Conditional Quantum Fisher Information

• Conditional Quantum Fisher Information (CQFI) is a framework that quantifies the rate of information acquisition per measurement in quantum systems with sequential, temporally correlated dynamics.
• It extends standard ensemble-averaged QFI by incorporating measurement back-action, coherence effects, and trajectory-resolved statistical fluctuations.
• CQFI has practical applications in quantum thermometry and Rabi frequency estimation, where optimizing measurement intervals and feedback can enhance sensitivity.

The Conditional Quantum Fisher Information (CQFI) is a refinement of quantum parameter estimation theory, quantifying the rate of information acquisition about a parameter $\theta$ in metrological protocols where measurements are temporally correlated due to sequential dynamics and measurement back-action. Unlike the standard, ensemble-averaged Quantum Fisher Information (QFI), which mimics an i.i.d. scenario, the CQFI framework encompasses both the statistical and trajectory-resolved aspects of parameter sensitivity in non-i.i.d. quantum processes, and can be formulated both in terms of sequential probability distributions and trajectory-conditioned quantum states (O'Connor et al., 2024, Melo et al., 18 Jan 2026). It provides a rigorous approach to quantifying the informational content and metrological limits of quantum systems under continuous monitoring or stroboscopic probing, explicitly accounting for quantum coherence, measurement history, and stochastic realization-dependent effects.

1. Formal Definition and Mathematical Structure

The CQFI admits two complementary definitions:

A. Sequential Measurement Setting.—

For a quantum probe subjected to $N$ sequential, generally non-commuting, $\theta$-dependent quantum channels $\mathcal{E}_\theta$ interleaved with generalized measurements $\{L_m\}$, the measurement record $\mathbf{m}_{1:N} = (m_1, \ldots, m_N)$ is distributed as: $$P(\mathbf{m}_{1:N};\theta) = \prod_{k=1}^N P(m_k | \mathbf{m}_{1:k-1};\theta)$$ with conditional probabilities defined by the cascaded application of channels and measurements: $$P(m_{k+1} | \mathbf{m}_{1:k};\theta) = \text{Tr}\left[ \rho_S(\mathbf{m}_{1:k}) M_{m_{k+1}} \right],$$ where $\rho_S(\mathbf{m}_{1:k})$ is the probe state after $k$ outcomes, $M_m = L_m^\dagger L_m$.

The total Fisher information for the entire record is: $$F_\theta(\mathbf{m}_{1:N}) = \sum_{\mathbf{m}_{1:N}} P(\mathbf{m}_{1:N};\theta) \left[ \partial_\theta \ln P(\mathbf{m}_{1:N};\theta) \right]^2.$$ If the process possesses finite Markov order $\mathcal{M}$, in the limit $N \to \infty$: $$I_\text{cond}(\theta) := \lim_{N\to\infty} \frac{1}{N} F_\theta(\mathbf{m}_{1:N}) = F_\theta(m_{\mathcal{M}+1}|\mathbf{m}_{1:\mathcal{M}})$$ where the one-step conditional Fisher information is

$$F_\theta(m_{\mathcal{M}+1}|\mathbf{m}_{1:\mathcal{M}}) = \sum_{\mathbf{m}_{1:\mathcal{M}}} P(\mathbf{m}_{1:\mathcal{M}};\theta) \sum_{m_{\mathcal{M}+1}} P(m_{\mathcal{M}+1}|\mathbf{m}_{1:\mathcal{M}};\theta) \left[\partial_\theta \ln P(m_{\mathcal{M}+1}|\mathbf{m}_{1:\mathcal{M}};\theta)\right]^2.$$

This quantity is the Fisher information rate per measurement, i.e., the CQFI (O'Connor et al., 2024).

B. Trajectory-Conditioned Quantum Perspective.—

Given a quantum system following a pure or mixed conditional state $\rho_t^{(k)}(\theta)$ along trajectory $k$, the trajectory-conditioned Symmetric Logarithmic Derivative (SLD) $L_\theta^{(k)}$ satisfies: $$\frac{\partial \rho_t^{(k)}}{\partial \theta} = \frac{1}{2} \left\{ L_\theta^{(k)}, \rho_t^{(k)} \right\}$$ The trajectory CQFI is then: $$I_Q^{(k)}(\theta) = \mathrm{Tr}\left[ \rho_t^{(k)} (L_\theta^{(k)})^2 \right]$$ This definition naturally generalizes the classical stochastic Fisher information to single quantum trajectories (Melo et al., 18 Jan 2026).

2. Decomposition and Physical Content

The spectrum-based decomposition of the CQFI elucidates three physically distinct contributions: $$I_Q^{(k)} = f^{IC}_{Q,(k)} + f^{C}_{Q,(k)} + f^{X}_{Q,(k)}$$

• Incoherent (Population) Term $f^{IC}_{Q,(k)}$:

Originates from the variation of eigenvalues (populations) $p_n(\theta)$; directly analogous to classical Fisher information.

• Coherent (Basis-Rotation) Term $f^{C}_{Q,(k)}$:

Accounts for the parameter dependence of the eigenbasis $|n(\theta)\rangle$; encodes additional information from quantum coherence (“geometry”).

• Interference (Cross) Term $f^{X}_{Q,(k)}$:

Reflects transient destructive or constructive interference between incoherent and coherent channels; this term can be negative for individual trajectories, but averages to zero over the ensemble (Melo et al., 18 Jan 2026).

