0⟩+ϵ∣α⟩)\lvert\psi_{\mathrm{aSCS}}\rangle=\mathcal{N}(\sqrt{\epsilon}\lvert0\rangle+\epsilon\lvert\alpha\rangle), with α=10\alpha=10 (red). The latter peaks at ϵ≃1/2\epsilon\simeq1/\sqrt{2} (displaced cat), while the former is maximal for the pure squeezed vacuum (ϵ=1\epsilon=1).

Preparation constraints through photon-number envelopes

The behavior of the states analyzed in the previous sections exposes the tension inherent in using the mean photon number as the figure of merit. This ambiguity disappears once the finite resources available for state preparation are taken into account, such as finite pumping [18, 19, 20, 21, 22] or bounded squeezing [23]. In particular, the expected photon number of a state cannot exceed the number of photons the source is able to generate. Likewise, the probability of populating Fock states above the mean photon number produced by the source cannot decay more slowly than the source’s own photon-number distribution. Depending on the platform, this envelope can take the form of a hard cutoff, as in Law-Eberly-type protocols [24, 25], or of an exponential (or faster) decay, as for coherent photon sources [18, 19, 20, 21, 22] or heralding protocols [20].

To unify and model these constraints, we introduce a general treatment based on a point-wise envelope, a non-negative sequence (gn)n(g_n)_n that bounds the photon-number distribution qn:=∣⟨n∣ψ⟩∣2q_n := |\langle n|\psi\rangle|^2. Notably, this approach provides a coarse description of preparation restrictions without specifying the photon source. For a given threshold Nmax⁡N_{\max}, we require the envelope to satisfy gn=1g_n=1 for 1≤n≤Nmax⁡1\le n\le N_{\max}, 0≤gn≤10\le g_n\le1 for n>Nmax⁡n>N_{\max}, and ∑n≥Nmax⁡n2gn<∞\sum_{n\ge N_{\max}}n^2g_n<\infty. Then, the admissible distributions are

Qg={qn≥0:∑n≥0qn=1, qn≤gn (n≥Nmax⁡)}.(10)\mathcal{Q}_g=\left\{q_n\ge0:\sum_{n\ge0}q_n=1,\ q_n\le g_n\ (n\ge N_{\max})\right\}. \tag*{(10)}

Thus, the range 0≤n≤Nmax⁡0\le n\le N_{\max} forms a free window, including an unconstrained vacuum population, while a nonzero tail allows occupation above Nmax⁡N_{\max}. This constraint defines the class considered here, although additional restrictions imposed by a specific device may narrow it.

In Sec. VI, we illustrate how the framework translates apparatus-level restrictions into a common constraint on the probes used for phase estimation. We consider a preparation model built from a universal optical gate set, where bounds on the available gate resources constrain the output photon-number distributions and thereby determine an envelope.

Limits on the precision

Since Fθ=4Var⁡[n^]\mathcal{F}_\theta=4\operatorname{Var}[\hat{n}], Eq. (3), the second moment controls the sensitivity; accounting for both typical photon numbers and rare high-energy events. To compare probes under the same preparation constraints, we consider the largest second moment allowed by the envelope rather than the moment of a particular probe. Taking its square root gives a photon-number scale fixed by these constraints, which we define as the envelope’s root-mean-square photon budget,

NRMS(g):=(max⁡q∈QgEq[n^2])1/2.(11)N_{\mathrm{RMS}}(g):=\left(\max_{q\in\mathcal{Q}_g}\mathbb{E}_q[\hat{n}^2]\right)^{1/2}. \tag*{(11)}

Here, Eq[n^]:=∑n≥0qnn2\mathbb{E}_q[\hat{n}]:=\sum_{n\ge0}q_n n^2 denotes the expectation with respect to the photon-number distribution qq. In Appendix C, we construct the unique maximizing distribution, which need not maximize the QFI.

For every admissible probe we derive the following upper-bound,

Fθ≤4⟨n^2⟩≤4NRMS2(g).(12)\mathcal{F}_\theta\le4\langle\hat{n}^2\rangle\le4N_{\mathrm{RMS}}^2(g). \tag*{(12)}

Thus, we obtain a Heisenberg limit in terms of the root-mean-square photon budget, fixed by the preparation constraints independently of the chosen probe. The condition ∑n>Nmax⁡n2gn<∞\sum_{n>N_{\max}}n^2g_n<\infty ensures that this bound is finite and holds uniformly over all admissible probes. In particular, rare high-photon-number components cannot produce unbounded sensitivity for a fixed envelope budget.

The quadratic scaling with NRMS(g)N_{\mathrm{RMS}}(g) is attainable up to a constant factor by states whose photon-number distributions belong to Qg\mathcal{Q}_g. Denoting by tt the distribution that maximizes Eq. (11), we construct the pure probe

∣ψt⟩=1+t02∣0⟩+∑n≥1tn2eiϕn∣n⟩.(13)\lvert\psi_t\rangle=\sqrt{\frac{1+t_0}{2}}\lvert0\rangle+\sum_{n\ge1}\sqrt{\frac{t_n}{2}}e^{i\phi_n}\lvert n\rangle. \tag*{(13)}

This state is admissible for arbitrary phases and satisfies

Fθ(ψt)=2NRMS2(g)−(∑n≥0ntn)2≥NRMS2(g),(14)\mathcal{F}_\theta(\psi_t)=2N_{\mathrm{RMS}}^2(g)-\left(\sum_{n\ge0}nt_n\right)^2\ge N_{\mathrm{RMS}}^2(g), \tag*{(14)}

where the inequality follows from Et[n^]2≤Et[n^2]=NRMS2(g)\mathbb{E}_t[\hat{n}]^2\le\mathbb{E}_t[\hat{n}^2]=N_{\mathrm{RMS}}^2(g).

To define scaling, consider a specified family of envelopes g(λ)g^{(\lambda)}, where λ\lambda is a tunable parameter of the preparation apparatus, for example, λ=Nmax⁡\lambda=N_{\max} at fixed tail shape and decay parameters. For each λ\lambda, we select an admissible probe and keep both the envelope and the state fixed throughout the mm repetitions. Scaling compares these experiments as NRMS(g(λ))N_{\mathrm{RMS}}(g^{(\lambda)}) increases.

For a hard cutoff, gn=0g_n=0 for n>Nmax⁡n>N_{\max}, the RMS budget reduces to NRMS=Nmax⁡N_{\mathrm{RMS}}=N_{\max}, and Eq. (12) yields Fθ≤4Nmax⁡2\mathcal{F}_{\theta}\le4N_{\max}^{2}. A sharper bound follows from the bounded photon-number support. By Popoviciu’s inequality [26], Var⁡(n^)≤Nmax⁡2/4\operatorname{Var}(\hat{n})\le N_{\max}^{2}/4, hence Fθ≤Nmax⁡2\mathcal{F}_{\theta}\le N_{\max}^{2}. This bound is attained for the balanced superposition (∣0⟩+∣Nmax⁡⟩)/2(|0\rangle+|N_{\max}\rangle)/\sqrt{2} [14].

We now return to the state families of Sec. II. The envelope constraint excludes heavy-tailed probes with divergent second moments, since we require ∑n>1n2gn<∞\sum_{n>1}n^{2}g_n<\infty. For the asymmetrically superposed states, fixing a common envelope makes NRMS(g)N_{\mathrm{RMS}}(g) independent of the vacuum weight, even though the probe’s actual photon-number moments vary with ϵ\epsilon. Increasing the excitation while keeping the probe’s mean photon number fixed cannot produce a divergent QFI under this envelope. Indeed,

⟨n^2⟩ψϵ⟨n^⟩ϕ=⟨n^2⟩ψϵ≤NRMS2(g),(15)\frac{\langle\hat{n}^{2}\rangle_{\psi_{\epsilon}}}{\langle\hat{n}\rangle_{\phi}}=\langle\hat{n}^{2}\rangle_{\psi_{\epsilon}}\le N_{\mathrm{RMS}}^{2}(g), \tag*{(15)}

so the QFI in Eq. (7) remains bounded. Thus, the fixed-mean divergence discussed in Sec. II requires an unbounded increase in NRMS(g)N_{\mathrm{RMS}}(g), revealing a growing resource requirement that the mean photon number alone does not capture.

Non-Gaussian advantage

We now use this framework to characterize non-Gaussian advantage. Within a fixed envelope, every pure Gaussian state ∣ψG⟩|\psi_G\rangle satisfies ⟨n^⟩∣ψG⟩≤Nˉ(g)\langle\hat{n}\rangle_{|\psi_G\rangle}\le\bar{N}(g), where Nˉ(g)=max⁡q∈QgEq[n^]\bar{N}(g)=\max_{q\in\mathcal{Q}_g}E_q[\hat{n}] is attained by the distribution tt that maximizes the RMS photon budget (see Appendix C). Since x(x+1)x(x+1) is increasing for x≥0x\ge0, Eq. (4) yields

Fθ(ψG)≤8Nˉ(g)(Nˉ(g)+1).(16)\mathcal{F}_{\theta}(\psi_G)\le8\bar{N}(g)\bigl(\bar{N}(g)+1\bigr). \tag*{(16)}

Note that Nˉ(g)\bar{N}(g) is an auxiliary quantity fixed by the envelope, not an additional resource. Combining Eqs. (14) and (16), we find

Fθ(ψt)Fθ(ψG)≥NRMS2(g)8Nˉ(g)(Nˉ(g)+1).(17)\frac{\mathcal{F}_{\theta}(\psi_t)}{\mathcal{F}_{\theta}(\psi_G)}\ge\frac{N_{\mathrm{RMS}}^{2}(g)}{8\bar{N}(g)\bigl(\bar{N}(g)+1\bigr)}. \tag*{(17)}

Since the right-hand side depends only on the envelope gg, this inequality provides a sufficient criterion for non-Gaussian advantage directly from the envelope constraint. Whenever it exceeds one, the probe ∣ψt⟩|\psi_t\rangle of Eq. (13), which attains optimal scaling, outperforms every admissible Gaussian state. Along a family g(λ)g^{(\lambda)} with increasing NRMS(g(λ))N_{\mathrm{RMS}}(g^{(\lambda)}), the condition Nˉ(g)/NRMS(g)→0\bar{N}(g)/N_{\mathrm{RMS}}(g)\to0 prevents Gaussian probes from achieving Heisenberg scaling in the resource, while ∣ψt⟩|\psi_t\rangle retains it. Physically, the envelope allows large second moments without a comparable increase in the first moment, which limits Gaussian sensitivity.

