Shadow Inequalities: A Unifying Mathematical Motif

• Shadow inequalities are mathematical constraints derived from shadow constructions that encode lower-dimensional, combinatorial, or dual information across varied domains such as quantum information, convex geometry, topology, and polyhedral projections.
• They enable rigorous nonexistence proofs, optimality conditions, and structural characterizations—for instance, in quantum error correction, multiqubit state purity, random polar volumes, and 4-manifold topology.
• The methodologies span enumerator positivity tests, Kravchuk transform constraints, convexity and monotonicity arguments, and combinatorial edge-criteria in sparse linear inequality systems.

Searching arXiv for the provided topic and cited papers to ground the article. “Shadow inequalities” is a context-dependent term rather than a single canonical theorem. In current usage it denotes several mathematically distinct frameworks in which a “shadow” encodes lower-dimensional, combinatorial, or dual information and the resulting quantities satisfy nontrivial constraints. In quantum information, shadow inequalities are positivity conditions on shadow enumerators and reductions of multipartite states, with direct applications to absolutely maximally entangled states and quantum codes (Huber et al., 2017). In multiqubit correlation theory, they appear as Kravchuk-transform constraints on sector lengths and are complemented by additional monogamy inequalities that fully characterize the pure-state region for $N\leq 5$ qubits (Serrano-Ensástiga et al., 16 Jul 2025). In convex geometry, shadow inequalities arise from shadow systems and govern convexity of $L_p$-support functions, concavity of slice volumes of $L_p$-polar bodies, and extremal properties of volumes of polars of random sets (Guo et al., 2024, Cordero-Erausquin et al., 2013). In low-dimensional topology, the term is tied to Turaev shadows and to bounds on special shadow-complexity for corks and exotic $4$-manifolds (Naoe, 2017). In polyhedral geometry, shadows are $2$-dimensional projections of high-dimensional polytopes, and sparse inequality descriptions can still force exponentially large shadows (Gärtner et al., 2013). This suggests a broad unifying motif—constraints induced by shadow constructions—while the technical content remains field-specific.

1. Quantum shadow inequalities and shadow enumerators

For multipartite quantum systems, the modern formalism begins with a pure or mixed state on $n$ parties and an orthonormal local operator basis $\{e_j\}$ satisfying

$\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$

Tensor products of these basis elements form a local error basis $\mathcal E$, and any operator $M$ admits the Bloch expansion

$L_p$0

The weight $L_p$1 is the size of the support of $L_p$2, and from this one defines the Shor–Laflamme coefficients

$L_p$3

together with their generating polynomials $L_p$4 and $L_p$5. Rains’ unitary enumerators are built from subsystem reductions,

$L_p$6

and the shadow enumerator is then defined by

$L_p$7

The generalized shadow inequalities assert that for any positive semidefinite Hermitian $L_p$8 and every fixed subset $L_p$9,

$L_p$0

Equivalently, all shadow coefficients satisfy $L_p$1. For states, these are consistency constraints on the purities of all marginals. The same paper derives the quantum MacWilliams identity in Bloch form,

$L_p$2

and the shadow transform

$L_p$3

These formulas place the shadow inequalities inside the same enumerator machinery as quantum error correction (Huber et al., 2017).

A principal application concerns absolutely maximally entangled states. A pure state $L_p$4 on $L_p$5 parties of local dimension $L_p$6 is absolutely maximally entangled if every reduction to $L_p$7 parties is maximally mixed, equivalently

$L_p$8

For AME$L_p$9, the unitary enumerators are fixed: $4$0 so the shadow coefficients become

$4$1

with $4$2 the Krawtchouk polynomials. Negativity of any $4$3 excludes existence. In this way the paper gives new nonexistence results, including

$4$4

$4$5

and

$4$6

It also derives Scott’s necessary bound

$4$7

from positivity of $4$8, and presents a mixed-dimensional maximally entangled example on $4$9, certified by an iterative semidefinite program (Huber et al., 2017).

A common misconception is to treat shadow-enumerator positivity as a complete characterization. In this framework it is necessary, not sufficient: passing the shadow test does not guarantee the existence of a corresponding state or quantum code, whereas violation gives a definitive impossibility result.