Term	Mathematical Structure	Physical Origin
$f^{IC}_{Q,(k)}$	$\sum_n |c_n^{(k)}|^2 (\partial_\theta p_n/p_n)^2$	Population change (classical-like)
$f^{C}_{Q,(k)}$	$\sum_{n} |\sum_{m\neq n} \cdots|^2$	Coherent (basis-rotation, geometry)
$f^{X}_{Q,(k)}$	$\sum_{n\neq m} \mathrm{Re}\left\{\cdots\right\}$	Interference between population and coherence channels

This decomposition highlights that, in contrast to ensemble QFI, the CQFI on each trajectory resolves nontrivial fluctuations—including negative interference contributions—that are inaccessible to conventional, ensemble-averaged analysis.

3. Physical Interpretation and Implications

The CQFI $I_\text{cond}(\theta)$, as the Fisher information rate, quantifies the average information gain per measurement about parameter $\theta$ in a sequential protocol. Sequential measurements induce temporal correlations and back-action that alter the informativeness of outcomes compared to i.i.d. (reset-based) strategies. The trajectory-level CQFI, in turn, provides a granular resolution of informational fluctuations, revealing unique features such as destructive interference (negative cross-terms) in single-shot quantum trajectories (O'Connor et al., 2024, Melo et al., 18 Jan 2026).

Pragmatically, the CQFI is tightly tied to the sensitivity of measurement transition probabilities to $\theta$. For Markovian dynamics induced by projective measurement, the CQFI rate is fully determined by the steady-state distribution and parameter-sensitivity of the transition matrix. This sets the metric for how “fast” information about $\theta$ is acquired as measurements proceed.

4. Applications

CQFI has been analytically evaluated in several paradigmatic quantum metrology settings:

A. Quantum Thermometry:

• Probe: $D$-level system coupled to a bosonic bath.
• Probe dynamics: Davies map with temperature-sensitive transitions, monitored by stroboscopic projective measurements.
• The outcome sequence forms a Markov chain with analytically tractable transition probabilities.
• The CQFI rate (per measurement) is enhanced over i.i.d. protocols if the measurement interval $\tau$ is optimized and feedback is employed; partial thermalization yields increased sensitivity over infinite-$\tau$ (fully reset) cases.

B. Rabi Frequency Estimation:

• Probe: Qubit with coherent drive and spontaneous emission.
• Sequential projective measurements in the $\sigma_z$ basis induce a two-state Markov process with conditional transition probabilities dependent on $\Omega$ and emission rate.
• The CQFI rate is nontrivially dependent on measurement gap $\tau$; optimizing $\tau$ and measurement basis further enhances metrological performance (O'Connor et al., 2024).

5. Optimization Strategies

Several protocol parameters can be tuned to maximize the CQFI rate:

• Inter-measurement Interval ($\tau$): Maximizing $F_{2|1}$ by optimizing $\tau$ between measurements can significantly enhance information gain.
• Feedback Control: Adjusting $\tau$ adaptively based on previous outcomes further boosts the CQFI rate.
• Measurement Basis Choice: For Hamiltonian parameter estimation, dynamically adjusting the measurement basis enhances transition sensitivity and thus the attainable CQFI.
• Real-time Monitoring of CQFI Decomposition: In the stochastic trajectory framework, monitoring $f^{IC}_{Q,(k)}$, $f^{C}_{Q,(k)}$, and $f^{X}_{Q,(k)}$ in real-time enables adaptive measurement strategies responsive to instantaneous trajectory features (Melo et al., 18 Jan 2026).

6. Comparison with Ensemble-Averaged QFI

In i.i.d. protocols, the Fisher information accumulates linearly: $F^{\mathrm{iid}} \times N$ for $N$ repetitions, setting the standard Cramér-Rao scaling. CQFI exhibits fundamentally different behavior:

• Temporal Correlations: CQFI naturally incorporates non-i.i.d. dynamics, including back-action and memory effects, leading to the possibility of both informational enhancement and suppression relative to $F^{\mathrm{iid}}$.
• Constraints: CQFI is restricted by the structure of (possibly realistic) sequential local measurements, rather than hypothetical global measurements.
• Trajectory-resolved Information: The decomposition at the stochastic trajectory level resolves information fluctuations masked in $F^{\mathrm{QFI}}$; crucially, the cross-term $f^X$ cancels only in ensemble averages (Melo et al., 18 Jan 2026).
• Geometric and Dynamical Implications: The CQFI metric defines a stochastic information geometry, with trajectory-dependent thermodynamic length and action that yield trajectory-level quantum speed limits, sometimes tighter or looser than ensemble bounds.

7. Stochastic Quantum Information Geometry and Quantum Speed Limits

Each trajectory $k$ in a monitored quantum system gives rise to a random information metric: $ds_k^2 = \frac{1}{4} I_Q^{(k)}(t)\, dt^2$ This framework introduces trajectory-level thermodynamic length

$$\ell_k(\tau) = \int_0^\tau \frac{1}{2} \sqrt{I_Q^{(k)}(t)}\, dt$$

and action

$$j_k(\tau) = \frac{\tau}{4} \int_0^\tau I_Q^{(k)}(t)\, dt$$

with the Cauchy–Schwarz bound $j_k(\tau) \geq \ell_k^2(\tau)$.

Associated quantum speed limits (QSLs) at trajectory level constrain observable dynamics: $$|\dot o_k(t)| \leq \Delta_{\rho^{(k)}_t} O\, \sqrt{I_Q^{(k)}(t)}$$ Integration yields geometric speed limits linked to thermodynamic length.

This stochastic geometry, and the associated QSLs and CQFI decomposition, provide a refined lens for analyzing quantum-dynamical estimation limits beyond the ensemble-mean paradigm, relevant for real-time metrological optimization and for the study of rare, information-rich quantum trajectories (Melo et al., 18 Jan 2026).