Tight Entropic Uncertainty Relations

• Tight Entropic Uncertainty Relations are precise lower bounds on the entropy of measurement outcome distributions in quantum systems, ensuring operational relevance in cryptography, steering, and entanglement detection.
• They extend classical bounds like the Maassen–Uffink inequality by incorporating multiple entropy measures (e.g., Rényi, Tsallis, min-entropy) and addressing both state-independent and quantum-memory-assisted scenarios.
• Recent advances use geometric optimization and majorization methods to derive bounds that are either analytically tight or numerically optimal, thereby improving security proofs and coherence quantification.

Tight entropic uncertainty relations are lower bounds on entropic measures of measurement outcome distributions that are attained by physical states, coincide with the exact minimum uncertainty, or converge to that minimum in a controlled limit. In their standard finite-dimensional form they bound expressions such as $H(A)+H(B)$, where $H$ is the Shannon entropy of the Born probabilities of two incompatible observables $A$ and $B$; more general versions involve min-entropy, smooth min-entropy, Rényi and Tsallis entropies, POVMs, continuous variables, and quantum side information. Across these settings, the central problem is not only to obtain nonzero lower bounds, but to characterize when such bounds are genuinely tight and therefore operationally meaningful for cryptography, steering, entanglement detection, and coherence theory [(Riccardi et al., 3 May 2026), 0612014, (Yang et al., 31 Jan 2026)].

1. Foundational formulation and the meaning of tightness

For two orthonormal measurement bases $A=\{|a_i\rangle\}$ and $B=\{|b_j\rangle\}$ in a $d$-dimensional Hilbert space, the Maassen–Uffink relation states

$$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$

In the mutually unbiased case, $c=1/\sqrt d$, so the lower bound becomes $\log_2 d$. This formulation already exhibits two features that make entropic uncertainty relations structurally different from variance-based ones: they depend only on the Born statistics of the outcomes and the lower bound is typically state independent (Ding et al., 2019, Riccardi et al., 3 May 2026).

Within the literature, “tight” is used in several related senses. A bound can be tight because equality is attained by explicit states; because it exactly characterizes the minimum of the uncertainty sum for a given measurement family; because it is numerically optimal up to preassigned precision; or because a one-parameter family converges to the exact minimum in a limit. These distinctions are not merely terminological. They separate analytically solvable settings, such as certain stabilizer or qubit families, from general finite-dimensional measurement scenarios where tightness is currently obtained algorithmically rather than in closed form (Niekamp et al., 2011, Abbott et al., 2015, Yang et al., 31 Jan 2026).

The classical two-observable Maassen–Uffink inequality is not uniformly tight for arbitrary observable pairs, but it is tight for important structured families. Niekamp et al. showed that it is tight for the measurement in any pair of stabilizer bases. Their proof uses the fact that the relevant overlaps satisfy $H$0, which implies equality in the Maassen–Uffink relation on basis states of one of the measurements (Niekamp et al., 2011). By contrast, for more than two observables, generic extensions obtained by combining pairwise bounds are often not tight; this theme recurs throughout the modern theory (2207.13469).

2. Exact state-independent results in finite dimensions

A large part of the theory concerns families of observables for which the exact state-independent Shannon lower bound is known. For qubits, Abbott, Alzieu, Hall, and Branciard derived tight state-independent uncertainty relations that completely characterize the obtainable uncertainty values for arbitrary pairs and triples of $H$1-valued observables. Their method first characterizes the allowed region of expectation values as an ellipse or ellipsoid in Bloch space and then converts the result into an entropic relation. For any two qubit observables $H$2 and $H$3, one obtains

$H$4

which is fully tight and is strictly stronger than the usual Maassen–Uffink bound wherever $H$5 (Abbott et al., 2015).

For specific low-dimensional families, exact constants are also known. Riccardi et al. derived tight Shannon-entropic bounds for spin observables $H$6 and for observables with mutually unbiased eigenbases in dimensions $H$7. They also identified the states saturating the inequalities, including null-projection states for integer-spin cases and explicit superposition families for MUB settings (Riccardi et al., 2017). Complementary qubit and qutrit MUB families furnish further exact multi-observable examples (2207.13469).

