A Quadratic Lower Bound for Noncommutative Circuits

Abstract: We prove that every fan-in $2$ noncommutative arithmetic circuit computing the palindrome polynomial has size $Ω(n^2)$. The proof builds on and refines a previous work of the author. The new ingredients in the proof were generated by Gemini 3.1 Pro.

• The paper establishes an optimal Ω(n²) lower bound for fan-in 2 noncommutative arithmetic circuits computing the palindrome polynomial.
• It employs a refined coefficient space paradigm and a novel counting argument that parameterizes circuit-generated elements via degree truncation.
• The results resolve longstanding gaps in noncommutative circuit complexity and pave the way for future research extending these techniques to broader circuit classes.

A Quadratic Lower Bound for Fan-in 2 Noncommutative Arithmetic Circuits

Introduction and Context

The study establishes a tight $\Omega(n^2)$ lower bound for fan-in 2 noncommutative arithmetic circuits tasked with computing the palindrome polynomial $\Pal_{n,n}(X,Y)$. Whereas general noncommutative circuit lower bounds for explicit degree-$\Theta(n)$ polynomials have remained at $\Omega(n\log n)$ for decades—beginning with Strassen [Stra73], Baur-Strassen [BS83], and encapsulated in Nisan's foundational separation between formulas and circuits [nisan-stoc1991]—this work pushes past that frontier for fan-in 2 circuits, even without homogeneity or syntactic degree restrictions.

Prior advances, such as the $\Omega(n^2)$ lower bounds for homogeneous circuits by Chatterjee and Hrubeš [chatterjee-hrubes-ccc2023] and the nearly quadratic bounds for bounded syntactic degree circuits by the author [shastri-itcs2026], left open the quantitative limitations for unrestricted fan-in 2 circuits. This paper resolves that longstanding question by leveraging and further refining the coefficient space and spanning set frameworks originating from Nisan, and by developing a precise counting argument that integrates degree-truncation with a sharp parameterization of circuit-generated elements.

Main Result

The principal theorem demonstrated is:

Every fan-in 2 noncommutative arithmetic circuit computing $\Pal_{n,n}(X,Y)$ must have circuit size $s = \Omega(n^2)$.

This lower bound matches exactly the known $O(n^2)$ upper bound arising from the natural recursive construction of the palindrome polynomial, closing the complexity gap for this archetypal problem.

Proof Overview and Technical Approach

The proof operates in the abstract algebraic setting $F\langle X,Y\rangle$, where $X$ and $\Pal_{n,n}(X,Y)$0 are disjoint collections of $\Pal_{n,n}(X,Y)$1 noncommuting indeterminates, and considers circuits whose DAGs use additive and multiplicative gates (each with in-degree 2). The analysis proceeds via these key technical steps:

Coefficient Space Paradigm

• The polynomial $\Pal_{n,n}(X,Y)$2 is $\Pal_{n,n}(X,Y)$3-separated, so each nonzero monomial is constructed by a block of $\Pal_{n,n}(X,Y)$4 variables followed by a block of $\Pal_{n,n}(X,Y)$5 variables in reverse order.
• The set $\Pal_{n,n}(X,Y)$6 consists of all degree $\Pal_{n,n}(X,Y)$7 monomials over $\Pal_{n,n}(X,Y)$8, leading immediately to $\Pal_{n,n}(X,Y)$9.

Spanning Set Construction

• The previous construction [shastri-itcs2026] builds an ascending series of sets $\Theta(n)$0. Each $\Theta(n)$1 can be interpreted as polynomials in $\Theta(n)$2 derived from the circuit's gates, structured so that $\Theta(n)$3.
• These sets are assembled inductively by introducing terminal factors from circuit gates and repeatedly left-multiplying by $\Theta(n)$4-only components generated at product gates, strictly respecting syntactic degree class order.

Degree Truncation and Parameterization

• A central insight is that each nonzero element in $\Theta(n)$5 admits a unique parameterization of the form $\Theta(n)$6, where each $\Theta(n)$7 is an $\Theta(n)$8-only multiplier from a strictly decreasing sequence of syntactic degree classes and $\Theta(n)$9 is either a base or an $\Omega(n\log n)$0-only output of a product gate.
• By minimal degree analysis (Lemma~\ref{lem:degree-truncation}), once $\Omega(n\log n)$1 in any such parameterization, the degree $\Omega(n\log n)$2 component vanishes. Therefore, only parameterizations with $\Omega(n\log n)$3 can contribute to the span of degree $\Omega(n\log n)$4 elements.

Sharp Counting Argument

• The total number of distinct elements in $\Omega(n\log n)$5 whose degree $\Omega(n\log n)$6 projection is nonzero is upper bounded by $\Omega(n\log n)$7, where $\Omega(n\log n)$8 is the circuit size.
• Comparing this count to the required spanning dimension $\Omega(n\log n)$9 (inherited from $\Omega(n^2)$0), it is shown that $\Omega(n^2)$1 is necessary for the inclusion to be possible for large enough $\Omega(n^2)$2.

Significance, Implications, and Outlook

The demonstrated $\Omega(n^2)$3 lower bound for fan-in 2 noncommutative circuits resolving $\Omega(n^2)$4 shatters the extended impasse at $\Omega(n^2)$5 for size lower bounds in noncommutative arithmetic circuit complexity. Notably, the techniques avoid homogeneity and bounded syntactic degree restrictions, emphasizing their applicability in the general (and most challenging) setting.

This result tightly characterizes the fan-in 2 regime but does not yield superpolynomial lower bounds for general (unbounded fan-in) circuits—a milestone that remains out of reach. By clarifying the tradeoff between syntactic structure (fan-in and homogeneity) and computational power, the work refines the frontier for future advances, including:

• Potential adaptation or extension of these techniques to circuits with higher or unbounded fan-in, and to other explicit noncommutative polynomials;
• Exploration of coefficient space techniques for commutative circuit classes, where the critical dimension arguments are less sharp;
• Formulating new combinatorial or algebraic invariants that can robustly lower-bound the expressiveness of more general arithmetical computations.

Additionally, the proof process itself demonstrates the effective synergy with advanced AI systems in technical theorem invention, as noted in the paper's historical remarks on the genesis of several key lemmata via collaboration with Gemini 3.1 Pro.

Conclusion

This paper establishes an optimal $\Omega(n^2)$6 size lower bound for fan-in 2 noncommutative arithmetic circuits computing the explicit palindrome polynomial $\Omega(n^2)$7, integrating a refined parameterization and counting argument into the coefficient space lower bound method. This advancement represents a major increment in noncommutative circuit complexity, resolving a question that has persisted since the inception of the field (2604.20575).