Quantum Chain Rule in Quantum Information

• Quantum chain rule is a framework of inequalities and equalities that generalizes classical chain rules to describe multipartite entropy in quantum systems.
• It connects key quantities like sandwiched Rényi entropies, quantum relative entropy, and smooth min-entropy, underpinning operational tasks such as channel discrimination and privacy amplification.
• The framework is pivotal for quantifying security in quantum protocols and ensuring fundamental limits by excluding non-physical super-quantum correlations.

The quantum chain rule refers to a collection of inequalities and equalities that generalize the classical information-theoretic chain rule to the quantum setting. These chain rules connect multipartite entropic or divergence quantities—such as (smooth) min-entropy, Rényi entropies, mutual information, and quantum relative entropy—under various operational contexts and models for quantum systems, channels, and protocols.

1. Chain Rule for Quantum Rényi Conditional Entropy

For quantum systems $A, B, C$ and a quantum state $\rho_{ABC}$, the classical chain rule for the (Shannon or von Neumann) conditional entropy reads $H(AB|C) = H(A|BC) + H(B|C)$. For the quantum "sandwiched" Rényi conditional entropy, this identity is replaced by a chain rule inequality parameterized by Rényi indices:

For $\alpha, \beta, \gamma \in (1, \infty)$, subject to the conjugacy constraint

$$\frac{\alpha}{\alpha-1} = \frac{\beta}{\beta-1} + \frac{\gamma}{\gamma-1}\,,$$

and

$(\alpha-1)(\beta-1)(\gamma-1) > 0\,,$

the sandwiched Rényi chain rule states (Dupuis, 2014): $$H_\alpha(AB|C)_\rho \ge H_\beta(A|BC)_\rho + H_\gamma(B|C)_\rho.$$ If $(\alpha-1)(\beta-1)(\gamma-1) < 0$, the direction of the inequality is reversed.

The sandwiched Rényi conditional entropy is given by

$$H_\alpha(A|B)_\rho := -\inf_{\sigma_B} D_\alpha(\rho_{AB} \| I_A \otimes \sigma_B),$$

where the sandwiched Rényi divergence is

$$D_\alpha(\rho \| \sigma) := \frac{1}{\alpha-1} \log \| \sigma^{\frac{1-\alpha}{2\alpha}} \rho \sigma^{\frac{1-\alpha}{2\alpha}} \|_\alpha^\alpha,$$

with $\|X\|_\alpha = (\Tr |X|^\alpha)^{1/\alpha}$.

Operationally, these conditional Rényi entropies interpolate between the von Neumann entropy $(\alpha \to 1)$, min-entropy $(\alpha \to \infty)$, and govern one-shot settings such as hypothesis testing and privacy amplification (Dupuis, 2014).

2. Quantum Relative Entropy and Chain Rules for Channels

For quantum relative entropy (Umegaki entropy), the classical chain rule

$$D(P_{XY} \| Q_{XY}) = D(P_X \| Q_X) + \mathbb{E}_{x\sim P_X} D(P_{Y|X=x}\| Q_{Y|X=x})$$

is not available as an equality in the quantum case due to noncommutativity. Gasbarri and Hoogsteder-Riera establish that single-letter chain rule inequalities for quantum relative entropy can be formulated in terms of decompositions by POVMs or projectors (Gasbarri et al., 19 Oct 2025). For a CPTP map $\mathcal{M}$ and $\mathcal{N}$, and a POVM $G$: $$D(\rho\Vert\sigma)\;-\;D(\mathcal M(\rho)\Vert\mathcal N(\sigma)) \;\ge\;-\,\mathbb E_{j\sim P^G_\rho}\;D(\mathcal M(\rho_j)\Vert\mathcal N(\sigma_j)).$$ Further, in the channel context, the chain rule for the quantum relative entropy establishes that for channels $\mathcal{E}$ (TPCP) and $\mathcal{F}$ (CP): $$D(\mathcal{E}(\rho)\| \mathcal{F}(\sigma)) \le D(\rho\|\sigma) + D^{\mathrm{reg}}(\mathcal{E}\|\mathcal{F}),$$ with $$D^{\mathrm{reg}}(\mathcal{E}\|\mathcal{F}) = \lim_{n\to\infty} \frac1n D(\mathcal{E}^{\otimes n}\| \mathcal{F}^{\otimes n})$$ (Fang et al., 2019).

3. Chain Rules for Quantum Rényi Divergences and Quantum Channels

For quantum channels (CP or CPTP maps), divergence chain rules relate the divergence of channel outputs to that of inputs plus the divergence between the channels themselves. For the sandwiched Rényi divergence and tensor-stable maps $E,F$: $$D_\alpha(E(\rho) \| F(\sigma)) \le D_\alpha(\rho\|\sigma) + D_\alpha^\infty(E\|F),$$ with $$D_\alpha^\infty(E\|F) = \lim_{n\to\infty} \frac1n D_\alpha(E^{\otimes n} \| F^{\otimes n})$$ (Berta et al., 2022). This structure underlies the impossibility of adaptive quantum channel discrimination outperforming non-adaptive strategies in the asymptotic regime (Fang et al., 2019).