Setting	Tight lower bound	Equality statement
Two stabilizer bases on $H$8 qubits	$H$9	Equality is achieved for any basis state of $A$0 or $A$1 (Niekamp et al., 2011)
Three Pauli observables on a qubit	$A$2	Equality iff the state is an eigenstate of $A$3, $A$4, or $A$5 (Abbott et al., 2015, 2207.13469)
Spin observables $A$6 in $A$7	$A$8, $A$9, $B$0	Saturation by null-projection or maximal-$B$1 states, depending on $B$2 (Riccardi et al., 2017)
Complete MUB sets in $B$3	$B$4, $B$5, $B$6	Saturation by explicit one-parameter or superposition families (Riccardi et al., 2017)

A recent development replaces the Maassen–Uffink dependence on the single maximal overlap $B$7 by the full $B$8 operator norm of the overlap matrix $B$9. For $A=\{|a_i\rangle\}$0,

$A=\{|a_i\rangle\}$1

This bound is state independent, is strictly sharper than the Maassen–Uffink bound for every $A=\{|a_i\rangle\}$2, and becomes asymptotically tight for all observable pairs as $A=\{|a_i\rangle\}$3: $A=\{|a_i\rangle\}$4 The same norm-based construction also extends to Rényi entropies (Riccardi et al., 3 May 2026).

3. Multiple measurements and optimization beyond pairwise bounds

For multiple measurements, exact entropic tradeoffs are substantially harder. One important exact result concerns prime-power dimensions. Funder derived a tight lower bound on the sum of Shannon entropies for $A=\{|a_i\rangle\}$5 mutually unbiased measurements by first solving a purely classical optimization problem: minimizing $A=\{|a_i\rangle\}$6 under a fixed total collision-probability constraint. Combined with Larsen’s lemma,

$A=\{|a_i\rangle\}$7

this yields a tight MUB entropic uncertainty relation in prime-power dimension $A=\{|a_i\rangle\}$8, reducing to Maassen–Uffink when $A=\{|a_i\rangle\}$9 (Funder, 2011).

More generally, the multiple-measurement quantum-memory-assisted bound of Liu et al. is not always tight, particularly for MUB collections. Xie et al. introduced the simply constructed bound

$B=\{|b_j\rangle\}$0

and its optimized version

$B=\{|b_j\rangle\}$1

with

$B=\{|b_j\rangle\}$2

They proved that $B=\{|b_j\rangle\}$3 for arbitrary MUB measurements, and numerically found $B=\{|b_j\rangle\}$4 for Werner and Bell-diagonal families, while for large random samples $B=\{|b_j\rangle\}$5 touches the true uncertainty for a large fraction of states (Xie et al., 2021).

A different analytic strategy uses majorization and symmetric-group averaging. Xiao et al. obtained an admixture bound for multi-measurement entropic uncertainty relations by combining a universal majorization vector $B=\{|b_j\rangle\}$6 with permutation symmetrization of order-dependent overlap expressions. In the numerical scenarios reported in that work, the admixture bound strictly dominates previous multi-measurement lower bounds, and the same machinery extends to Rényi and Tsallis entropies because those entropies remain Schur-concave (Xiao et al., 2016).

At the opposite end of the spectrum, recent work formulates the exact finite-dimensional problem as a geometric optimization in quantum probability space. For an effective POVM $B=\{|b_j\rangle\}$7, the minimal entropy

$B=\{|b_j\rangle\}$8

is cast as a global concave minimization over a compact convex body. An outer-approximation algorithm then generates lower and upper bounds that converge to the true optimum with preassigned numerical precision. This approach yields tight entropic uncertainty bounds for general measurements in finite-dimensional systems, benchmarks favorably against analytical and majorization-based bounds, and strengthens steering tests (Yang et al., 31 Jan 2026).

The additivity structure changes sharply when more than two observables are included. According to tight joint-entropy inequalities for local measurements, additivity holds only for EURs that involve two observables. For $B=\{|b_j\rangle\}$9 or when a subsystem von Neumann entropy is added, the corresponding inequalities are not additive, and this non-additivity is directly connected with the presence of quantum correlations (2207.13469).

