Rank-One Determinant Expansion

• Rank-One Determinant Expansion is an innovative method that expresses a matrix determinant as a concise sum of products of linear forms, drastically reducing term counts compared to the Leibniz formula.
• It uncovers deep connections between tensor, Waring, and partition ranks, providing new insights into algebraic and combinatorial structures underlying determinants.
• Applications include efficient determinant updates via the matrix-determinant lemma and advances in algebraic complexity, symbolic computation, and combinatorial optimization.

A rank-one determinant expansion is an explicit formula for the determinant of a matrix that expresses it as a sum of products, each term constructed as a product of linear forms in the rows, or through explicitly summing over objects with “rank-one” algebraic or combinatorial structure. This expands the classical Leibniz formula, providing decompositions with significantly fewer terms, unveiling the determinant’s tensor structure and offering deep connections to tensor rank, Waring rank, and geometric complexity theory. Several prominent lines of research have developed, including new combinatorial expansions, analyses of rank-one affine perturbations, closure properties under approximation, and path-integral approaches for operators formed from rank-one components.

1. Historical Formulations and Bell Number Expansion

The Leibniz formula for the determinant of an $n \times n$ matrix $A$ is

$$\det(A) = \sum_{\sigma \in S_n} \mathrm{sgn}(\sigma) \prod_{i=1}^n a_{i,\sigma(i)},$$

a sum over all $n!$ permutations. In “A New Formula for the Determinant and Bounds on Its Tensor and Waring Ranks,” Houston, Goucher, and Johnston introduced a dramatically more concise expansion (Houston et al., 2023). Their formula exploits a summation over partial set partitions of $[n]$ with no singleton blocks (denoted $PP(n)$): $$\det(A) = \sum_{P \in PP(n)} \left( \prod_{S \in P} (-1)^{|S|+1} \right) |P|! \prod_{i=1}^n \begin{cases} \sum_{j \sim_P i; j \neq i} a_{i,j}, &\text{if}\ i\in \text{block of }P, \ a_{i,i} + \sum_{j \sim_P i} a_{i,j}, &\text{otherwise.} \end{cases}$$ The number of terms equals the $n$th Bell number $B_n$ (partitions of $[n]$), with $B_n$ scaling as $\exp(O(n \log n))$—superexponentially fewer than $n!$ for large $n$.

Term comparison for small $n$:

$n$	Leibniz formula terms	Bell number terms ($B_n$)
3	6	5
4	24	15
5	120	52

Each generic summand in the new expansion is a pure tensor: a product of $n$ linear forms, one from each row, thus aligning directly with rank-one tensor decompositions and bounding the tensor rank by $B_n$ (Houston et al., 2023).

2. Tensor, Partition, and Slice Ranks in Determinant Expansions

The determinant polynomial can be viewed as an order-$n$ tensor, which motivates the study of minimal-length expansions using products of linear or multilinear forms.

• Tensor Rank: The tensor rank (Trank) of the determinant tensor $T_{\det}$ is the minimal $r$ so that $T_{\det}$ is a sum of $r$ pure tensors. The partial-partition expansion gives $\operatorname{Trank}(T_{\det}) \leq B_n$ (Houston et al., 2023).
• Waring Rank: Via polarization, the Waring rank of $\det$ is bounded by $2^{n-1} B_n$ (Houston et al., 2023).
• Slice Rank: Lampert and Moshkovitz, in "Slice rank and partition rank of the determinant" (Lampert et al., 8 Sep 2025), established that the slice rank of $\det_n$ is exactly $n$, corresponding to the classical Laplace expansion. No nontrivial slice-rank expansion exists with fewer than $n$ summands; every minimal decomposition aligns, up to basis change, with Laplace.
• Partition Rank: Partition-rank decompositions are more flexible; they decompose the multilinear form as sums of products of multilinear forms over variable subsets. A logarithmic lower bound applies: $$\operatorname{prk}(\det_n) \geq \lceil \log_2 n \rceil + 1$$, and explicit quadratic (partition-rank 3) decompositions are known for $n=4$, with the determinant $\det_4$ coinciding with the 4th elementary symmetric polynomial over $\mathbb{F}_2$ (Lampert et al., 8 Sep 2025).

