---
title: Polynomial Complexity Theorem
url: https://www.emergentmind.com/topics/polynomial-complexity-theorem
type: topic
---

# Polynomial Complexity Theorem

In the literature represented here, “Polynomial Complexity Theorem” does not denote a single canonical theorem. It designates a recurring theorem pattern: a result establishing polynomial-time computability, polynomial-size proofs, polynomial bond dimension, polynomial tropical complexity, polynomial-hierarchy containment, or a sharp tractable/intractable dichotomy under explicitly stated structural hypotheses. Typical instances include exact integration over rational simplices under fixed structural parameters [0809.2083], polynomial tropical semiring complexity for Schur and skew Schur polynomials [1710.03312], the compact real analogue of Toda’s theorem \({\bf PH}^c_{\mathbb R}\subset {\bf P}_{\mathbb R}^{\#{\bf P}_{\mathbb R}^{\dagger}}\) [0812.1200], restricted polynomial-size propositional proofs for Kneser–Lovász formulas [1402.4338], algorithmic Polynomial Freiman–Ruzsa over \(\mathbb F_2^n\) [2604.04547], and polynomial bond-dimension consequences of spectral small-incremental-entangling for 1D long-range quantum systems [2509.12014]. This suggests that the phrase functions less as a standardized theorem name than as a theorem schema for polynomially controlled structure.

## 1. Taxonomy of polynomial-complexity statements

Across these works, several recurrent theorem types appear: structural upper bounds, complexity-class containments, dichotomy theorems, polynomial proof-complexity upper bounds in strong systems, and structure-vs-randomness transfer principles. The common feature is not the domain but the form of the conclusion: a quantitatively polynomial bound replaces an exponential, quasi-polynomial, or nonconstructive statement.

| Setting | Polynomial statement | Representative paper |
|---|---|---|
| Rational simplex integration | Polynomial time for fixed effective variable count or fixed degree; NP-hard in general | [0809.2083] |
| Tropical symmetric polynomials | \(\operatorname{Trop}(s_{\lambda/\mu})\) has complexity \(O(n^2\lambda_1)\) | [1710.03312] |
| Real semi-algebraic complexity | \({\bf PH}_{\mathbb R}^c \subset {\bf P}_{\mathbb R}^{\#{\bf P}_{\mathbb R}^{\dagger}}\) | [0812.1200] |
| Homomorphism polynomials | \(VAC\)/\(VNP\) dichotomy by fixed target graph \(H\) | [1210.7641] |
| Additive combinatorics | Polynomial-time algorithmic PFR recovery of a covering subspace | [2604.04547] |
| Long-range quantum systems | Polynomial bond-dimension approximation theorems | [2509.12014] |

A second recurring feature is that many such theorems are conditional on a structural regime rather than unconditional on all inputs. Fixed degree, fixed effective variable number, fixed proof system, bounded doubling, compactness, or \(\eta>2\) long-range decay are typical hypotheses. A common misconception is therefore that “polynomial complexity theorem” always means unrestricted polynomial-time solvability; in much of the literature, it instead means that a previously uncontrolled or quasi-polynomial phenomenon becomes polynomial once the correct invariant is identified.

## 2. Algebraic, tropical, and proof-theoretic constructions

In exact symbolic computation, Baldoni, Berline, De Loera, Köppe, and Vergne establish a sharp classification for integrating a polynomial over a rational simplex: exact integration is NP-hard for general polynomials, but polynomial-time solvable when the polynomial depends on a fixed number \(D\) of linear forms, and also polynomial-time solvable for fixed total degree \(M\) [0809.2083]. The theorem is paradigmatic because it isolates the precise source of tractability: not low ambient dimension, but low effective algebraic complexity of the integrand.

Woo and Yong prove a different kind of polynomial-complexity theorem in tropical algebraic combinatorics. If a homogeneous symmetric polynomial \(f\in \mathrm{Sym}_n\) is dominated by a Schur polynomial \(s_\lambda\), then \(\operatorname{Trop}(f)=\operatorname{Trop}(s_\lambda)\), and hence its tropical semiring complexity is at most \(O(n^2\lambda_1)\); for every skew shape \(\lambda/\mu\), they identify an explicit partition \(\beta\) with \(\beta_1=\lambda_1\) such that \(\operatorname{Trop}(s_{\lambda/\mu})=\operatorname{Trop}(s_\beta)\), yielding the same \(O(n^2\lambda_1)\) bound for all skew Schur polynomials [1710.03312]. The underlying mechanism is geometric: equality of Newton polytopes plus the saturated Newton polytope property collapses tropical complexity to that of a single dominating Schur polynomial.