4. High-order, smooth, and memory-assisted entropies

A major branch of the subject replaces Shannon entropy by min-entropy or smooth min-entropy, largely because cryptographic security parameters are naturally one-shot. Damgård, Fehr, Renner, Salvail, and Schaffner derived a new entropic quantum uncertainty relation involving min-entropy and described it as tight. They used it to prove the security of quantum $d$0-out-of-$d$1 Oblivious Transfer and quantum Bit Commitment in the bounded quantum-storage model according to new strong security definitions. The same relation also yields security of QKD against quantum-memory-bounded eavesdroppers while tolerating considerably higher error rates than the standard model with unbounded adversaries; for the six-state protocol with one-way communication, a bit-flip error rate of up to $d$2 can be tolerated, compared to $d$3 in the standard model. The relation additionally gives a lower bound on the min-entropy key uncertainty against known-plaintext attacks when quantum ciphers are composed [0612014].

Finite-size smooth-min-entropy bounds were sharpened by Ng, Berta, and Wehner. For BB84 measurements on $d$4 qubits, they proved the tight Rényi-entropy relation

$d$5

attained on an eigenstate of either the $d$6- or $d$7-basis on each qubit, and an analogous six-state relation

$d$8

Via a one-shot quantum AEP, these become smooth min-entropy lower bounds applicable to rather small block lengths. A concrete comparison given in that work is that achieving $d$9 at $$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$0 requires $$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$1 under a previous large-$$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$2 method, while the new bound with $$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$3 requires $$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$4 (Ng et al., 2012).

Quantum memory systematically tightens the structure of entropic bounds. In the bipartite setting, Berta’s relation

$$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$5

admits an improved form

$$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$6

with

$$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$7

The extra term depends on the gap between total correlation and the accessible information about the two measurement outcomes; it strengthens lower bounds on entanglement of formation and yields an upper bound on regularized one-way distillable common randomness (Adabi et al., 2016).

Tripartite and multipartite memory-assisted relations extend this pattern. Dolatkhah et al. proved

$$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$8

which improves the complementarity-only tripartite bound and was used to obtain a lower bound for the quantum secret key rate, as well as equality conditions linked to strong subadditivity and the Koashi–Winter relations (Dolatkhah et al., 2020). Xu et al. then introduced tripartite and multipartite quantum-memory-assisted EURs for multiple measurements and arbitrary POVMs, with lower bounds tighter than those of Zhang et al. [Phys. Rev. A 108, 012211 (2023)], and applied them to relative entropy of unilateral coherence and QKD protocols (Xu et al., 6 Jan 2025).

5. Generalized entropies, POVMs, and continuous variables

Tightening strategies based on generalized entropies often proceed through auxiliary quantities such as the index of coincidence. Rastegin developed an information-diagram method in which a probability distribution $$H(A)+H(B)\ge -2\log_2 c, \qquad c=\max_{i,j}|\langle a_i|b_j\rangle|.$$9 is represented by the pair $c=1/\sqrt d$0, where

$c=1/\sqrt d$1

and $c=1/\sqrt d$2 is the Tsallis $c=1/\sqrt d$3-entropy. Instead of the crude Jensen-type bound

$c=1/\sqrt d$4

he obtained a piecewise-linear bound

$c=1/\sqrt d$5

and the corresponding Rényi inequality. The construction reproduces the Harremoës–Topsøe Shannon bound at $c=1/\sqrt d$6 and yields improved uncertainty relations for MUBs, SIC-POVMs, and equiangular tight frames (Rastegin, 2023).

For general SIC-POVMs and mutually unbiased measurements, Huang, Chen, and Wu derived state-dependent Rényi EURs from inequalities between Rényi entropy and the index of coincidence. Their lower bounds are analytically tight on intervals of sufficiently mixed states. In the Shannon limit, the MUM/MUB bound takes a piecewise form and improves the Maassen–Uffink and Wu–Yu–Mølmer bounds except at special coincidence points (Huang et al., 2020). Closely related ETF-based POVMs interpolate between orthonormal bases and SIC-POVMs. For an ETF $c=1/\sqrt d$7, the index of coincidence satisfies

$c=1/\sqrt d$8

leading to both state-dependent and state-independent Shannon, Rényi, Tsallis, and min-entropy uncertainty relations, as well as applications to entanglement detection, steering inequalities, and coherence quantification (Rastegin, 2021).