As an example, consider the algebraically decaying envelope for n>Nmax⁡n>N_{\max},

gn=(Nmax⁡n)p,(18)g_n=\left(\frac{N_{\max}}{n}\right)^p, \tag*{(18)}

with fixed p>3p>3, so that the envelope has a finite second moment (∑n≥Nmax⁡n2gn<∞)(\sum_{n\ge N_{\max}}n^{2}g_n<\infty). Increasing Nmax⁡N_{\max} enlarges the free window while preserving the envelope profile as a function of n/Nmax⁡n/N_{\max}. Both Nˉ(g)\bar{N}(g) and NRMS(g)N_{\mathrm{RMS}}(g) grow proportionally to Nmax⁡p/(p−1)N_{\max}^{p/(p-1)}, as we show in Appendix D, and for sufficiently large Nmax⁡N_{\max} the right-hand side of Eq. (17) satisfies

NRMS2(g)8Nˉ(g)[Nˉ(g)+1]∼(p−2)28(p−1)(p−3).(19)\frac{N_{\mathrm{RMS}}^{2}(g)}{8\bar{N}(g)\bigl[\bar{N}(g)+1\bigr]}\sim\frac{(p-2)^{2}}{8(p-1)(p-3)}. \tag*{(19)}

This limit exceeds one for 3<p<2+8/73<p<2+\sqrt{8/7}, certifying that ∣ψt⟩|\psi_t\rangle outperforms every admissible Gaussian probe for sufficiently large Nmax⁡N_{\max}. Thus, Eq. (17) establishes a constant-factor advantage at fixed tail exponent, although it does not yield a scaling separation in NRMS(g)N_{\mathrm{RMS}}(g).

A second criterion can be derived when the envelope decays at least exponentially at large photon numbers. We define its asymptotic exponential rate and corresponding tail length by

ρg=lim sup⁡n→∞gn1/n,ℓg=−1/log⁡ρg,(20)\rho_g=\limsup_{n\to\infty}g_n^{1/n},\qquad\ell_g=-1/\log\rho_g, \tag*{(20)}

with 0<ρg<10<\rho_g<1. The condition ρg<1\rho_g<1 means that an exponential upper bound holds sufficiently far in the envelope tail. Algebraic tails, and subexponential tails more generally, have ρg=1\rho_g=1 and are not covered by this second condition.

For any admissible Gaussian probe, the squeezing parameter rr satisfies tanh⁡r≤ρg\tanh r\le\rho_g, see Appendix E, independently of displacement. The envelope thus limits squeezing even when it allows NRMS(g)N_{\mathrm{RMS}}(g) to grow. As we further show in Appendix E, this restriction yields an upper bound on the QFI of Gaussian states in terms of ℓg\ell_g and NRMS(g)N_{\mathrm{RMS}}(g),

Fθ(ψG)≤2ℓg2+4(1+2ℓg)NRMS(g).(21)\mathcal{F}_{\theta}(\psi_G)\le2\ell_g^{2}+4(1+2\ell_g)N_{\mathrm{RMS}}(g). \tag*{(21)}

In this bound, the first term comes from squeezing. It depends only on how fast the tail of the envelope decays, and not on NRMS(g)N_{\mathrm{RMS}}(g). The second term comes from displacement and grows linearly with NRMS(g)N_{\mathrm{RMS}}(g). Thus, increasing the budget at fixed tail decay, for example by enlarging the free window, allows at most linear growth of the Gaussian QFI with NRMS(g)N_{\mathrm{RMS}}(g).

Combining this with Eq. (14) gives

Fθ(ψt)Fθ(ψG)≥NRMS2(g)2ℓg2+4(1+2ℓg)NRMS(g).(22)\frac{\mathcal{F}_{\theta}(\psi_t)}{\mathcal{F}_{\theta}(\psi_G)}\ge\frac{N_{\mathrm{RMS}}^{2}(g)}{2\ell_g^{2}+4(1+2\ell_g)N_{\mathrm{RMS}}(g)}. \tag*{(22)}

Thus, at bounded ℓg\ell_g, the QFI of Gaussian probes is limited to O(NRMS)O(N_{\mathrm{RMS}}), whereas ∣ψt⟩|\psi_t\rangle achieves O(NRMS2)O(N_{\mathrm{RMS}}^{2}). More generally, ℓg/NRMS(g)→0\ell_g/N_{\mathrm{RMS}}(g)\to0 excludes Gaussian Heisenberg scaling. The advantage arises because the envelope permits increasing photon-number fluctuations while restricting Gaussian squeezing. To illustrate the distinction between the two criteria, consider the exponentially decaying envelope for n>Nmax⁡n>N_{\max},

gn=2−(n−Nmax⁡).(23)g_n=2^{-(n-N_{\max})}. \tag*{(23)}

As we show in Appendix § 2.13, the distribution tNmax⁡+j=2−jt_{N_{\max}+j}=2^{-j} gives: Nˉ(g)=Nmax⁡+2\bar{N}(g)=N_{\max}+2, NRMS2(g)=Nmax⁡2+4Nmax⁡+6N_{\mathrm{RMS}}^2(g)=N_{\max}^2+4N_{\max}+6. As Nmax⁡→∞N_{\max}\to\infty, the right-hand side of Eq. (17) approaches 1/81/8 and therefore does not certify an advantage. Instead, computing ρg=1/2\rho_g=1/2 and ℓg=1/ln⁡2\ell_g=1/\ln2, Eq. (22) certifies an advantage growing at least linearly with NRMS(g)N_{\mathrm{RMS}}(g). Indeed, as we detail in Appendix § 2.13, the best admissible Gaussian probe has Fθ(ψG)∼NRMS\mathcal{F}_\theta(\psi_G)\sim N_{\mathrm{RMS}}, whereas Fθ(ψt)∼NRMS2\mathcal{F}_\theta(\psi_t)\sim N_{\mathrm{RMS}}^2. The distinction between the two criteria arises because Eq. (17) does not capture the envelope’s restriction on squeezing.

Both bounds, Eqs. (17) and (22), compare probes under the same envelope and provide sufficient conditions for non-Gaussian advantage. A lower bound of at most one therefore does not rule out an advantage. These results show that preparation constraints can impose a fundamental separation between Gaussian and non-Gaussian metrology. In particular, at fixed exponential tail decay, Gaussian probes are limited to shot-noise-like scaling in NRMS(g)N_{\mathrm{RMS}}(g), while admissible non-Gaussian probes attain Heisenberg scaling under the same constraints.

Example: Universal optical gate set

We now apply the previous results to a concrete preparation model using the universal gate set G≡{D^,S^,R^,K^}\mathcal{G}\equiv\{\hat{D},\hat{S},\hat{R},\hat{K}\} of displacements, squeezing, rotations, and Kerr gates [27, 28] (see Appendix G for the proofs of the section). Circuit resources are quantified by the total squeezing rtot=∑i∣ri∣r_{\mathrm{tot}}=\sum_i|r_i| and displacement ∣αtot∣=∑j∣αj∣|\alpha_{\mathrm{tot}}|=\sum_j|\alpha_j|, which determine the envelope for Gaussian circuits. Rotations and Kerr gates preserve photon-number probabilities but alter phases relevant to subsequent squeezing and displacement. We show that the same envelope remains valid when these gates are interleaved in circuits without squeezing. For general circuits, we assume this property based on a heuristic argument in Appendix G. Table I includes only cases for which we show that the envelope remains valid.

Firstly, we compute the decay rate of the tail fixed by the resources, so that, for all n>Nmax⁡n>N_{\max}, the admissible probes satisfy:

qn≤ϵ(1−e−s∗)e−s∗(n−Nmax⁡−1),(24)q_n\le\epsilon\left(1-e^{-s_*}\right)e^{-s_*(n-N_{\max}-1)}, \tag*{(24)}

where 0<s∗<κ=ln⁡coth⁡rtot0<s_*<\kappa=\ln\coth r_{\mathrm{tot}}. The rate s∗s_* must be strictly smaller than κ\kappa, and it is found by minimizing NRMSN_{\mathrm{RMS}}. The threshold, in contrast, depends explicitly

on both the total displacement and squeezing applied. Requiring Eq. (24) for every probe gives:

Nmax⁡∼e2rtot∣αtot∣2.(25)N_{\max}\sim e^{2r_{\mathrm{tot}}}|\alpha_{\mathrm{tot}}|^2. \tag*{(25)}

Evaluating Eq. (11) for this envelope, we obtain the root-mean-square photon budget:

NRMS2=Nmax⁡2+ϵ[2Nmax⁡1−e−s∗+1+e−s∗(1−e−s∗)2].(26)N_{\mathrm{RMS}}^2=N_{\max}^2+\epsilon\left[\frac{2N_{\max}}{1-e^{-s_*}}+\frac{1+e^{-s_*}}{(1-e^{-s_*})^2}\right]. \tag*{(26)}

Hence, the resources that the setup can supply, ∣αtot∣|\alpha_{\mathrm{tot}}| and rtotr_{\mathrm{tot}}, bound the achievable QFI through NRMSN_{\mathrm{RMS}}. Table I compares representative admissible probes at equal NRMS(g)N_{\mathrm{RMS}}(g). The non-Gaussian probes scale quadratically, whereas the prefactor of the squeezed vacuum, 2/(ln⁡ϵ−1+2rtot)22/(\ln\epsilon^{-1}+2r_{\mathrm{tot}})^2, decreases with both the tolerance and the squeezing.