Bürgisser’s factor theorem supplies an algebraic-complexity closure principle of a different kind. Over characteristic-zero fields, if \(g\) is a degree-\(d\) factor of a polynomial of low approximative complexity, then \(\underline{L}(g)\) is polynomially bounded in \(d\) and in the approximative complexity of a nonzero multiple of \(g\), specifically
\[
\underline{L}(g)\le O\big(M(d)M(d^4)\,\underline{(g)}+d^{2\gamma}M(d)^2\big),
\]
and this yields graph-checkability \(\Rightarrow\) approximative computability for bounded-degree polynomial families [1812.06828]. The theorem extends Kaltofen’s factor-complexity result by removing multiplicity dependence at the price of passing from straight-line complexity to approximative complexity.

A further constructive variant appears in the new proof of the Schwartz–Zippel lemma. For a degree-\(d\) circuit \(P\) in \(n\) variables and a cube \(S_q^n\), the paper constructs a polynomial-time computable and polynomial-time invertible surjection
\[
dec_{P,\overline a}: C_{n,d,q}\twoheadrightarrow Z_{P,q},
\qquad
C_{n,d,q}=[n]\times[d]\times S_q^{\,n-1},
\]
onto the root set \(Z_{P,q}\), formalizable in \(S^1_2\) [2411.07966]. This yields the exact compression bound \(|Z_{P,q}|\le ndq^{n-1}\), allows \(S^1_2+\mathrm{dWPHP(PV)}\) to prove polynomial-size circuits for PIT, proves existence of small hitting sets for every explicitly described class of polynomial-degree circuits, and shows that existence of such hitting sets is equivalent over \(S^1_2\) to \(\mathrm{dWPHP(PV)}\). The constructive content is stronger than the classical root count: it turns a counting lemma into a feasible coding theorem.

## 3. Complexity containments and numerical continuation

The real analogue of Toda’s theorem is one of the clearest uses of “polynomial complexity theorem” as a containment theorem. Basu proves
\[
{\bf PH}_{\mathbb R}^c \subset {\bf P}_{\mathbb R}^{\#{\bf P}_{\mathbb R}^{\dagger}},
\]
where \(\#{\bf P}_{\mathbb R}^{\dagger}\) is defined by Poincaré polynomials of fibers rather than finite cardinality, and also proves that for every \(\omega>0\), \({\bf GDP}_\omega^c\) reduces in deterministic polynomial time in the Blum–Shub–Smale model to computing Poincaré polynomials [0812.1200]. The theorem replaces discrete counting by topological invariants—Betti numbers and Poincaré polynomials—as the operative polynomial resource in semi-algebraic complexity. The compactness restriction is essential rather than cosmetic; the paper does not prove the full noncompact analogue.

Beltrán and Kozhasov give an average-case polynomial-complexity refinement for Smale’s 17th problem via homotopy continuation. Their randomized algorithm for square homogeneous polynomial systems has average number of continuation steps bounded by
\[
C D^{3/2} n N^{1/2},
\]
improving the earlier \(O(D^{3/2}nN)\) bound, and hence average arithmetic complexity \(O(D^{3/2}nN^{3/2})\) since one Newton step costs \(O(N)\) [1507.03896]. They also reprove the deterministic Bürgisser–Cucker theorem by homotopy methods alone: there is a deterministic real-number algorithm that computes an approximate zero in average time \(N^{O(\log\log N)}\), and polynomial average time when \(D\le n^{1/(1+\varepsilon)}\) or \(D\ge n^{1+\varepsilon}\). Here the polynomial-complexity content lies not in worst-case decision complexity but in average certified continuation complexity governed by condition length.