4. Smooth Min- and Max-Entropy Chain Rules

For the smooth min-entropy $H_{\min}^\epsilon(A|B)$ of $A$ given $B$, classical chain rules fail due to smoothness and worst-case conditioning. However, Marwah & Dupuis derive a universal chain rule for a variant $H_{\min}^{\downarrow,\epsilon}$: $$H_{\min}^{\downarrow, g_1(\epsilon)}(A_1^n | B) \geq \sum_{k=1}^n H_{\min}^{\downarrow,\epsilon}(A_k|A_1^{k-1}B) - n g_2(\epsilon) - k(\epsilon),$$ with all corrections independent of $n$ and $g_1,g_2(\epsilon)\to0$ as $\epsilon\to0$ (Marwah et al., 2024). The dual relation holds for the smooth max-entropy.

In operational settings involving interactive leakage, the change of min-entropy under a quantum protocol is quantified as

$$H_{\min}(A_0|B_r) \geq H_{\min}(A_0|B_0) - \min\{ m_A + m_B, 2 m_A \},$$

where $m_A, m_B$ are the total communication from Alice to Bob and vice versa (Lai et al., 2018).

5. Chain Rules for Quantum Mutual Information and Security Applications

Chain rules also govern how the addition or removal of leakage or side registers affects security in quantum cryptography. For a quantum state $\rho_{SEL}$, the smooth min-entropy with an additional register $L$ satisfies

$$H_{\min}^{\epsilon+\delta}(S|LE) \geq H_{\min}^\epsilon(S|E) - I_{\max}^\delta(SE;L) - \log \frac{4}{\delta^2},$$

where $I_{\max}^\delta$ is the smooth max-information (Arqand et al., 2024).

Further, in protocols producing output registers via a sequence of channels each emitting leakage, the accumulated smooth max-information between output and leakage is bounded by a sum of single-round Rényi mutual informations. These results enable security analyses for both device-dependent and device-independent quantum key distribution under imperfect or leaky devices (Arqand et al., 2024), incorporating such chain rules into generalized entropy accumulation theorems.

6. Quantum Chain Rule and Physical Principles

The chain rule is not merely technical; it is fundamental to the structure of quantum (or classical) information theory. For generalized mutual information (GMI) with an operational (channel-coding) definition, imposing the chain rule excludes super-quantum (beyond Tsirelson’s bound) correlations. Wakakuwa and Murao show that the chain rule for GMI, together with data-processing and no-signalling conditions, is logically equivalent to bounding nonlocal correlations below the quantum Tsirelson limit (Wakakuwa et al., 2012).

7. Proof Methods and Technical Ingredients

Proofs of quantum chain rules utilize operator norm interpolations (generalized Riesz–Thorin theorems), variational representations of Schatten $p$-norms, spectral pinching, matrix analysis, and convex optimization. For sandwiched Rényi entropies, one constructs holomorphic families of operators whose norms are related by conjugacy of Rényi parameters; for relative entropy chain rules, regularization over tensor powers and asymptotic equipartition appear crucial (Dupuis, 2014, Berta et al., 2022, Gasbarri et al., 19 Oct 2025).

Summary Table: Key Quantum Chain Rule Results

Setting / Quantity	Chain Rule Statement	Reference
Sandwiched Rényi entropy	$$H_\alpha(AB|C)_\rho \geq H_\beta(A|BC)_\rho + H_\gamma(B|C)_\rho$$ (parameters)	(Dupuis, 2014)
Quantum relative entropy, channel	$$D(\mathcal{E}(\rho)\| \mathcal{F}(\sigma)) \le D(\rho\|\sigma) + D^{\text{reg}}(\mathcal{E}\|\mathcal{F})$$	(Fang et al., 2019)
Quantum mutual info (GMI)	$I_G(A:BC) = I_G(A:C) + I_G(A:B|C)$	(Wakakuwa et al., 2012)
Universal chain rule (smooth min)	$$H_{\min}^{\downarrow,g_1}(A_1^n|B) \geq \sum_k H_{\min}^{\downarrow,\epsilon}(A_k|A_1^{k-1}B) - \delta(n)$$	(Marwah et al., 2024)
Interactive min-entropy leakage	$$H_{\min}(A_0|B_r) \ge H_{\min}(A_0|B_0) - \min\{m_A+m_B,2m_A\}$$	(Lai et al., 2018)
Chain rules for channels (Rényi)	$$D_\alpha(E(\rho)\|F(\sigma)) \le D_\alpha(\rho\|\sigma)+D_\alpha^\infty(E\|F)$$	(Berta et al., 2022)
Mutual info under leakage	$$H_{\min}^{\epsilon+\delta}(S|LE) \ge H_{\min}^\epsilon(S|E) - I_{\max}^\delta(SE;L) - O(\log 1/\delta)$$	(Arqand et al., 2024)

Significance and Applications

Quantum chain rules are foundational in quantum information theory. They govern the structure of multipartite entropy, mutual information, and distinguishability measures—underpinning the analyses of quantum communication, channel discrimination, privacy amplification, cryptographic key rates, and the physical limitations of quantum correlations. The current body of single-shot and regularized quantum chain rules extends their applicability to nonasymptotic scenarios, device-imperfect protocols, and physical principle derivations such as the exclusion of super-Tsirelson correlations.