ProbeStateFθ\mathcal{F}_\theta
Squeezed vacuumS^(r)∣0⟩\hat{S}(r)|0\rangle≃2NRMS2/(ln⁡ϵ−1+2rtot)2\simeq2N_{\mathrm{RMS}}^2/(\ln\epsilon^{-1}+2r_{\mathrm{tot}})^2
Displaced cat∣0⟩+∣α⟩2(1+e−∣α∣2/2)\frac{|0\rangle+|\alpha\rangle}{\sqrt{2(1+e^{-|\alpha|^2/2})}}≃NRMS2\simeq N_{\mathrm{RMS}}^2
ON(∣0⟩+∣Nmax⁡⟩)/2(|0\rangle+|N_{\max}\rangle)/\sqrt{2}Nmax⁡2∼NRMS2N_{\max}^2\sim N_{\mathrm{RMS}}^2

Table 1. Quantum Fisher information of representative admissible probes. The comparison is therefore at equal resource NRMS(g)N_{\mathrm{RMS}}(g). All rows use s=s∗s=s_* at fixed tolerance ϵ\epsilon, and $\simeq denotes asymptotic equivalence along each path. ϵ\epsilon refers to the quantity of probability density above Nmax⁡N_{\max}.

Discussion

In this work, we incorporated preparation constraints into local optical phase estimation through upper envelopes on the probe’s photon-number probabilities. We defined a finite root-mean-square photon budget fixed by these constraints and derived a Heisenberg limit in this budget whose scaling is attainable. Divergent QFI at fixed mean photon number therefore requires an unbounded preparation budget. We also established sufficient conditions for non-Gaussian probes to outperform every admissible Gaussian state under the same envelope. In particular, for exponentially decaying envelopes with fixed decay length, Gaussian QFI grows at most linearly with the budget, whereas admissible non-Gaussian probes can attain quadratic scaling. Preparation constraints can thus make non-Gaussianity necessary to achieve Heisenberg scaling.

Using an optical preparation model, we illustrated how experimentally relevant resources can be translated into photon-number envelopes. This connects our framework to the capabilities of optical setups and provides sensitivity benchmarks in terms of available preparation resources. A natural next step is to identify preparation and measurement protocols that realize the predicted advantage in concrete devices and assess its robustness to photon loss.

Acknowledgments

The authors thank Martí Perarnau-Llobet, John Calsamiglia and Ricard Ravell Rodríguez for fruitful discussions, advice and feedback on the manuscript. This research was supported by the ATRAE Program (Grant ATR2024154621), funded by MI-CIU/AEI/10.13039/501100011033.

The authors acknowledge the use of generative AI tools (Claude and ChatGPT) for the purposes of literature search, language editing of the manuscript and assistance in deriving equations. All AI-assisted content was reviewed, verified, and revised by the authors, who take full responsibility for the content of this work.

References

On the heavy-tailed states

In this appendix, we show how to construct heavy-tailed probes with infinite QFI even at a finite mean photon number. To this end, we require a divergent photon-number variance while preserving a finite mean photon number. First, we impose a finite mean photon number,

∣ψγ⟩=N∑n=1∞1nγ∣n⟩⟹⟨ψγ∣n^∣ψγ⟩=N2∑n=1∞1n2γ−1.(27)|\psi_\gamma\rangle= \mathcal{N}\sum_{n=1}^{\infty}\frac{1}{n^\gamma}|n\rangle\Longrightarrow\langle\psi_\gamma|\hat{n}|\psi_\gamma\rangle= \mathcal{N}^2\sum_{n=1}^{\infty}\frac{1}{n^{2\gamma-1}}. \tag*{(27)}

The series ∑r=1∞r−μ\sum_{r=1}^{\infty}r^{-\mu} converges only if μ>1\mu>1, which here requires 2γ−1>12\gamma-1>1, that is, γ>1\gamma>1. Second, we require a divergent photon-number variance,

⟨ψγ∣n^2∣ψγ⟩=N2∑n=1∞1n2γ−2,(28)\langle\psi_\gamma|\hat{n}^2|\psi_\gamma\rangle= \mathcal{N}^2\sum_{n=1}^{\infty}\frac{1}{n^{2\gamma-2}}, \tag*{(28)}

which diverges for 2γ−2≤12\gamma-2\leq1, that is, γ≤32\gamma\leq\frac{3}{2}. Hence, the expansion in Eq. (5) with exponent γ∈(1,32]\gamma\in(1,\frac{3}{2}] yields a finite mean photon number and a divergent QFI.

On the details of asymmetrically superposed states

In this appendix, we study how the weight parameter ϵ\epsilon in Eq. (6) affects the QFI for a given excitation ∣ϕ⟩|\phi\rangle. Since the QFI in Eq. (7) is concave in ϵ2\epsilon^2, setting its derivative with respect to ϵ2\epsilon^2 to zero gives the maximum at

ϵ∗2=min⁡{1,⟨n^2⟩ϕ2⟨n^⟩ϕ2}.(29)\epsilon_*^2=\min\left\{1,\frac{\langle\hat{n}^2\rangle_\phi}{2\langle\hat{n}\rangle_\phi^2}\right\}. \tag*{(29)}

If ⟨n^2⟩ϕ≥2⟨n^⟩ϕ2\langle\hat{n}^2\rangle_\phi\geq2\langle\hat{n}\rangle_\phi^2 (excitations with broad number distributions), the optimal value is ϵ∗=1\epsilon_*=1 and the superposition reduces to ∣ϕ⟩|\phi\rangle. Whenever ⟨n^2⟩ϕ≈⟨n^⟩ϕ2\langle\hat{n}^2\rangle_\phi\approx\langle\hat{n}\rangle_\phi^2, exact for ∣ϕ⟩=∣N⟩|\phi\rangle=|N\rangle, the optimum lies at ϵ∗≈2−1/2\epsilon_*\approx2^{-1/2}, the balanced superposition. Since ⟨n^2⟩ϕ≥⟨n^⟩ϕ2\langle\hat{n}^2\rangle_\phi\geq\langle\hat{n}\rangle_\phi^2, we have 1/2≤ϵ∗2≤11/2\leq\epsilon_*^2\leq1 for every excitation, so the optimum never lies in the ϵ→0\epsilon\to0 regime that appears when ⟨n^⟩\langle\hat{n}\rangle is taken as the resource.

On the distribution attaining the RMS photon budget

In this appendix, we construct the distribution t∈Qgt\in\mathcal{Q}_g attaining the maximum in the RMS photon budget NRMS(g)N_{\mathrm{RMS}}(g), defined in Eq. (11). We also show that the same distribution maximizes the mean photon number. For simplicity, we use the Nmax⁡=1N_{\max}=1 notation, writing the envelope constraints as qn≤gnq_n\leq g_n for n≥1n\geq1 and leaving the vacuum population unconstrained. An arbitrary free window is included by setting gn=1g_n=1 for 1≤n≤Nmax⁡1\leq n\leq N_{\max}, so the argument remains general. Throughout, we assume ∑n≥1n2gn<∞\sum_{n\geq1}n^2g_n<\infty. First, consider the case ∑n≥1gn≤1\sum_{n\geq1}g_n\leq1. Take the distribution tt as

tn=gn(n≥1),t0=1−∑n≥1gn.(30)t_n=g_n\quad(n\geq1),\qquad t_0=1-\sum_{n\geq1}g_n. \tag*{(30)}

This distribution is normalized and admissible. For any q∈Qgq\in\mathcal{Q}_g and k∈{1,2}k\in\{1,2\},

∑n≥0nktn−∑n≥0nkqn=∑n≥1nk(gn−qn)≥0.(31)\sum_{n\geq0}n^k t_n-\sum_{n\geq0}n^kq_n=\sum_{n\geq1}n^k(g_n-q_n)\geq0. \tag*{(31)}

Thus, tt maximizes both the first and second moments, with N‾(g)=∑n≥1ngn\overline{N}(g)=\sum_{n\geq1}ng_n and NRMS2(g)=∑n≥1n2gnN_{\mathrm{RMS}}^2(g)=\sum_{n\geq1}n^2g_n. Equality requires qn=gnq_n=g_n for every n≥1n\geq1, while normalization fixes q0=t0q_0=t_0. Now consider ∑n≥1gn>1\sum_{n\geq1}g_n>1. Choose an integer m≥1m\geq1 satisfying

∑n>mgn≤1≤∑n≥mgn.(32)\sum_{n>m}g_n\leq1\leq\sum_{n\geq m}g_n. \tag*{(32)}

To show that such an mm exists, define the tail mass T(k):=∑n≥kgnT(k) := \sum_{n\ge k} g_n. It is non-increasing in kk, it starts at T(1)=∑n≥1gn>1T(1) = \sum_{n\ge1} g_n > 1, and T(k)→0T(k) \to0 as k→∞k \to\infty since ∑n≥1gn<∞\sum_{n\ge1} g_n < \infty. Hence, the set {k≥1:T(k)≥1}\{k \ge1 : T(k) \ge1\} is non-empty and bounded, and its largest element mm satisfies the previous inequality, i.e., T(m)≥1≥T(m+1)T(m) \ge1 \ge T(m + 1). Next, take the distribution tt as

tn={0,0≤n<m,1−∑j>mgj,n=m,gn,n>m.(33)t_n = \begin{cases} 0, & 0 \le n < m,\\ 1-\sum_{j>m}g_j, & n=m,\\ g_n, & n>m. \end{cases} \tag*{(33)}

The distribution fills the envelope above mm and places the remaining probability at mm. It is normalized and admissible, since Eq. (32) ensures 0≤tm≤gm0 \le t_m \le g_m. For any q∈Qgq \in\mathcal{Q}_g, normalization gives ∑n≥0(tn−qn)=0\sum_{n\ge0}(t_n-q_n)=0. Hence, for k∈{1,2}k \in\{1,2\},

∑n≥0nk(tn−qn)=∑n≥0(nk−mk)(tn−qn)=∑n>m(nk−mk)(gn−qn)+∑0≤n<m(mk−nk)qn≥0.(34)\begin{aligned} \sum_{n\ge0}n^k(t_n-q_n)=\sum_{n\ge0}(n^k-m^k)(t_n-q_n) \\ &=\sum_{n>m}(n^k-m^k)(g_n-q_n)+\sum_{0\le n<m}(m^k-n^k)q_n\ge0. \tag*{(34)} \end{aligned}

Both sums are non-negative. Therefore, tt again maximizes both moments, giving

N‾(g)=max⁡q∈QgEq[n^]=∑n≥0ntn=m+∑n>m(n−m)gn,(35)\overline{N}(g)=\max_{q\in\mathcal{Q}_g}\mathbb{E}_q[\hat{n}]=\sum_{n\ge0}nt_n=m+\sum_{n>m}(n-m)g_n, \tag*{(35)}
NRMS2(g)=∑n≥0n2tn=m2+∑n>m(n2−m2)gn.(36)N_{\mathrm{RMS}}^2(g)=\sum_{n\ge0}n^2t_n=m^2+\sum_{n>m}(n^2-m^2)g_n. \tag*{(36)}

Equality in Eq. (34) requires qn=gnq_n=g_n for n>mn>m and qn=0q_n=0 for n<mn<m; normalization then fixes qm=tmq_m=t_m. Thus, in either case, only q=tq=t attains equality.