These containment and continuation theorems exhibit a characteristic pattern. Polynomial complexity is not stated as a direct bound on naïve symbolic manipulation; it is mediated by an invariant—Poincaré polynomial, condition number, or condition length—that compresses the effective hardness of the problem.

## 4. Proof complexity and algebraic search complexity

In propositional proof complexity, Istrate and Crăciun show that formulas encoding the Kneser–Lovász theorem form a genuine generalization of the pigeonhole principle: \(Kneser_{1,n}\) is essentially \(PHP_{n-1}^n\), and for every fixed \(k\ge 1\),
\[
res(Kneser_{n,k})=2^{\Omega(n)},
\]
while bounded-depth Frege proofs require size \(\Omega(2^{n^{\epsilon_d}})\) for every fixed depth \(d\) [1402.4338]. At the same time, they prove polynomial-size upper bounds only in restricted low-\(k\) cases: \(Kneser^{onto}_{2,n}\) has polynomial-size Frege proofs, and \(Kneser^{onto}_{3,n}\) has polynomial-size extended Frege proofs. The paper does not prove polynomial-size Frege or EF proofs for arbitrary fixed \(k\), and it explicitly leaves polynomial-size ordinary Frege proofs for \(k=3\) open. The resulting “polynomial complexity theorem” is therefore sharply delimited: the theorem isolates exactly where polynomial proof length is known, against a background of exponential lower bounds in weaker systems.

Göös, Kaminski, and Sosnovec provide a complementary complexity classification for algebraic total search. They prove that PPA-CIRCUIT CNSS and PPA-CIRCUIT CHEVALLEY, two search problems derived from the Combinatorial Nullstellensatz and Chevalley–Warning theorem over \(\mathrm{GF}(2)\), are PPA-complete [1710.08602]. The key technical device is the PPA-circuit: an arithmetic circuit whose maximal parse subcircuits can be paired in polynomial time, forcing parity of top-degree monomials and hence totality. This is not a polynomial-time solvability theorem, but it is a polynomial-time *classification* theorem: the algebraic parity principle is shown to sit exactly at PPA-complete search complexity.

These results correct another common misconception. In proof complexity, a polynomial-complexity theorem may mean a theorem about *proof size* or *search-class completeness*, not about decision-time tractability. The notion is thus proof-theoretic and structural as much as algorithmic.

## 5. Dichotomies, additive structure, and polynomial patterns

A dichotomy theorem gives perhaps the cleanest form of polynomial-complexity classification. De Rugy-Altherre proves that for every fixed graph \(H\), the homomorphism polynomial family \(f_n^H\) lies in \(VAC\) if \(H\) has a loop or no edges, and is \(VNP\)-complete for c-reductions otherwise, over \(\mathbb Q\) [1210.7641]. The theorem classifies an entire family of generating functions by a simple graph-theoretic invariant of the target. In the same paper, the cut polynomial \(Cut_n\) is shown to be \(VNP\)-complete over \(\mathbb Q\), illustrating strong field dependence in arithmetic complexity.

The algorithmic Polynomial Freiman–Ruzsa theorem of Bhattacharyya, Ghosh, and Xie turns a structural additive-combinatorics theorem into an explicit recovery algorithm. Given \(A\subseteq \mathbb F_2^n\) with doubling constant \(K\), there is a randomized polynomial-time algorithm that returns a basis of a subspace \(V\subseteq \mathbb F_2^n\) satisfying \(|V|\le |A|\) and such that \(A\) can be covered by \(2K^C\) translates of \(V\), for a universal constant \(C>1\); the informal theorem gives runtime \(\widetilde O(n^4)\), \(O(\log|A|)\) random samples, and \(\widetilde O(\log^2|A|)\) queries to \(A\) [2604.04547]. The same framework also yields algorithmic versions of the polynomial \(U^3\)-inverse theorem, approximate Freiman homomorphism classification, and quadratic structure-vs-randomness decompositions. Here the polynomial-complexity theorem is genuinely algorithmic: a recent existential theorem is turned into a structure-recovery procedure.

