Quantum Speed Limit under Calibration Uncertainty
Salman Sajad Wani, Saif Al-Kuwari · 2026
Read Quantum Speed Limit under Calibration Uncertainty by Salman Sajad Wani, Saif Al-Kuwari (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-27423.txt.
Introduction
Salman Sajad Wani and Saif Al-Kuwari set speed limits for quantum sensors when calibration is uncertain. They project out nuisance parameters on a quotient manifold using quantum Fisher information, deriving bounds for general Markovian evolution. Applied to Jaynes-Cummings sensors, the framework yields detuning tolerances and speed limits from field-dependent Purcell loss.
Catalog record
| CREATOR | Salman Sajad Wani, Saif Al-Kuwari |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 79391 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.27423
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quantum Speed Limit under Calibration Uncertainty Salman Sajad Wani, Saif Al-Kuwari Abstract Standard quantum speed limits presuppose exactly known parameters, overestimating operational speed under calibration uncertainty. We introduce a projected speed limit based on the quantum Fisher information that profiles out these nuisance parameters on a quotient manifold. We derive constructive bounds for general Markovian evolution using sensitivity equations. Applying this to Jaynes–Cummings sensors, we obtain explicit detuning tolerances and quantify speed limits arising from field-dependent Purcell loss. This framework turns geometric bounds into concrete design rules for calibration and interrogation time. I Introduction Quantum speed limit (QSL) quantify the minimal time for a quantum state to evolve into a distinguishable state. The earliest bounds, due to Mandelstam–Tamm and Margolus–Levitin, relate the orthogonalization time to the energy variance and, respectively, to the mean energy above the ground state 28 ; 2 ; 29 . Geometric treatments recast these results in the Bures/Uhlmann geometry and use the symmetric logarithmic derivative (SLD) quantum Fisher information (QFI) metric to fix the line element 8 ; 45 ; 6 ; 34 . Subsequent work extended this framework to general open-system dynamics, yielding QSLs expressed directly in terms of the instantaneous statistical speed for completely positive trace-preserving evolutions 44 ; 14 ; 12 . Related developments also exploited the nonuniqueness of contractive distinguishability metrics to construct generalized geometric QSL families for unitary and nonunitary dynamics, including bounds that can be tighter than the conventional QFI/Bures choice 35 . For open dynamics, studies have emphasized complementary desiderata such as tightness, robustness under composition and mixing, and computational or experimental feasibility 9 . Related work also sharpened the operational interpretation of quantum speed limits, including in open-system settings 30 ; 40 . More recent Liouville-space formulations have derived exact and inexact QSL families for completely positive trace-preserving dynamics and clarified attainability for time-optimal CPTP evolutions 42 . These developments have deepened the link between QSLs, quantum estimation theory, and monotone information metrics 22 ; 34 ; 33 . They also underpin many analyses of quantum metrology, control, and dissipation 19 ; 27 ; 20 . A common feature across the standard Bures/QFI setting, generalized geometric families, and more recent open-system and CPTP formulations is the assumption of a fixed, known generator 44 ; 14 ; 12 ; 35 ; 9 ; 42 . Operationally, distinguishability along laboratory time is therefore confounded by nuisance directions in parameter space: an experimenter can partially compensate for changes in time by co-varying these parameters within their estimation uncertainty. Consequently, the physical Bures speed computed at a fixed generator generally overestimates the operational speed relevant for distinguishing states modulo nuisance reparametrizations. In this paper we formulate a nuisance-aware projected quantum speed limit. We treat time and nuisance parameters jointly within the multiparameter SLD-QFI and project the Bures metric onto the time direction via a Schur complement 11 ; 26 ; 46 . This construction yields an effective information F_{\mathrm{eff}} and an associated projected speed v_{\mathrm{quo}}=\tfrac{1}{2}\sqrt{F_{\mathrm{eff}}} . This speed never exceeds the standard Bures speed and remains invariant under reparametrizations of the nuisance manifold. Crucially, F_{\mathrm{eff}} coincides with the partial SLD Fisher information, which sets the ultimate precision limit when nuisance parameters are profiled out 43 ; 41 . In contrast to standard formulations 44 ; 14 ; 12 , which bound evolution for a frozen generator, our framework quantifies operational distinguishability modulo nuisances. We implement this for general Markovian dynamics using GKSL sensitivity equations 27 ; 20 ; 32 ; 16 and analyze two paradigmatic sensors: a unitary Jaynes-Cummings (JC) model and a dispersive JC cavity with loss. For the examples studied here, the nuisance penalty is more pronounced in