Continuous-variable tightness raises additional issues because differential entropies must account for correlations between canonically conjugate or rotated quadratures. The entropy-power formulation

$c=1/\sqrt d$9

is equivalent to the Białynicki-Birula–Mycielski relation, but it is not saturated by general rotated Gaussian pure states. A tighter form,

$\log_2 d$0

incorporates the covariance $\log_2 d$1, implies the Schrödinger–Robertson uncertainty relation exactly, and is saturated for all Gaussian pure states. The derivation is partial: it relies on a concavity assumption for the relevant functional, but is supported by extensive numerical evidence (Hertz et al., 2017). A broader review and extension to arbitrary vectors of intercommuting quadratures states a rigorous proof for all Gaussian states and extends the general inequality to arbitrary states conditionally on two assumptions, global minimality of pure Gaussian states and concavity of the functional $\log_2 d$2 (Hertz et al., 2018).

6. Applications, experiments, and current limitations

The operational role of tight entropic uncertainty relations is most visible in cryptography. Higher lower bounds on an adversary’s uncertainty translate into stronger privacy amplification statements, higher key-rate guarantees, and improved noise thresholds. In bounded- or noisy-storage two-party protocols, the number of extractable key bits is essentially governed by $\log_2 d$3, so finite-size improvements immediately lower demands on the adversary’s storage and reduce required block lengths (Ng et al., 2012). In QKD against memory-bounded eavesdroppers, the min-entropy relation of Damgård et al. allows significantly higher tolerable error rates than the unbounded-adversary model [0612014]. Multipartite quantum-memory-assisted tightenings similarly improve secret-key-rate lower bounds through the Devetak–Winter framework (Dolatkhah et al., 2020, Xu et al., 6 Jan 2025).

Beyond cryptography, tight or near-tight entropic relations have become tools for entanglement detection, steering, and coherence theory. The non-additivity of multi-observable EURs under local composition is directly exploited for bipartite and multipartite entanglement criteria (2207.13469). The geometric optimization framework yields steering thresholds with improved noise tolerance; for a qutrit two-measurement family $\log_2 d$4 at $\log_2 d$5, the majorization bound gives $\log_2 d$6, whereas the optimal bound yields $\log_2 d$7, a $\log_2 d$8 improvement (Yang et al., 31 Jan 2026). ETF- and SIC/MUM-based relations further connect uncertainty with experimentally accessible purity-dependent quantities and with the Brukner–Zeilinger approach to information (Rastegin, 2021, Huang et al., 2020).

Experimental verification of tightened entropic bounds has also been carried out. In an all-optical platform, Zhang et al. prepared Bell-like and Bell-like diagonal states, performed local projective measurements over a complete set of mutually unbiased bases, and reconstructed pre- and post-measurement density matrices by maximum-likelihood tomography. The reported average state fidelity was $\log_2 d$9. They found that lower bounds augmented by Holevo and mutual-information terms always lie above the original bounds and can nearly coincide with the measured uncertainty sum; for a Bell-diagonal state with $H$00, they reported $H$01 bits, $H$02 bits, and improved $H$03 bits, giving a perfect match within error. The same experiment also observed that entropic uncertainty is inversely correlated with coherence (Ding et al., 2019).

Several limitations remain explicit in the literature. Multiple-observable bounds are often not tight outside special families, and additivity breaks down once three or more observables are involved (2207.13469). Some of the strongest continuous-variable “tight” formulations are rigorously established only for Gaussian states or conditionally on concavity and global-minimality assumptions (Hertz et al., 2017, Hertz et al., 2018). For general finite-dimensional POVMs, current exactness is frequently numerical rather than analytic, obtained through convergent outer-approximation schemes rather than closed-form formulas (Yang et al., 31 Jan 2026). This suggests that the subject is best viewed not as a single theorem, but as a hierarchy of exact, asymptotically exact, and numerically optimal uncertainty principles adapted to different entropy functionals and measurement models.