2. Multiqubit sector lengths and monogamy beyond shadow inequalities

For $2$0-qubit systems, the same terminology has a specialized form in the Pauli correlation representation. Writing

$2$1

with $2$2 and $2$3, one groups correlations by Hamming weight. The sector length

$2$4

is the average squared $2$5-body correlation content. Purity satisfies

$2$6

so for pure states

$2$7

The corresponding shadow enumerators are

$2$8

where

$2$9

are the Kravchuk polynomials. For $n$0, one recovers the time-reversal overlap

$n$1

The inverse transform is exact: $n$2 In this formulation, shadow inequalities are positivity constraints on the Kravchuk transform of the vector of sector lengths (Serrano-Ensástiga et al., 16 Jul 2025).

The paper then derives additional monogamy inequalities from reduced purities and time-reversal overlaps. For all $n$3,

$n$4

$n$5

and

$n$6

Intersecting the region $n$7 defined by these new inequalities with the shadow region $n$8 and the pure-state equalities yields a polytope

$n$9

that contains the true pure-state range $\{e_j\}$0. For $\{e_j\}$1, the characterization is complete: $\{e_j\}$2 (Serrano-Ensástiga et al., 16 Jul 2025).

The explicit small-$\{e_j\}$3 descriptions are particularly sharp. For $\{e_j\}$4,

$\{e_j\}$5

and the allowed region is

$\{e_j\}$6

Its vertices are realized by $\{e_j\}$7, $\{e_j\}$8, and the symmetric tetrahedron state

$\{e_j\}$9

For $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$0,

$\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$1

and the allowed region is

$\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$2

Its vertices are realized by $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$3, $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$4, and $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$5. Because the shadow enumerators are linear in the $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$6, extremization of quantities such as average linear entropy or $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$7 reduces to evaluation at the polytope vertices (Serrano-Ensástiga et al., 16 Jul 2025).

For $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$8, completeness fails. At $\mathrm{Tr}(e_j^\dagger e_k)=\delta_{jk}\,D .$9, neither the shadow region $\mathcal E$0 nor the new region $\mathcal E$1 contains the other; their intersection is strictly larger than the known pure-state set, and an additional constraint

$\mathcal E$2

is reported for pure states but is not implied by shadow plus new inequalities. The paper also identifies a vertex $\mathcal E$3 in the enlarged polytope with no known realizing state. This directly addresses a recurrent misunderstanding: shadow inequalities are powerful, but even when supplemented by these monogamy relations they do not yet fully determine the geometry for larger qubit numbers (Serrano-Ensástiga et al., 16 Jul 2025).

3. Convex-geometric shadow systems and polar-body inequalities

In convex geometry, a shadow system is a one-parameter family of convex bodies generated by motion along a fixed direction. The general shadow system of Rogers–Shephard and Shephard is

$\mathcal E$4

where $\mathcal E$5 and $\mathcal E$6 is a bounded speed function. A parallel chord movement is the special case

$\mathcal E$7

with $\mathcal E$8 chosen so that the moved chords remain convex. These constructions are closely tied to Steiner symmetrization through the representation

$\mathcal E$9

The $M$0-support function introduced by Berndtsson, Mastrantonis and Rubinstein is

$M$1

and the associated $M$2-polar body is defined through the gauge

$M$3

The volume representation is

$M$4

As $M$5, one recovers the classical support function and classical polar body (Guo et al., 2024).

The central shadow inequalities in this setting are convexity and concavity statements along the shadow system parameter. For a parallel movement $M$6 and fixed $M$7, the function

$M$8

is convex. For the slices of the $M$9-polar bodies one has, for all $L_p$00 and $L_p$01,

$L_p$02

The proofs use midpoint convexity, a functional shadow inclusion,

$L_p$03

Ball’s inequality for measurable functions, and the Brunn–Minkowski inequality on $L_p$04 (Guo et al., 2024).