On the algebraically decaying envelope example

In this appendix, we detail the computations for the algebraically decaying envelope example in § 2.3. We derive the asymptotic behavior of N‾(g)\overline{N}(g) and NRMS(g)N_{\mathrm{RMS}}(g) for the envelope of Eq. (18) and obtain the limit in Eq. (19). Throughout, p>3p>3 is fixed and o(1)o(1) denotes terms that vanish as Nmax⁡→∞N_{\max}\to\infty. Note that for n>Nmax⁡n>N_{\max} we have n2gn=Nmax⁡pn2−pn^2g_n=N_{\max}^p n^{2-p}, so the condition ∑n≥Nmax⁡n2gn<∞\sum_{n\ge N_{\max}}n^2g_n<\infty holds precisely because p>3p>3.

We first identify the distribution tt attaining both N‾(g)\overline{N}(g) and NRMS(g)N_{\mathrm{RMS}}(g). Since g1=1g_1=1 and the tail of the envelope is strictly positive, ∑n≥1gn>1\sum_{n\ge1}g_n>1, so tt is given by Eq. (33), with mm the largest integer satisfying T(m)≥1T(m)\ge1. The last level of the free window carries gNmax⁡=1g_{N_{\max}}=1; hence,

T(Nmax⁡)=1+∑n>Nmax⁡gn>1,(37)T(N_{\max})=1+\sum_{n>N_{\max}}g_n>1, \tag*{(37)}

and therefore m≥Nmax⁡m\ge N_{\max}. Thus, every level n>mn>m belongs to the algebraic tail, where gn=Nmax⁡pn−pg_n=N_{\max}^p n^{-p}, and also gm=(Nmax⁡/m)pg_m=(N_{\max}/m)^p, including at m=Nmax⁡m=N_{\max}, where both sides equal one.

From Eqs. (35) and (36), we can write N‾(g)\overline{N}(g) and NRMS(g)N_{\mathrm{RMS}}(g) in terms of mm and the tail sums

Sj:=∑n>mnjgn=Nmax⁡p∑n>mnj−p,j=0,1,2.(38)S_j:=\sum_{n>m}n^j g_n=N_{\max}^p\sum_{n>m}n^{j-p},\qquad j=0,1,2. \tag*{(38)}

Indeed, splitting ∑n>m(nk−mk)gn=Sk−mkS0\sum_{n>m}(n^k-m^k)g_n=S_k-m^kS_0, both moments read

N‾(g)=S1+m(1−S0),NRMS2(g)=S2+m2(1−S0).(39)\begin{aligned} \overline{N}(g)&=S_1+m(1-S_0),\\ N_{\mathrm{RMS}}^2(g)&=S_2+m^2(1-S_0). \tag*{(39)} \end{aligned}

To estimate the tail sums, note that for j≤2<p−1j\le2<p-1 the function f(x)=xj−pf(x)=x^{j-p} is decreasing, so that ∫nn+1f≤f(n)≤∫n−1nf\int_n^{n+1}f\le f(n)\le\int_{n-1}^n f. Summing over n≥m+1n\ge m+1 and using ∫k∞xj−pdx=kj+1−p/(p−j−1)\int_k^\infty x^{j-p}dx=k^{j+1-p}/(p-j-1), we obtain

Nmax⁡p(m+1)j+1−pp−j−1≤Sj≤Nmax⁡pmj+1−pp−j−1.(40)\frac{N_{\max}^p(m+1)^{j+1-p}}{p-j-1}\le S_j\le\frac{N_{\max}^p m^{j+1-p}}{p-j-1}. \tag*{(40)}

The two bounds differ by the factor (m/(m+1))p−j−1=1−(p−j−1)/m+O(m−2)(m/(m+1))^{p-j-1}=1-(p-j-1)/m+O(m^{-2}), which approaches one since m≥Nmax⁡→∞m\ge N_{\max}\to\infty. Hence,

Sj=Nmax⁡pmj+1−pp−j−1(1+O(1m))(41)S_j=\frac{N_{\max}^{p}m^{j+1-p}}{p-j-1}\left(1+O\left(\frac{1}{m}\right)\right) \tag*{(41)}

Next, we determine how mm grows with Nmax⁡N_{\max}. Since T(m+1)=S0T(m+1)=S_0 and T(m)=S0+gmT(m)=S_0+g_m, Eq. (32) reads

1−gm≤S0≤1.(42)1-g_m\le S_0\le1. \tag*{(42)}

Combining S0≤1S_0\le1 with the lower bound in Eq. (40) for j=0j=0 gives (m+1)p−1≥Nmax⁡p/(p−1)(m+1)^{p-1}\ge N_{\max}^{p}/(p-1), that is,

m+1Nmax⁡≥(Nmax⁡p−1)1/(p−1)→Nmax⁡→∞∞.(43)\frac{m+1}{N_{\max}}\ge\left(\frac{N_{\max}}{p-1}\right)^{1/(p-1)} \xrightarrow[N_{\max}\to\infty]{}\infty. \tag*{(43)}

Thus, m/Nmax⁡→∞m/N_{\max}\to\infty and gm=(Nmax⁡/m)p→0g_m=(N_{\max}/m)^p\to0. Then, Eq. (42) yields S0→1S_0\to1. Moreover, the integral bounds on T(m)T(m) and T(m+1)T(m+1) give ∣m−(Nmax⁡p/(p−1))1/(p−1)∣≤1\left|m-(N_{\max}^{p}/(p-1))^{1/(p-1)}\right|\le1, which sharpens the asymptotic corrections. Eq. (41) with j=0j=0 then fixes mm,

Nmax⁡p=(p−1)mp−1(1+O(m−1)),N_{\max}^{p}=(p-1)m^{p-1}(1+O(m^{-1})),
m=(p−1)−1p−1Nmax⁡pp−1(1+O(Nmax⁡−pp−1)).(44)m=(p-1)^{-\frac{1}{p-1}}N_{\max}^{\frac{p}{p-1}}\left(1+O\left(N_{\max}^{-\frac{p}{p-1}}\right)\right). \tag*{(44)}

Writing Nmax⁡pmj+1−p=mjNmax⁡pm1−pN_{\max}^{p}m^{j+1-p}=m^jN_{\max}^{p}m^{1-p} and inserting the first line of Eq. (44) into Eq. (41), we find

S1=p−1p−2m(1+O(m−1)),S_1=\frac{p-1}{p-2}m(1+O(m^{-1})),
S2=p−1p−3m2(1+O(m−1)).(45)S_2=\frac{p-1}{p-3}m^2(1+O(m^{-1})). \tag*{(45)}

We can now evaluate Nˉ(g)\bar{N}(g) and NRMS(g)N_{\mathrm{RMS}}(g). By Eq. (42), 0≤1−S0≤gm=O(m−1)0\le1-S_0\le g_m=O(m^{-1}), so the terms m(1−S0)m(1-S_0) and m2(1−S0)m^2(1-S_0) in Eq. (39) give relative corrections of order O(m−1)O(m^{-1}) compared with S1∝mS_1\propto m and S2∝m2S_2\propto m^2. Therefore,

Nˉ(g)=p−1p−2m(1+O(m−1)),\bar{N}(g)=\frac{p-1}{p-2}m(1+O(m^{-1})),
NRMS2(g)=p−1p−3m2(1+O(m−1)).(46)N_{\mathrm{RMS}}^2(g)=\frac{p-1}{p-3}m^2(1+O(m^{-1})). \tag*{(46)}

and, using Eq. (44), we see that both quantities scale as Nmax⁡p/(p−1)N_{\max}^{p/(p-1)} for Nmax⁡≫1N_{\max}\gg1,

Nˉ(g)=(p−1)p−2p−1p−2Nmax⁡pp−1(1+O(Nmax⁡−pp−1)),\bar{N}(g)=\frac{(p-1)^{\frac{p-2}{p-1}}}{p-2}N_{\max}^{\frac{p}{p-1}}\left(1+O\left(N_{\max}^{-\frac{p}{p-1}}\right)\right),
NRMS(g)=(p−1)p−32(p−1)p−3Nmax⁡pp−1(1+O(Nmax⁡−pp−1)).(47)N_{\mathrm{RMS}}(g)=\frac{(p-1)^{\frac{p-3}{2(p-1)}}}{\sqrt{p-3}}N_{\max}^{\frac{p}{p-1}}\left(1+O\left(N_{\max}^{-\frac{p}{p-1}}\right)\right). \tag*{(47)}

This growth, faster than Nmax⁡N_{\max}, reflects that tt leaves the whole free window empty and places its weight deep in the tail, since m/Nmax⁡→∞m/N_{\max}\to\infty. Since tn=0t_n=0 for n<mn<m, we have Nˉ(g)≥m≥Nmax⁡\bar{N}(g)\ge m\ge N_{\max}, so 1/Nˉ(g)→01/\bar{N}(g)\to0 and Nˉ(g)+1=Nˉ(g)[1+O(m−1)]\bar{N}(g)+1=\bar{N}(g)[1+O(m^{-1})].

Inserting Eq. (46) into the right-hand side of Eq. (17), the factors m2m^2 cancel and

NRMS2(g)8Nˉ(g)[Nˉ(g)+1]=(p−1)/(p−3)8(p−1)2/(p−2)2[1+O(Nmax⁡−pp−1)]\frac{N_{\mathrm{RMS}}^2(g)}{8\bar{N}(g)[\bar{N}(g)+1]}= \frac{(p-1)/(p-3)}{8(p-1)^2/(p-2)^2} \left[1+O\left(N_{\max}^{-\frac{p}{p-1}}\right)\right]
∼(p−2)28(p−1)(p−3).(48)\qquad\sim\frac{(p-2)^2}{8(p-1)(p-3)}. \tag*{(48)}

which is Eq. (19) for Nmax⁡≫1N_{\max}\gg1. This limit exceeds one when (p−2)2>8(p−1)(p−3)(p-2)^2>8(p-1)(p-3). Writing u=p−2>1u=p-2>1, so that (p−1)(p−3)=u2−1(p-1)(p-3)=u^2-1, the condition becomes u2>8(u2−1)u^2>8(u^2-1), i.e., u2<8/7u^2<8/7. Hence, Eq. (17) certifies a non-Gaussian advantage for 3<p<2+8/7≈3.0693<p<2+\sqrt{8/7}\approx3.069 and sufficiently large Nmax⁡N_{\max}.