the dispersive open JC sensor, since the field controls both the coherent cavity-frequency shift and the effective decay rate. In both regimes, the projected QSL converts abstract geometric limits into concrete design rules. One example is the tolerance |\Delta|t\lesssim 0.3 for retaining 99\% of the physical speed. II Formalism We work in the Bures–SLD geometry of quantum states. The Bures metric, induced by the Uhlmann fidelity, provides the local statistical distance on the space of density operators 8 ; 45 ; 6 ; 34 . The corresponding symmetric logarithmic derivative (SLD) operators L_{\mu} are defined by \partial_{\mu}\rho=\tfrac{1}{2}(\rho L_{\mu}+L_{\mu}\rho), (1) and the associated SLD quantum Fisher information matrix (QFIM) F_{\mu\nu} determines the local line element dL^{2}=\tfrac{1}{4}F_{\mu\nu}\,dx^{\mu}dx^{\nu}. (2) For any parameter direction, this quadratic form gives the local sensitivity of the state to infinitesimal displacements; for time evolution, its square root is the instantaneous statistical speed that underlies geometric quantum speed limits 6 ; 44 ; 13 . We take the coordinate vector x^{\mu}=(t,\bm{\lambda}) to include both laboratory time and the nuisance parameters \bm{\lambda} (e.g., detunings, loss rates). In our setting, \bm{\lambda} are not targets of joint estimation but sources of uncertainty. Throughout this work, the nuisance parameters \bm{\lambda} are treated as run-wise fixed but imperfectly known calibration parameters that label the generator of the dynamics. They are therefore distinct from explicitly time-dependent stochastic fluctuations, which belong to a different modeling layer and are not profiled by the present Schur-complement construction. In this multiparameter setting, the QFIM takes the block form F=\begin{pmatrix}F_{tt}&\bm{f}^{\top}\\ \bm{f}&F_{\lambda\lambda}\end{pmatrix},\qquad\bm{f}:=(F_{t\alpha})_{\alpha},\;F_{\lambda\lambda}:=(F_{\alpha\beta})_{\alpha,\beta}. (3) Standard “physical” QSLs are obtained by fixing \bm{\lambda} and using F_{tt} alone. Operationally, however, an experimenter can re-tune \bm{\lambda} to reduce distinguishability along nuisance directions. Thus the relevant speed is set not by F_{tt} , but by the information component that is orthogonal to the nuisance manifold. To isolate this component, we minimize dL^{2} over d\bm{\lambda} at fixed dt , profiling the Bures quadratic form in analogy with profile likelihood 11 ; 26 ; 1 . Solving the normal equations yields the Schur complement F_{\mathrm{eff}}:=F_{tt}-\bm{f}^{\top}F_{\lambda\lambda}^{-1}\bm{f}, (4) where F_{\lambda\lambda}^{-1} denotes the Moore–Penrose pseudoinverse on the estimable nuisance subspace 3 ; 46 . Geometrically, this Schur complement is the squared norm of the time tangent after orthogonal projection away from the nuisance directions in the Bures–SLD geometry. It therefore removes the distinguishability attributable to nuisance reparametrizations. Equivalently, the profiled quadratic form induces the local quotient metric obtained by identifying tangent variations that differ only by nuisance components: dL_{\mathrm{quo}}^{2}=1/4F_{\mathrm{eff}}\,dt^{2}. F_{\mathrm{eff}} satisfies 0\leq F_{\mathrm{eff}}\leq F_{tt} and is invariant under smooth reparametrizations of \bm{\lambda} . When F is invertible, F_{\mathrm{eff}}=((F^{-1})_{tt})^{-1} , recovering the standard profiled information. Rank-deficient states (including pure states) are handled by restricting all constructions to \operatorname{supp}(\rho) 22 ; 37 . In the SLD formalism, F_{\mathrm{eff}} coincides with the partial quantum Fisher information for time. Projecting L_{t} orthogonally to the span of nuisance scores \{L_{\alpha}\} via the inner product \langle\!\langle A,B\rangle\!\rangle_{\rho}:=\tfrac{1}{2}\Tr[\rho\{A,B\}] yields the efficient score \widetilde{L}_{t} , whose norm is \langle\!\langle\widetilde{L}_{t},\widetilde{L}_{t}\rangle\!\rangle_{\rho}=F_{tt}-\bm{f}^{\top}F_{\lambda\lambda}^{-1}\bm{f}=F_{\mathrm{eff}}. (5) Thus, the projected metric coefficient sets the ultimate precision limit \mathrm{Var}(\hat{t})\geq 1/(NF_{\mathrm{eff}}) attainable when nuisance parameters are profiled out 43 ; 41 . To embed this geometry into Markovian dynamics, we consider trajectories generated by the GKSL master equation \partial_{t}\rho(t,\bm{\lambda})=\mathcal{L}_{\bm{\lambda}}[\rho]:=-\tfrac{i}{\hbar}[H(\bm{\lambda}),\rho]+\sum_{k}\mathcal{D}[F_{k}(\bm{\lambda})]\rho. (6) The time derivative entering the QFIM is \partial_{t}\rho=\mathcal{L}_{\bm{\lambda}}[\rho] , while the nuisance derivatives \rho^{\prime}_{\alpha}:=\partial_{\lambda^{\alpha}}\rho satisfy the linear sensitivity equation \partial_{t}\rho^{\prime}_{\alpha}(t)=(\partial_{\lambda^{\alpha}}\mathcal{L}_{\bm{\lambda}})[\rho(t)]+\mathcal{L}_{\bm{\lambda}}[\rho^{\prime}_{\alpha}(t)]. (7) This equation allows F_{\mathrm{eff}}(t) to be evaluated efficiently along any trajectory using the Duhamel formula 32 ; 16 . At each fixed time, the profiling step remains explicit: one solves the normal equations F_{\lambda\lambda}\,d\bm{\lambda}_{*}=-\,\bm{f}\,dt and substitutes the result into