These facts imply monotonicity under Steiner symmetrization. If $L_p$05 is symmetric and the movement is chosen so that $L_p$06, then

$L_p$07

The paper identifies this as recovering the $L_p$08 Blaschke–Santaló inequality by iteration. It also proves a reverse Rogers–Shephard type inequality: if $L_p$09 and $L_p$10 have opposite barycenters,

$L_p$11

then

$L_p$12

An upper bound in the reverse direction is also obtained by applying Rogers–Shephard’s section inequality. Here “shadow inequality” therefore refers not to an enumerator positivity constraint, but to functional convexity, volume monotonicity, and Rogers–Shephard type inequalities driven by the shadow-system deformation (Guo et al., 2024).

4. Random polars, generalized shadow systems, and Blaschke–Santaló type consequences

A second convex-geometric usage concerns generalized shadow systems and volumes of the polar of random sets. For vectors $L_p$13 and a coefficient set $L_p$14, define

$L_p$15

so that

$L_p$16

The generalized shadow system is obtained from a centrally symmetric closed convex set $L_p$17, a direction $L_p$18, and the maps

$L_p$19

The decisive analytic input is a Busemann-type theorem for convex measures: if $L_p$20 has even density and is $L_p$21-concave, then

$L_p$22

defines a norm on $L_p$23. From this, one derives that for a shadow system of centrally symmetric convex sets,

$L_p$24

is convex; in the generalized setting, the same map is convex on $L_p$25, and under additional symmetry it is even. A key corollary is that for $L_p$26, unconditional $L_p$27, and $L_p$28, the map

$L_p$29

is even and convex (Cordero-Erausquin et al., 2013).

This shadow convexity is then combined with Christ’s rearrangement principle to compare random models. Let $L_p$30 be independent random vectors whose laws belong to $L_p$31, the class of Borel probability measures on $L_p$32 with densities bounded by $L_p$33. Let $L_p$34 be unconditional, and let $L_p$35 be spherically invariant with decreasing radial density. If $L_p$36 are i.i.d. uniform on the volume-one Euclidean ball $L_p$37, then

$L_p$38

Under the additional assumption that $L_p$39 is convex for the radial density $L_p$40, there is also stochastic dominance: $L_p$41 For $L_p$42, this specializes to

$L_p$43

and the expected volume of $L_p$44 is maximized when the $L_p$45 are i.i.d. uniform on $L_p$46 (Cordero-Erausquin et al., 2013).

The same framework yields a random extension of the Blaschke–Santaló inequality. If $L_p$47 are i.i.d. uniform on a fixed symmetric convex body $L_p$48 with $L_p$49, then $L_p$50 in Hausdorff metric almost surely. Passing to the limit gives

$L_p$51

This recovers the classical Blaschke–Santaló inequality from the random extremal model. In this literature, therefore, “shadow inequalities” are inequalities for polar volumes and related functionals that are induced by convexity properties of shadow systems together with rearrangement methods (Cordero-Erausquin et al., 2013).

5. Shadow-complexity inequalities in $L_p$52-manifold topology

In $L_p$53-manifold topology, a shadow is a simple polyhedron $L_p$54 properly embedded in a compact, oriented smooth $L_p$55-manifold $L_p$56, locally flat in $L_p$57, and a strong deformation retract of $L_p$58. A region is a connected component of $L_p$59, and a simple polyhedron is special if all regions are open disks. Each internal region carries a half-integer gleam $L_p$60 satisfying the parity condition

$L_p$61

Turaev’s reconstruction associates to a shadowed polyhedron $L_p$62 a $L_p$63-manifold $L_p$64. This leads to the complexity invariants

$L_p$65

By definition,

$L_p$66

A central technical tool is the Ishikawa–Koda slope-length criterion. If $L_p$67 is a special shadow of a $L_p$68-manifold $L_p$69 with $L_p$70 true vertices and each region $L_p$71 satisfies

$L_p$72

where $L_p$73 counts adjacent true vertices with multiplicity, then

$L_p$74

This criterion is used to pass exact lower bounds from $L_p$75-manifold boundaries to $L_p$76-manifolds (Naoe, 2017).