On the bound on the QFI for Gaussian states in terms of tail decay

In this appendix, we show that every admissible pure Gaussian probe satisfies tanh⁡r≤ρg\tanh r \le\rho_g, and we derive the Gaussian bound in Eq. (21). Throughout, 0<ρg<10 < \rho_g < 1.

Up to a global phase, every pure Gaussian state can be written as ∣ψG⟩=D(α)S(ξ)∣0⟩|\psi_G\rangle= D(\alpha)S(\xi)|0\rangle with ξ=reiφ\xi= r e^{i\varphi} and r≥0r \ge0 [8, 29]. A phase rotation e−iφn^e^{-i\varphi\hat{n}} leaves the photon-number distribution unchanged and only shifts the phases of α\alpha and ξ\xi, so we may take ξ=r\xi= r real. We write nr:=sinh⁡2rn_r := \sinh^2 r, β:=tanh⁡r∈[0,1)\beta:= \tanh r \in[0,1), and α=u+iv\alpha= u + iv, and split the displacement along the anti-squeezed and squeezed quadratures, ∣α∣2=d++d−|\alpha|^2 = d_+ + d_- with d+=v2d_+ = v^2 and d−=u2d_- = u^2 (for real ξ\xi, S^(ξ)\hat{S}(\xi) squeezes the real quadrature, as in Eq. (75)). The photon-number mean and variance are then [8, 29]

⟨n^⟩ψG=nr+d++d−,\langle\hat{n}\rangle_{\psi_G} = n_r + d_+ + d_-,
Var⁡ψG(n^)=2nr(nr+1)+e2rd++e−2rd−.(49)\operatorname{Var}_{\psi_G}(\hat{n}) = 2n_r(n_r + 1) + e^{2r}d_+ + e^{-2r}d_-. \tag*{(49)}

To compare the photon-number distribution qn=∣⟨n∣ψG⟩∣2q_n = |\langle n|\psi_G\rangle|^2 with the envelope, we use its photon-number generating function,

G(z):=∑n≥0qnzn=⟨ψG∣zn^∣ψG⟩,z≥0.(50)G(z) := \sum_{n\ge0} q_n z^n = \langle\psi_G|z^{\hat{n}}|\psi_G\rangle,\qquad z \ge0. \tag*{(50)}

This is a power series whose coefficients are the photon-number probabilities. It satisfies G(1)=1G(1) = 1, and its derivatives at z=1z = 1 give the factorial moments of n^\hat{n}. Here, we only need its radius of convergence RR, which by the Cauchy–Hadamard formula satisfies 1/R=lim sup⁡n→∞qn1/n1/R = \limsup_{n\to\infty} q_n^{1/n}. The radius of convergence therefore measures how fast qnq_n decays at large photon numbers, in the same way as ρg\rho_g in Eq. (20) measures the decay of the envelope. Comparing RR with 1/ρg1/\rho_g thus compares the tail of a Gaussian distribution with the tail allowed by the envelope. For a pure Gaussian state, GG is known in closed form [30, 31],

G(z)=1−β21−β2z2exp⁡[d+(1−β)(z−1)1−βz+d−(1+β)(z−1)1+βz],(51)G(z) = \sqrt{\frac{1-\beta^2}{1-\beta^2z^2}}\exp\left[\frac{d_+(1-\beta)(z-1)}{1-\beta z}+\frac{d_-(1+\beta)(z-1)}{1+\beta z}\right], \tag*{(51)}

where the series converges for 0≤z<1/β0 \le z < 1/\beta.

We first determine the radius of convergence of GG. Suppose first that β>0\beta> 0. For 1≤z<1/β1 \le z < 1/\beta, both terms in the exponent of Eq. (51) are non-negative; hence,

G(z)≥1−β21−β2z2.(52)G(z) \ge\sqrt{\frac{1-\beta^2}{1-\beta^2z^2}}. \tag*{(52)}

The right-hand side diverges as zz approaches 1/β1/\beta from below. Thus, the series converges below 1/β1/\beta but not at 1/β1/\beta, so its radius of convergence is exactly R=1/βR = 1/\beta, and the Cauchy–Hadamard formula gives

lim sup⁡n→∞qn1/n=β=tanh⁡r,(53)\limsup_{n\to\infty}q_n^{1/n} = \beta= \tanh r, \tag*{(53)}

i.e., the photon-number distribution of a pure Gaussian state decays at the exponential rate ln⁡coth⁡r\ln\coth r, in agreement with κ\kappa in Eq. (24). An admissible probe satisfies qn≤gnq_n \le g_n for n>Nmax⁡n > N_{\max}; hence, qn1/n≤gn1/nq_n^{1/n} \le g_n^{1/n}. Taking the limit superior and using Eq. (20),

tanh⁡r≤ρg,(54)\tanh r \le\rho_g, \tag*{(54)}

which also holds trivially for β=0\beta= 0. This restriction holds independently of the displacement and of Nmax⁡N_{\max}, since it only involves the asymptotic decay of the envelope.

Next, we derive an upper bound on the QFI of Gaussian states in terms of the tail decay. The functions x2/(1−x2)x^2/(1-x^2) and (1+x)/(1−x)(1+x)/(1-x) are increasing on 0≤x<10 \le x < 1, and nr=β2/(1−β2)n_r = \beta^2/(1-\beta^2), e2r=(1+β)/(1−β)e^{2r} = (1+\beta)/(1-\beta). Hence,

nr≤ng:=ρg21−ρg2,e2r≤τg:=1+ρg1−ρg.(55)n_r \le n_g := \frac{\rho_g^2}{1-\rho_g^2},\qquad e^{2r} \le\tau_g := \frac{1+\rho_g}{1-\rho_g}. \tag*{(55)}

Moreover, since the photon-number distribution of ∣ψG⟩|\psi_G\rangle is admissible, Eq. (49) gives d++d−≤⟨n^⟩ψG≤Nˉ(g)d_+ + d_- \le\langle\hat{n}\rangle_{\psi_G} \le\bar{N}(g). Inserting these bounds into the QFI for phase estimation, Eq. (3), and using e−2r≤e2re^{-2r} \le e^{2r}, we obtain

Fθ(ψG)≤8nr(nr+1)+4e2r(d++d−)\mathcal{F}_{\theta}(\psi_G) \le8n_r(n_r+1) + 4e^{2r}(d_+ + d_-)
≤8ng(ng+1)+4τgNˉ(g).(56)\le8n_g(n_g+1) + 4\tau_g\bar{N}(g). \tag*{(56)}

The first term is the squeezing contribution and the second is the displacement contribution. Finally, we express these constants through the tail length. Writing ρg=e−1/ℓg\rho_g=e^{-1/\ell_g} and using the elementary inequalities sinh⁡x≥x\sinh x\ge x and ex−1≥xe^x-1\ge x,

8ng(ng+1)=8ρg2(1−ρg2)2=2sinh⁡2(1/ℓg)≤2ℓg2,(57)8n_g(n_g+1)=\frac{8\rho_g^2}{(1-\rho_g^2)^2}=\frac{2}{\sinh^2(1/\ell_g)}\le2\ell_g^2, \tag*{(57)}
τg=1+2e1/ℓg−1≤1+2ℓg.(58)\tau_g=1+\frac{2}{e^{1/\ell_g}-1}\le1+2\ell_g. \tag*{(58)}

Inserting the bounds in Eq. (E9) into Eq. (56) and using N‾(g)≤NRMS(g)\overline{N}(g)\le N_{\mathrm{RMS}}(g), which holds because tt attains both budgets and Et[n^]2≤Et[n^2]=NRMS2(g)\mathbb{E}_t[\hat n]^2\le\mathbb{E}_t[\hat n^2]=N_{\mathrm{RMS}}^2(g), we obtain

Fθ(ψG)≤2ℓg2+4(1+2ℓg)N‾(g)\mathcal{F}_{\theta}(\psi_G)\le2\ell_g^2+4(1+2\ell_g)\overline{N}(g)
≤2ℓg2+4(1+2ℓg)NRMS(g),(59)\le2\ell_g^2+4(1+2\ell_g)N_{\mathrm{RMS}}(g), \tag*{(59)}

which is Eq. (21).

On the exponentially decaying envelope example

In this appendix, we detail the computations for the exponentially decaying envelope example in § 2.3, Eq. (23). In particular, we compute both NRMS(g)N_{\mathrm{RMS}}(g) and N‾(g)\overline{N}(g), evaluate both criteria, Eqs. (17) and (22), and show that the best admissible Gaussian probe has a QFI linear in NRMS(g)N_{\mathrm{RMS}}(g).

We first identify the distribution tt attaining both N‾(g)\overline{N}(g) and NRMS(g)N_{\mathrm{RMS}}(g). Since g1=1g_1=1 and the tail of the envelope is strictly positive, ∑n≥1gn>1\sum_{n\ge1}g_n>1, so tt is given by Eq. (33), with mm the largest integer satisfying T(m)≥1T(m)\ge1. The tail masses just above the free window are

T(Nmax⁡+1)=∑j≥12−j=1,T(N_{\max}+1)=\sum_{j\ge1}2^{-j}=1,
T(Nmax⁡+2)=12;(60)T(N_{\max}+2)=\frac{1}{2}; \tag*{(60)}

hence, m=Nmax⁡+1m=N_{\max}+1. The remaining probability placed at mm is tm=1−∑n>mgn=1/2=gmt_m=1-\sum_{n>m}g_n=1/2=g_m, so tt fills the envelope exactly above Nmax⁡N_{\max},

tNmax⁡+j=2−j(j≥1),t_{N_{\max}+j}=2^{-j}\quad(j\ge1),
tn=0(0≤n≤Nmax⁡).(61)t_n=0\quad(0\le n\le N_{\max}). \tag*{(61)}

The free window, including the vacuum, is left unoccupied, and all the probability lies in the tail.