Leng’s quantitative polynomial Szemerédi theorem supplies a transfer principle of a different kind. For linearly independent \(P,Q\in\mathbb Z[x]\) with \(P(0)=Q(0)=0\), if \(A\subseteq \mathbb Z/N\mathbb Z\) contains no progression
\[
(x,\ x+P(y),\ x+Q(y),\ x+P(y)+Q(y)),
\]
then
\[
|A|\le O\!\left(\frac{N}{\log_{(O(1))}(N)}\right)
\]
for \(N\) prime [2205.05540]. More generally, the paper shows that if one can establish polynomial-type bounds on the true complexity of a polynomial progression, then one can establish polynomial-type bounds on Szemerédi’s theorem for that progression. The theorem thus links a quantitative generalized von Neumann statement to a density theorem; “polynomial complexity” refers to polynomial dependence in the controlling Gowers-uniformity parameter rather than directly to runtime.

In additive combinatorics and arithmetic circuit complexity alike, the decisive phenomenon is the same: once the correct structural invariant—doubling, true complexity, or fixed target graph—is exposed, a previously diffuse combinatorial statement collapses to a polynomially controlled classification.

## 6. Entanglement structure and polynomial simulation of long-range quantum systems

In 1D long-range quantum systems, the phrase takes yet another form: a polynomial-complexity theorem becomes a polynomial bond-dimension theorem. The spectral small-incremental-entangling theorem proves that for a bipartite Hamiltonian \(H=H_A+H_B+V_{AB}\),
\[
\frac{dE_\alpha(\psi_t)}{dt}\le c_\alpha\,\bar{\mathcal J}(V_{AB}), \qquad \alpha\ge \frac12,
\]
with \(c_{1/2}=2\), \(c_1=4/e\), and \(c_\infty=2\) [2509.12014]. Here \(\bar{\mathcal J}(V_{AB})\) is the spectral-entangling strength, defined through the maximal \(\ell_1\)-mass of the Schmidt vector created by \(V_{AB}\) on a product input. This yields a universal \(1/s^2\) tail on squared Schmidt coefficients, identifies \(\alpha=\tfrac12\) as the sharp threshold beyond which universal growth bounds fail, and produces rigorous truncation-error control.

From this spectral control, the paper derives polynomial bond-dimension approximation theorems for 1D power-law interactions \(J(r)=J_0 r^{-\eta}\) with \(\eta>2\). Ground states admit MPS approximants whose bond dimension can be chosen polynomial in \(n/\epsilon\); for time-evolved product states, there exists an MPS \(|\phi_t^{(D)}\rangle\) with
\[
\|e^{-iHt}|\phi\rangle-|\phi_t^{(D)}\rangle\|^2 \le \frac{2e^{2\tilde J t}}{D}\,n;
\]
and thermal states admit purified MPS approximants with bond dimension of order \((n/\epsilon)^{O(\beta)}\) [2509.12014]. The same framework yields the first rigorous precision-guarantee bound for t-DMRG. The paper is explicit, however, that ground-state and thermal-state results are primarily existential representation theorems, whereas the t-DMRG result is an explicit algorithmic guarantee. Polynomial bond dimension therefore implies polynomial representation complexity and polynomial local-observable evaluation, but not automatically a complete constructive polynomial-time algorithm for every state class.

A broadly similar lower-bound usage of the phrase occurs in pseudorandomness theory: Popoli proves that for every monic \(P\in \mathbb Z[X]\) of degree \(d\ge 2\) with \(P(\mathbb N)\subset \mathbb N\),
\[
M(\mathcal T_P,N)\gg N^{1/d},
\qquad
M(\mathcal P_{k,P},N)\gg N^{1/d},
\]
and hence in particular \(M(\mathcal R_P,N)\gg N^{1/d}\) for Rudin–Shapiro along polynomial values [2011.03457]. Here “polynomial complexity” no longer means efficient solvability, but a polynomial lower bound on predictive complexity along polynomial subsequences.

Taken together, these results show that “Polynomial Complexity Theorem” is best treated as an encyclopedic umbrella for theorems that polynomialize a complexity statement in a precise ambient theory. Sometimes the theorem proves tractability, sometimes it proves polynomial-size certificates or approximants, sometimes it establishes a dichotomy, and sometimes it quantifies a lower bound. What unifies the class is the replacement of uncontrolled growth by an explicit polynomial law tied to a structural invariant.

Source: https://www.emergentmind.com/topics/polynomial-complexity-theorem