Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Border determinantal complexity is at most determinantal complexity

Proved
GCT.dcBar_le_dc

by LG · Sep 14, 2026 · Mathlib 0df444a (Lean v4.33.1)

algebraic-geometrycomplexity-theorygeometric-complexity-theory

Since End(Cn2)⋅det⁡n\mathrm{End}(\mathbb{C}^{n^2}) \cdot \det_nEnd(Cn2)⋅detn​ is contained in its own closure End(Cn2)⋅det⁡n‾=Detn\overline{\mathrm{End}(\mathbb{C}^{n^2}) \cdot \det_n} = \mathrm{Det}_nEnd(Cn2)⋅detn​​=Detn​, every determinantal expression is in particular a border expression, so

dc‾(permm)  ≤  dc(permm)\overline{\mathrm{dc}}(\mathrm{perm}_m) \;\le\; \mathrm{dc}(\mathrm{perm}_m)dc(permm​)≤dc(permm​)

for all mmm. This inequality is what makes the GCT flagship conjecture a strengthening of Valiant's conjecture, and it is what transfers a lower bound for dc‾\overline{\mathrm{dc}}dc to one for dc\mathrm{dc}dc.

Preamble
import Definitions.Def_GCT_determinantal_complexity
Formal statement
namespace GCT

theorem dcBar_le_dc (m : ℕ) : dcBar m ≤ dc m := by sorry

end GCT
Source
J. M. Landsberg, *Geometric Complexity Theory: an introduction for geometers*, arXiv:1305.7387v3, https://arxiv.org/abs/1305.7387, p. 4, §2.1 ('Define dc(P) to be the smallest n such that l^(n-m) P ∈ End(W)·det_n, so dc-bar(P) ≤ dc(P)'); cf. J. M. Landsberg, *Geometry and Complexity Theory*, Cambridge Studies in Advanced Mathematics 169, CUP 2017, DOI 10.1017/9781108183192, p. 16, Definition 1.2.5.1.
Read-back

What the Lean code literally says, in plain math · Aristotle (Harmonic)

Provenance note — this read-back is NOT blind and NOT independent. It was written by the same agent that drafted the Lean statement it describes, at the explicit instruction of the human who owns this proposal, and not by a separate auditor with a fresh context. It is therefore self-testimony: any misreading made while drafting the statement may be repeated here. Reviewers should not treat it as independent confirmation that the formalization is faithful to the source.

The statement is phrased in terms of the following notions.

For a natural number mmm, the permanent polynomial is the element of the polynomial ring C[yij]\mathbb{C}[y_{ij}]C[yij​] in m2m^2m2 commuting indeterminates given by

permm  =  ∑σ∈Sm  ∏iyσ(i) i,\mathrm{perm}_m \;=\; \sum_{\sigma \in S_m} \; \prod_{i} y_{\sigma(i)\, i},permm​=σ∈Sm​∑​i∏​yσ(i)i​,

the sum running over all permutations of the mmm-element index set (for m=0m = 0m=0 the conventions give perm0=1\mathrm{perm}_0 = 1perm0​=1).

A second copy of the permanent is taken in the larger variable set consisting of the m2m^2m2 variables yijy_{ij}yij​ together with one extra variable ℓ\ellℓ; in that ring the padded permanent of size nnn is

pad(m,n)  =  ℓ n−m⋅permm,\mathrm{pad}(m, n) \;=\; \ell^{\,n - m} \cdot \mathrm{perm}_m,pad(m,n)=ℓn−m⋅permm​,

where n−mn - mn−m is truncated natural-number subtraction: if n<mn < mn<m the exponent is 000 and the padded permanent is just permm\mathrm{perm}_mpermm​.

For a polynomial ppp in any variable set:

  • ppp has a linear determinantal representation of size nnn when there exists an n×nn \times nn×n matrix AAA of polynomials such that every entry AijA_{ij}Aij​ is homogeneous of degree 111 (the zero polynomial counts as homogeneous of degree 111, so entries may vanish, but nonzero constants are not allowed) and det⁡A=p\det A = pdetA=p.
  • ppp has a border linear determinantal representation of size nnn when there is a sequence A(k)A^{(k)}A(k), k∈Nk \in \mathbb{N}k∈N, of n×nn \times nn×n matrices all of whose entries are homogeneous of degree 111, such that for every monomial exponent vector ddd the scalars coeffd(det⁡A(k))\mathrm{coeff}_d(\det A^{(k)})coeffd​(detA(k)) converge in C\mathbb{C}C, as k→∞k \to \inftyk→∞, to coeffd(p)\mathrm{coeff}_d(p)coeffd​(p). Convergence is required coefficientwise, for each exponent vector separately; no uniformity across exponents is required.
  • ppp has an affine determinantal representation of size nnn when there is an n×nn \times nn×n matrix AAA of polynomials with deg⁡Aij≤1\deg A_{ij} \le 1degAij​≤1 for all i,ji, ji,j (constants permitted) and det⁡A=p\det A = pdetA=p.

Finally,

dc(m)=inf⁡{ n∈N  :  m≤n and pad(m,n) has a linear determinantal representation of size n },\mathrm{dc}(m) = \inf\{\, n \in \mathbb{N} \;:\; m \le n \text{ and } \mathrm{pad}(m, n) \text{ has a linear determinantal representation of size } n \,\},dc(m)=inf{n∈N:m≤n and pad(m,n) has a linear determinantal representation of size n}, dc‾(m)=inf⁡{ n∈N  :  m≤n and pad(m,n) has a border linear determinantal representation of size n },\overline{\mathrm{dc}}(m) = \inf\{\, n \in \mathbb{N} \;:\; m \le n \text{ and } \mathrm{pad}(m, n) \text{ has a border linear determinantal representation of size } n \,\},dc(m)=inf{n∈N:m≤n and pad(m,n) has a border linear determinantal representation of size n},

each the least element of a set of natural numbers. This is a total function: if the set is empty the value is 000 by convention, so every assertion below also speaks about that junk value whenever the relevant set happens to be empty.

For every natural number mmm, the assertion is the inequality of natural numbers

dc‾(m)  ≤  dc(m),\overline{\mathrm{dc}}(m) \;\le\; \mathrm{dc}(m),dc(m)≤dc(m),

with no hypothesis on mmm. Both sides are least elements of sets of natural numbers, each taking the value 000 when its set is empty; in particular the inequality holds trivially at any mmm where the border set is empty, and it is a genuine constraint only where dc(m)\mathrm{dc}(m)dc(m) is the size of an actual representation.

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me