To evaluate the moments of tt, we use the geometric sums ∑j≥1xj=x/(1−x)\sum_{j\ge1}x^j=x/(1-x), ∑j≥1jxj=x/(1−x)2\sum_{j\ge1}jx^j=x/(1-x)^2, and ∑j≥1j2xj=x(1+x)/(1−x)3\sum_{j\ge1}j^2x^j=x(1+x)/(1-x)^3, obtained by differentiating the first one. At x=1/2x=1/2, they give

∑j≥12−j=1,∑j≥1j2−j=2,∑j≥1j22−j=6.(62)\sum_{j\ge1}2^{-j}=1,\qquad\sum_{j\ge1}j2^{-j}=2,\qquad\sum_{j\ge1}j^22^{-j}=6. \tag*{(62)}

Expanding (Nmax⁡+j)k(N_{\max}+j)^k for k=1,2k=1,2 and using Eq. (62),

N‾(g)=∑j≥1(Nmax⁡+j)2−j=Nmax⁡+2,\overline{N}(g)=\sum_{j\ge1}(N_{\max}+j)2^{-j}=N_{\max}+2,
NRMS2(g)=∑j≥1(Nmax⁡+j)22−j=Nmax⁡2+4Nmax⁡+6.(63)N_{\mathrm{RMS}}^2(g)=\sum_{j\ge1}(N_{\max}+j)^22^{-j}=N_{\max}^2+4N_{\max}+6. \tag*{(63)}

Since (Nmax⁡+2)2≤NRMS2(g)≤(Nmax⁡+3)2(N_{\max}+2)^2\le N_{\mathrm{RMS}}^2(g)\le(N_{\max}+3)^2, the RMS budget and the threshold differ only by a constant,

Nmax⁡+2≤NRMS(g)≤Nmax⁡+3.(64)N_{\max}+2\le N_{\mathrm{RMS}}(g)\le N_{\max}+3. \tag*{(64)}

Moreover, Eq. (14) gives the QFI of the probe ∣ψt⟩|\psi_t\rangle exactly,

Fθ(ψt)=2NRMS2(g)−N‾2(g)=Nmax⁡2+4Nmax⁡+8.(65)\mathcal{F}_{\theta}(\psi_t)=2N_{\mathrm{RMS}}^2(g)-\overline{N}^2(g)=N_{\max}^2+4N_{\max}+8. \tag*{(65)}

Next, we evaluate the mean-based criterion. Inserting Eq. (63) into the right-hand side of Eq. (17), we obtain

NRMS2(g)8N‾(g)[N‾(g)+1]=Nmax⁡2+4Nmax⁡+68(Nmax⁡2+5Nmax⁡+6)<18,(66)\frac{N_{\mathrm{RMS}}^{2}(g)}{8\overline{N}(g)\left[\overline{N}(g)+1\right]}=\frac{N_{\max}^{2}+4N_{\max}+6}{8\left(N_{\max}^{2}+5N_{\max}+6\right)}<\frac{1}{8}, \tag*{(66)}

where the inequality holds for every Nmax⁡≥1N_{\max}\geq1, and the left-hand side tends to 1/81/8 as Nmax⁡→∞N_{\max}\to\infty. Thus, Eq. (17) never certifies an advantage for this envelope: the mean of tt grows as fast as its RMS value, so the mean-based Gaussian bound, Eq. (16), also grows quadratically.

We now turn to the tail-decay-based criterion, Eq. (22). For n>Nmax⁡n>N_{\max}, gn1/n=2−(n−Nmax⁡)/n=2−1+Nmax⁡/ng_{n}^{1/n}=2^{-(n-N_{\max})/n}=2^{-1+N_{\max}/n}, which converges to 1/21/2 as n→∞n\to\infty. Hence, the limit superior in Eq. (20) is a limit, and

ρg=12,ℓg=1ln⁡2,(67)\rho_g=\frac{1}{2},\qquad\ell_g=\frac{1}{\ln2}, \tag*{(67)}

independently of Nmax⁡N_{\max}. Inserting these values into Eqs. (21) and (22), we obtain

Fθ(ψG)≤2ln⁡22+4(1+2ln⁡2)NRMS(g),(68)F_\theta(\psi_G)\leq\frac{2}{\ln^2 2}+4\left(1+\frac{2}{\ln2}\right)N_{\mathrm{RMS}}(g), \tag*{(68)}
Fθ(ψt)Fθ(ψG)≥NRMS2(g)2/ln⁡22+4(1+2/ln⁡2)NRMS(g).(69)\frac{F_\theta(\psi_t)}{F_\theta(\psi_G)}\geq\frac{N_{\mathrm{RMS}}^2(g)}{2/\ln^2 2+4\left(1+2/\ln2\right)N_{\mathrm{RMS}}(g)}. \tag*{(69)}

The right-hand side of Eq. (69) grows linearly with NRMS(g)N_{\mathrm{RMS}}(g). Numerically, the denominator is 4.16+15.54NRMS(g)4.16+15.54N_{\mathrm{RMS}}(g), so the ratio exceeds one for NRMS(g)>15.805N_{\mathrm{RMS}}(g)>15.805, which Eq. (64) guarantees for Nmax⁡≥14N_{\max}\geq14.

Finally, we show that the Gaussian bound in Eq. (68) is attained up to constants, so that the separation is genuinely one of scaling. Consider a coherent state ∣α⟩|\alpha\rangle with A:=∣α∣2≤Nmax⁡/2A:=|\alpha|^2\leq N_{\max}/2, whose photon-number distribution is qn=e−AAn/n!q_n=e^{-A}A^n/n!. The levels 1≤n≤Nmax⁡1\leq n\leq N_{\max} lie in the free window and impose no condition. For n≥1n\geq1, maximizing e−AAne^{-A}A^n over AA (the maximum lies at A=nA=n) and using Stirling’s lower bound, n!≥2πn(n/e)nn!\geq\sqrt{2\pi n}(n/e)^n, we obtain

qn≤e−nnnn!≤12πn<12.(70)q_n\leq\frac{e^{-n}n^n}{n!}\leq\frac{1}{\sqrt{2\pi n}}<\frac{1}{2}. \tag*{(70)}

For the tail levels, the ratio of consecutive populations is qNmax⁡+k/qNmax⁡+k−1=A/(Nmax⁡+k)≤1/2q_{N_{\max}+k}/q_{N_{\max}+k-1}=A/(N_{\max}+k)\leq1/2 for k≥2k\geq2, since A≤Nmax⁡/2A\leq N_{\max}/2. Combining this with Eq. (70) at n=Nmax⁡+1n=N_{\max}+1, we find, for every j≥1j\geq1,

qNmax⁡+j≤qNmax⁡+12−(j−1)<2−j=gNmax⁡+j,(71)q_{N_{\max}+j}\leq q_{N_{\max}+1}2^{-(j-1)}<2^{-j}=g_{N_{\max}+j}, \tag*{(71)}

so the state is admissible. Its QFI is Fθ=4Var⁡(n^)=4AF_\theta=4\operatorname{Var}(\hat{n})=4A, and taking A=Nmax⁡/2A=N_{\max}/2 gives Fθ=2Nmax⁡≥2NRMS(g)−6F_\theta=2N_{\max}\geq2N_{\mathrm{RMS}}(g)-6 by Eq. (64). Together with Eq. (68), the best admissible Gaussian probe, ∣ψG∗⟩|\psi_G^*\rangle, satisfies

Fθ(ψG∗)≥2NRMS(g)−6,F_\theta(\psi_G^*)\geq2N_{\mathrm{RMS}}(g)-6,
Fθ(ψG∗)≤2ln⁡22+4(1+2ln⁡2)NRMS(g),(72)F_\theta(\psi_G^*)\leq\frac{2}{\ln^2 2}+4\left(1+\frac{2}{\ln2}\right)N_{\mathrm{RMS}}(g), \tag*{(72)}

that is, Fθ(ψG∗)∼NRMS(g)F_\theta(\psi_G^*)\sim N_{\mathrm{RMS}}(g), whereas Eqs. (64) and (65) give Fθ(ψt)∼NRMS2(g)F_\theta(\psi_t)\sim N_{\mathrm{RMS}}^2(g). Enlarging the free window increases the RMS budget, but the fixed tail decay keeps tanh⁡r≤1/2\tanh r\leq1/2, so admissible Gaussian probes can only profit from it through displacement, which yields a QFI linear in the resource.

On the details of the example

In this appendix, we derive the results of the example in Sec. VI. First, we compute the tail decay rate imposed by the setup. Second, we find the threshold Nmax⁡N_{\max} that bounds the effective space in which the accessible states are defined. Next, we establish the relation between Nmax⁡N_{\max} and NRMSN_{\mathrm{RMS}} for this setup. Finally, we derive the scaling of the quantum Fisher information with Nmax⁡N_{\max} for different classes of states.