The paper constructs, for each positive integer $L_p$77, an infinite family $L_p$78 of Mazur-type corks. The twisting parameters are chosen so that

$L_p$79

and $L_p$80 is defined by a parity-dependent piecewise formula with leading scale $L_p$81. An explicit special shadow $L_p$82 gives the upper bound

$L_p$83

which is packaged as

$L_p$84

where

$L_p$85

In particular,

$L_p$86

with $L_p$87 independent of $L_p$88. The lower bound comes from the boundary: $L_p$89 and hence

$L_p$90

The boundaries are shown to be mutually nonhomeomorphic as $L_p$91 varies via Casson invariants and the surgery formula

$L_p$92

The same methods produce exotic pairs $L_p$93 of $L_p$94-manifolds with boundary, obtained by attaching a $L_p$95-framed $L_p$96-handle to a meridian of the dotted circle or of the $L_p$97-framed circle. Because $L_p$98 is a cork, $L_p$99 is the cork-twist of $L_p$00. Their complexities satisfy

$L_p$01

and

$L_p$02

Therefore,

$L_p$03

where $L_p$04 (Naoe, 2017).

The paper also addresses the small-complexity regime. It exhibits an exotic pair $L_p$05 with

$L_p$06

showing that shadow-complexity zero is compatible with exotic pairs when one does not require special shadows. By contrast, the lowest special shadow-complexity among exotic pairs of $L_p$07-manifolds with boundary is proved to be $L_p$08 or $L_p$09. This distinguishes ordinary shadow-complexity from special shadow-complexity and clarifies that the large-complexity cork families complement, rather than contradict, the existence of exotic behavior at very small non-special shadow-complexity (Naoe, 2017).

6. Large planar shadows from sparse linear inequalities

In polyhedral geometry, a shadow is the image of a polytope under a $L_p$10-dimensional linear projection. If

$L_p$11

and $L_p$12 are linearly independent, then

$L_p$13

defines the shadow $L_p$14. A vertex $L_p$15 survives as a vertex of $L_p$16 if there exists $L_p$17 such that $L_p$18 uniquely maximizes $L_p$19 over $L_p$20. The paper uses this criterion to study the combinatorial size of shadows for sparse inequality systems (Gärtner et al., 2013).

A $L_p$21-sparse inequality system is one in which each row of $L_p$22 has at most $L_p$23 nonzero entries. The starting point is the contrast between the $L_p$24-sparse Goldfarb cube and the $L_p$25-sparse Klee–Minty cube. Goldfarb’s construction yields a polytope whose $L_p$26-dimensional shadow contains all $L_p$27 vertices. The paper then proves that the same exponential shadow complexity already occurs in a $L_p$28-sparse system. The relevant Klee–Minty variant is

$L_p$29

with $L_p$30. Its vertices are indexed by bit vectors $L_p$31 through the recursion

$L_p$32

The main theorem chooses the projection plane using

$L_p$33

and proves that the shadow

$L_p$34

has $L_p$35 vertices. Equivalently, every vertex $L_p$36 of the Klee–Minty cube appears as a vertex of the planar shadow. The proof uses an edge-local optimality criterion: a vertex uniquely maximizes a linear functional if and only if it strictly beats all its edge-neighbors. By constructing, for each $L_p$37, a vector $L_p$38 with carefully separated $L_p$39-scales, the argument shows that the sign of $L_p$40 is negative for every neighboring vertex. This yields a maximal $L_p$41-vertex shadow from inequalities with only two variables per constraint (Gärtner et al., 2013).

The same paper proves that this behavior disappears for $L_p$42-sparse systems. If

$L_p$43

is an axis-parallel box, then every $L_p$44-dimensional shadow has at most $L_p$45 vertices. The bound is tight for generic projections. The contrast is therefore sharp: with one variable per inequality, shadow size is $L_p$46, whereas with two variables per inequality it can be $L_p$47. The paper interprets this as a limit result for shadow-based linear-programming methods, especially the shadow-vertex method of Gass–Saaty. Large shadows imply exponentially many bends in the parametric path, so sparsity alone does not preclude worst-case shadow complexity (Gärtner et al., 2013).