Shape of the envelope

To find the decay of the photon-number distribution generated by the universal gate set G\mathcal{G}, we first compute the moment-generating function of the state,

M(s)=∑mqmesm.(73)M(s)=\sum_m q_m e^{sm}. \tag*{(73)}

For a decay qm∝e−κmq_m \propto e^{-\kappa m}, M(s)M(s) converges only for s<κs < \kappa. Since every term of the sum is positive, each term is bounded by the whole sum, qmesm≤M(s)q_m e^{sm} \le M(s). Defining Ns=s−1ln⁡M(s)N_s = s^{-1}\ln M(s), we can rewrite this bound as

qm≤e−s(m−Ns),(74)q_m \le e^{-s(m-N_s)}, \tag*{(74)}

which holds for every mm and every ss at which M(s)M(s) is finite. The supremum κ\kappa of such ss gives the asymptotic decay rate of qmq_m. For a squeezed state, the coefficients are

c2n=1cosh⁡r(2n)!2nn!(−eiϑtanh⁡r)n,c2n+1=0.(75)c_{2n} = \frac{1}{\sqrt{\cosh r}}\frac{\sqrt{(2n)!}}{2^n n!}\left(-e^{i\vartheta}\tanh r\right)^n,\qquad c_{2n+1}=0. \tag*{(75)}

The moment-generating function then reads

M(s)=1cosh⁡r∑n(2nn)14n(e2stanh⁡2r)n.(76)M(s)=\frac{1}{\cosh r}\sum_n \binom{2n}{n}\frac{1}{4^n}\left(e^{2s}\tanh^2 r\right)^n. \tag*{(76)}

This series converges only if s<ln⁡coth⁡rs < \ln\coth r, which fixes the supremum of the decay rate. The displacement multiplies the coefficients by a Hermite-polynomial factor [30] that grows more slowly than any exponential. This factor leaves the radius of convergence of M(s)M(s), and hence κ\kappa, unchanged [Eq. (E5)]. When the displacement has a component $d_+ > 0$ along the anti-squeezed quadrature, however, the generating function in Eq. (E3) diverges as $\exp[d_+(1-\beta)^2/\beta(1-\beta z)]$ for $z \to 1/\beta$. A bound $q_n \le C e^{-\kappa n}$ for all large $n$ would instead allow at most a simple pole, $G(z) \le \sum_{n\le N}q_nz^n+C(\beta z)^{N+1}/(1-\beta z)$. Consequently, no such bound with finite $C$ holds, and the decay rate of the envelope must satisfy $0 < s < \kappa$ strictly. For any such $s$, $M(s)$ is finite and, for a displaced squeezed state, follows in closed form from {xref}Eq. (E3)` at z=esz=e^s.

We therefore fix 0<s<κ0 < s < \kappa and define gn=Ae−s(n−Nmax⁡−1)g_n=Ae^{-s(n-N_{\max}-1)} for n>Nmax⁡n>N_{\max}. Requiring that the probability weight above Nmax⁡N_{\max} equal a tolerance ϵ\epsilon gives

∑n>Nmax⁡gn=A1−e−s=ϵ⇒A=ϵ(1−e−s).(77)\sum_{n>N_{\max}}g_n=\frac{A}{1-e^{-s}}=\epsilon\quad\Rightarrow\quad A=\epsilon(1-e^{-s}). \tag*{(77)}

Every probe then satisfies qn≤M(s)e−sn≤gnq_n\le M(s)e^{-sn}\le g_n for n>Nmax⁡n>N_{\max}, provided that M(s)e−s(Nmax⁡+1)≤ϵ(1−e−s)M(s)e^{-s(N_{\max}+1)}\le\epsilon(1-e^{-s}).

Threshold Nmax⁡N_{\max}

For Gaussian circuits, the threshold depends only on the displacement and squeezing gates. Each squeezing gate stretches phase-space distances by at most e∣ri∣e^{|r_i|}, and rotations preserve them. The final squeezing therefore satisfies ∣ξ′∣≤rtot|\xi'|\le r_{\mathrm{tot}}, and each displacement is amplified by at most ertote^{r_{\mathrm{tot}}}, so that the final displacement satisfies ∣α′∣≤ertot∣αtot∣|\alpha'|\le e^{r_{\mathrm{tot}}}|\alpha_{\mathrm{tot}}|. Since Eq. (51) increases with the squeezing and with the anti-squeezed displacement, the worst case is the aligned state.

Rotations and Kerr gates leave the photon-number distribution unchanged when they act, but they modify the phases on which subsequent squeezing and displacement operate, and hence the final distribution. For circuits with Kerr gates, we assume that this interleaving does not enlarge the envelope beyond that of the aligned Gaussian state with the same totals. A phase-space argument supports this assumption: Kerr gates preserve the angle-averaged Wigner function, squeezing enlarges phase-space distances by at most ere^r, and displacements shift them by at most ∣α∣|\alpha|. This argument is heuristic, because Kerr gates generate Wigner negativity, and tail probabilities of quasiprobability distributions are not monotone under phase-space maps. For circuits without squeezing, the assumption holds. Number-diagonal gates preserve the factorial moments ∥a^kψ∥2=⟨n^(n^−1)⋯(n^−k+1)⟩\|\hat{a}^k\psi\|^2=\langle\hat{n}(\hat{n}-1)\cdots(\hat{n}-k+1)\rangle, and D^(γ)=D^(γ+γ)\hat{D}(\gamma)=\hat{D}(\gamma+\gamma) gives ∥a^kD^(γ)ψ∥≤∑j(kj)∣γ∣k−j∥a^jD^(γ)ψ∥\|\hat{a}^k\hat{D}(\gamma)\psi\|\le\sum_j\binom{k}{j}|\gamma|^{k-j}\|\hat{a}^j\hat{D}(\gamma)\psi\|. By induction from the vacuum, ∥a^kψ∥≤∣αtot∣k\|\hat{a}^k\psi\|\le|\alpha_{\mathrm{tot}}|^k, and expanding G(z)G(z) in factorial moments gives, for z≥1z\ge1,

G(z)≤e∣αtot∣2(z−1),(78)G(z)\le e^{|\alpha_{\mathrm{tot}}|^2(z-1)}, \tag*{(78)}

the generating function of the coherent state ∣αtot⟩|\alpha_{\mathrm{tot}}\rangle. Since this coherent state belongs to the Gaussian class, these circuits satisfy the same envelope.

For the purpose of computing Nmax⁡N_{\max}, we can then describe the state-generation protocol as D(αeff)S(−rtot)∣0⟩D(\alpha_{\mathrm{eff}})S(-r_{\mathrm{tot}})|0\rangle, with

αeff≡ertotαtot=e∑i∣ri∣∑j∣αj∣.(79)\alpha_{\mathrm{eff}}\equiv e^{r_{\mathrm{tot}}}\alpha_{\mathrm{tot}}=e^{\sum_i|r_i|}\sum_j|\alpha_j|. \tag*{(79)}

In the anti-squeezed quadrature x^=(a^+a^†)/2\hat{x} = (\hat{a} + \hat{a}^{\dagger})/\sqrt{2}, this state has mean μ=2ertotαtot\mu= \sqrt{2}e^{r_{\mathrm{tot}}}\alpha_{\mathrm{tot}} and variance σ2=e2rtot/2\sigma^2 = e^{2r_{\mathrm{tot}}}/2, whereas the orthogonal quadrature has variance e−2rtot/2e^{-2r_{\mathrm{tot}}}/2. For strong anti-squeezing, e2rtot≫1e^{2r_{\mathrm{tot}}} \gg1, this quadrature dominates the photon number, n^≈x^2/2\hat{n} \approx\hat{x}^2/2, and the tail probability is approximately

P(n>Nmax⁡)≈12erfc⁡(2Nmax⁡−μ2σ)+12erfc⁡(2Nmax⁡+μ2σ).(80)P(n > N_{\max}) \approx\frac{1}{2}\operatorname{erfc}\left(\frac{\sqrt{2N_{\max}}-\mu}{\sqrt{2}\sigma}\right) + \frac{1}{2}\operatorname{erfc}\left(\frac{\sqrt{2N_{\max}}+\mu}{\sqrt{2}\sigma}\right). \tag*{(80)}

The second term never exceeds the first. Using 12erfc⁡(y)≃e−y2/(2πy)\frac{1}{2}\operatorname{erfc}(y) \simeq e^{-y^2}/(2\sqrt{\pi}y) for y≫1y \gg1 and setting the tail probability equal to ε\varepsilon, the argument of the first term satisfies y2=ln⁡ε−1+O(ln⁡ln⁡ε−1)y^2 = \ln\varepsilon^{-1} + O(\ln\ln\varepsilon^{-1}). To leading order in ln⁡ε−1\ln\varepsilon^{-1}, this gives

Nmax⁡≃e2rtot(∣αtot∣+12ln⁡ε−1)2.(81)N_{\max} \simeq e^{2r_{\mathrm{tot}}}\left(|\alpha_{\mathrm{tot}}|+\sqrt{\frac{1}{2}\ln\varepsilon^{-1}}\right)^2. \tag*{(81)}

This quantile bounds only the total tail weight, not each level, and therefore does not guarantee Eq. (24). The certified threshold, Eq. (25), imposes the condition of Sec. G 1 with M(s)M(s) given by Eq. (51) for the aligned state at z=esz=e^s,

ln⁡M(s)=12ln⁡1−β21−β2e2s+e2rtot∣αtot∣2(1−β)(es−1)1−βes.(82)\ln M(s) = \frac{1}{2}\ln\frac{1-\beta^2}{1-\beta^2e^{2s}} + e^{2r_{\mathrm{tot}}}|\alpha_{\mathrm{tot}}|^2\frac{(1-\beta)(e^s-1)}{1-\beta e^s}. \tag*{(82)}

Since NRMSN_{\mathrm{RMS}} diverges both as s→0s \to0 and as s→κs \to\kappa, it attains a minimum at some s∗∈(0,κ)s_* \in(0,\kappa). For large displacement, the resulting threshold remains close to Eq. (81).

Relation between NRMS and Nmax

To find NRMSN_{\mathrm{RMS}}, we maximize ⟨n2⟩\langle n^2\rangle under the envelope constraint using Appendix C. The tail masses are T(Nmax⁡)=1+ε≥1T(N_{\max}) = 1+\varepsilon\ge1 and T(Nmax⁡+1)=ε≤1T(N_{\max}+1) = \varepsilon\le1, so the pivot is m=Nmax⁡m=N_{\max}. The maximizing distribution therefore fills the tail completely, with total weight ε\varepsilon, and places the remaining probability 1−ε1-\varepsilon at Nmax⁡N_{\max}. Writing n=Nmax⁡+Jn=N_{\max}+J in the tail, with x=e−sx=e^{-s}, the excess J≥1J\ge1 follows the normalized geometric law hJ=(1−x)xJ−1h_J=(1-x)x^{J-1}, whose first two moments are

E[J]=11−x,E[J2]=1+x(1−x)2.(83)\mathbb{E}[J] = \frac{1}{1-x}, \qquad\mathbb{E}[J^2] = \frac{1+x}{(1-x)^2}. \tag*{(83)}

The RMS photon budget then reads

NRMS2=(1−ε)Nmax⁡2+εE[(Nmax⁡+J)2]=Nmax⁡2+ε(2Nmax⁡E[J]+E[J2])=Nmax⁡2+ε[2Nmax⁡1−e−s+1+e−s(1−e−s)2].(84)\begin{aligned} N_{\mathrm{RMS}}^2 &= (1-\varepsilon)N_{\max}^2+\varepsilon\mathbb{E}\left[(N_{\max}+J)^2\right] \\ &= N_{\max}^2+\varepsilon\left(2N_{\max}\mathbb{E}[J]+\mathbb{E}[J^2]\right) \\ &= N_{\max}^2+\varepsilon\left[\frac{2N_{\max}}{1-e^{-s}}+\frac{1+e^{-s}}{(1-e^{-s})^2}\right]. \tag*{(84)} \end{aligned}

Scaling of the representative probes

First, we compute the scaling of the ON state. This state has no probability weight above Nmax⁡N_{\max}, so its variance and QFI scale with Nmax⁡N_{\max} as

max⁡N∈{0,…,Nmax⁡}Var⁡(∣0⟩+∣N⟩)/2(n^)=Nmax⁡24⟹Fθ=Nmax⁡2.(85)\max_{N\in\{0,\ldots,N_{\max}\}}\operatorname{Var}_{(|0\rangle+|N\rangle)/\sqrt{2}}(\hat{n})=\frac{N_{\max}^2}{4} \quad\Longrightarrow\quad F_\theta=N_{\max}^2. \tag*{(85)}

Since NRMS≃Nmax⁡N_{\mathrm{RMS}}\simeq N_{\max} along this path, this gives Fθ≃NRMS2F_\theta\simeq N_{\mathrm{RMS}}^2. We use this state only as a benchmark and do not analyze its preparation within G\mathcal{G}.

Similarly, consider the displaced cat state ∣C⟩=(∣0⟩+∣α⟩)/2(1+e−A/2)|C\rangle=(|0\rangle+|\alpha\rangle)/\sqrt{2(1+e^{-A/2})}, with A=∣α∣2A=|\alpha|^2, whose normalization includes the overlap ⟨0∣α⟩=e−A/2\langle0|\alpha\rangle=e^{-A/2}. Its first two moments are

⟨n^⟩C=A2(1+e−A/2),⟨n^2⟩C=A2+A2(1+e−A/2).(86)\langle\hat{n}\rangle_C=\frac{A}{2(1+e^{-A/2})}, \qquad\langle\hat{n}^2\rangle_C=\frac{A^2+A}{2(1+e^{-A/2})}. \tag*{(86)}

so that

Fθ=4Var⁡C(n^)=A2+2A+O(A2e−A/2).(87)\mathcal{F}_{\theta}=4\operatorname{Var}_{C}(\hat{n})=A^{2}+2A+O\left(A^{2}e^{-A/2}\right). \tag*{(87)}

For n≥1n \ge1, its photon-number probabilities are qn=pn(A)/[2(1+e−A/2)]≤pn(A)q_{n}=p_{n}(A)/[2(1+e^{-A/2})]\le p_{n}(A), where pn(A)=e−AAn/n!p_{n}(A)=e^{-A}A^{n}/n! are those of the coherent state ∣α⟩|\alpha\rangle. The cat state therefore satisfies the envelope whenever ∣α⟩|\alpha\rangle does. Taking ∣α∣=∣αtot∣|\alpha|=|\alpha_{\mathrm{tot}}|, the coherent state belongs to the Gaussian class, and Eq. (25) gives, for large displacement, Nmax⁡≃e2rtot∣αtot∣2N_{\max}\simeq e^{2r_{\mathrm{tot}}}|\alpha_{\mathrm{tot}}|^{2} for ∣αtot∣2≫ln⁡ε−1|\alpha_{\mathrm{tot}}|^{2}\gg\ln\varepsilon^{-1}. Hence Fθ≃e−4rtotNmax⁡2≃e−4rtotNRMS2\mathcal{F}_{\theta}\simeq e^{-4r_{\mathrm{tot}}}N_{\max}^{2}\simeq e^{-4r_{\mathrm{tot}}}N_{\mathrm{RMS}}^{2}, which reduces to Fθ≃NRMS2\mathcal{F}_{\theta}\simeq N_{\mathrm{RMS}}^{2} when the setup supplies no squeezing.

Finally, the QFI of the squeezed vacuum is Fθ=8⟨n⟩(⟨n⟩+1)\mathcal{F}_{\theta}=8\langle n\rangle(\langle n\rangle+1). Along the path αtot=0\alpha_{\mathrm{tot}}=0 with increasing rtotr_{\mathrm{tot}}, s∗s_{*} approaches κ≃2e−2rtot\kappa\simeq2e^{-2r_{\mathrm{tot}}}, and Eq. (25) gives Nmax⁡≃e2rtot(ln⁡ε−1+2rtot)/2N_{\max}\simeq e^{2r_{\mathrm{tot}}}(\ln\varepsilon^{-1}+2r_{\mathrm{tot}})/2, up to corrections logarithmic in ln⁡ε−1+2rtot\ln\varepsilon^{-1}+2r_{\mathrm{tot}}. Hence ⟨n⟩≃Nmax⁡/[2(ln⁡ε−1+2rtot)]\langle n\rangle\simeq N_{\max}/[2(\ln\varepsilon^{-1}+2r_{\mathrm{tot}})], and

Fθ≃2NRMS2(ln⁡ε−1+2rtot)2.(88)\mathcal{F}_{\theta}\simeq\frac{2N_{\mathrm{RMS}}^{2}}{(\ln\varepsilon^{-1}+2r_{\mathrm{tot}})^{2}}. \tag*{(88)}

For ε=10−6\varepsilon=10^{-6} and rtot=2r_{\mathrm{tot}}=2, the exact minimization gives Fθ/NRMS2≈5.3×10−3\mathcal{F}_{\theta}/N_{\mathrm{RMS}}^{2}\approx5.3\times10^{-3}.

References

  1. [1]V. Giovannetti, S. Lloyd, and L. Maccone, Nat. Photonics 5, 222 (2011).
  2. [2]C. L. Degen, F. Reinhard, and P. Cappellaro, Rev. Mod. Phys. 89, 035002 (2017).
  3. [3]C. R. Rao, Bull. Calcutta Math. Soc. 37, 81 (1945).
  4. [4]H. Cramér, Mathematical Methods of Statistics (Princeton University Press, Princeton, NJ, 1946).
  5. [5]S. L. Braunstein and C. M. Caves, Phys. Rev. Lett. 72, 3439 (1994).
  6. [6]R. Demkowicz-Dobrzański, M. Jarzyna, and J. Kołodyński, in Progress in Optics, Vol. 60, edited by E. Wolf (Elsevier, Amsterdam, 2015) pp. 345–435.
  7. [7]M. Fadel, N. Roux, and M. Gessner, Rep. Prog. Phys. 88, 106001 (2025).
  8. [8]A. Manson, Phys. Rev. A 73, 033821 (2006).
  9. [9]J. Guo, S. Liu, B. Jing, Q. He, and M. Gessner, Metrological sensitivity beyond Gaussian limits with cubic phase states (2025), arXiv:2512.03769 [quant-ph].arxiv.org/abs/2512.03769
  10. [10]R. Gordillo-Hachuel, E. Torrentequi, C. de Dios, and R. Puebla, Quantum metrological advantage of high-order squeezed states (2026), arXiv:2604.09958 [quant-ph].
  11. [11]K. Park, T. Krisnanda, Y. Gao, and R. Filip, Quantum phase estimation beyond the Gaussian limit (2025), arXiv:2508.13046 [quant-ph].arxiv.org/abs/2508.13046
  12. [12]Á. Rivas and A. Luis, New J. Phys. 14, 093052 (2012).
  13. [13]A. Luis, Physical Review A 95, 10.1103/phys-reva.95.032113 (2017).
  14. [14]C. Lee, C. Oh, H. Jeong, C. Rockstuhl, and S.-Y. Lee, J. Phys. Commun. 3, 115008 (2019).
  15. [15]D. Branford and J. Rubio, New J. Phys. 23, 123041 (2021).
  16. [16]M. Jarzyna and R. Demkowicz-Dobrzański, Phys. Rev. A 85, 011801(R) (2012).
  17. [17]C. M. Caves, Phys. Rev. D 23, 1693 (1981).
  18. [18]C. Gerry and P. Knight, Introductory Quantum Optics (Cambridge University Press, Cambridge, 2004).
  19. [19]M. O. Scully and M. S. Zubairy, Quantum Optics (Cambridge University Press, Cambridge, 1997).
  20. [20]G. Harder, T. J. Bartley, A. E. Lita, S. W. Nam, T. Gerrits, and C. Silberhorn, Phys. Rev. Lett. 116, 143601 (2016).
  21. [21]K. Chihni and N. Quesada, Phys. Rev. A 110, 013712 (2024).
  22. [22]D. F. Walls and R. Barakat, Phys. Rev. A 1, 446 (1970).
  23. [23]H. Vahlbruch, M. Mehmet, K. Danzmann, and R. Schnabel, Phys. Rev. Lett. 117, 110801 (2016).
  24. [24]C. K. Law and J. H. Eberly, Phys. Rev. Lett. 76, 1055 (1996).
  25. [25]M. Hofheinz, H. Wang, M. Ansmann, R. C. Bialczak, E. Lucero, M. Neeley, A. D. O’Connell, D. Sank, J. Wenner, J. M. Martinis, and A. N. Cleland, Nature 459, 546 (2009).
  26. [26]T. Popoviciu, Mathematica (Cluj) 9, 129 (1935).
  27. [27]S. Lloyd and S. L. Braunstein, Phys. Rev. Lett. 82, 1784 (1999).
  28. [28]R.-W. Wu, T.-J. Tarn, and C.-W. Li, Phys. Rev. A 73, 012719 (2006).
  29. [29]O. Pinel, P. Jian, N. Treps, C. Fabre, and D. Braun, Phys. Rev. A 88, 040102(R) (2013).
  30. [30]P. Marian and T. A. Marian, Phys. Rev. A 47, 4474 (1993).
  31. [31]E. Fitzke, F. Niederschuh, and T. Walther, APL Photonics 8, 026106 (2023).

Paper details

vlnc:W_ufackf2d
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 8

downloads · 0

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

  • Physics and Astronomy

Claude, ChatGPT

View statement
Looking for related papers…
Contents