The significance is that deterministic tensors (structure) can asymptotically separate partition rank and analytic rank, with $$\operatorname{prk}(\det_n)/\operatorname{ark}(\det_n)$$ diverging as $n \to \infty$.

3. Rank-One Determinantal Families and Border Complexity

A canonical family is the set of polynomials expressible as $\det(A_0 + \sum_{i=1}^n A_i x_i)$ where each $A_i$ is rank one. Such rank-one determinantal families, denoted $\mathrm{Det}_1$ or $\mathrm{VBP}^1$, have a natural expansion via Cauchy–Binet: $$\det(U X V^T) = \sum_{S \subseteq [n], |S| = r} \det(U_S) \det(V_S) \prod_{i \in S} x_i,$$ where $U$ and $V$ collect the rank-one vectors and the $\det(U_S)$, $\det(V_S)$ are the Plücker coordinates (Chatterjee et al., 2023). The expansion is the Hadamard product of Plücker embeddings in the Grassmannian.

The border complexity problem asks whether the closure under approximation (limit of a sequence of such polynomials) stays within the same class. It is proven that $\mathrm{Det}_1$ is closed under approximation: any polynomial obtained as a limit of rank-one determinantal polynomials has a genuine rank-one determinantal representation (Chatterjee et al., 2023). The key technical tools are valuated matroids, Plücker coordinate closure, and Murota’s weight-splitting theorem, which facilitate separating limits in the Hadamard coordinates.

4. Rank-One Updates and the Matrix-Determinant Lemma

Matrix perturbations of the form $A + u v^T$—a rank-one correction—admit an explicit determinant update formula: $\det(A + u v^T) = \det A \cdot (1 + v^T A^{-1} u),$ assuming $A$ invertible (Filar et al., 2019). This is the matrix-determinant lemma and underpins efficient determinant and derivative computations in numerical and optimization contexts, especially in Markov chain analysis.

For optimization tasks (e.g., Hamiltonian cycle problem), the lemma allows the determinant, gradient, and Hessian with respect to $u$ or $v$ to be obtained in $O(N^3)$ time via a single LU decomposition—a drastic efficiency improvement over naive approaches (Filar et al., 2019). When $A$ is singular, refinements with Moore–Penrose inverses or basis changes restore validity, subject to the regularizability of $A + u v^T$.

5. Combinatorial and Path-Integral Expansions

Rank-one constructions naturally generalize in combinatorial settings, such as operators formed from polynomials of rank-one maps. Burman developed a discrete path-integral formula for the characteristic polynomial of operators $M = P(M_1, ..., M_N)$, where each $M_i$ is rank one. The expansion expresses coefficients as sums over discrete oriented one-dimensional manifolds with boundary (DOOMBs), encapsulating unions of cycles and chains on $\{1,\ldots,N\}$ (Burman, 2012): $$\chi_M(\lambda) = \sum_{r=0}^n (-1)^r \left[ \sum_{G \in \mathcal D_{n,r}} W_P(G) \det A_G \right] \lambda^{n-r}$$ where $W_P(G)$ is a combinatorial path weight and $A_G$ collects inner products along the graph's edges.

This framework generalizes classical results such as:

• The matrix-tree theorem (Kirchhoff), expressing minors of the Laplacian as sums over spanning forests.
• The Forman–Kenyon formula for cycle-rooted spanning forests, incorporating holonomy.
• Level-2 analogues, where determinants reflect counts of triangulated nodal surfaces with prescribed combinatorics.

6. Applications in Algebraic Complexity, Optimization, and Graph Theory

Rank-one determinant expansions have essential applications in:

• Algebraic Complexity Theory: Upper and lower bounding tensor and Waring ranks; showing closure under border complexity for restricted determinantal classes (Houston et al., 2023, Chatterjee et al., 2023).
• Symbolic Computation and Identity Testing: Efficient computations in the class $\mathrm{Det}_1$ with deterministic black-box root detection (Chatterjee et al., 2023).
• Combinatorial Optimization: Fast determinant and derivative evaluation for rank-one perturbations in Markov chain generators, relevant to combinatorial optimization algorithms (Filar et al., 2019).
• Graph Invariants and Laplacians: Connecting combinatorial enumeration to spectral graph invariants via discrete path-integration over rank-one mapped operators (Burman, 2012).

7. Contemporary Directions and Open Questions

Recent progress includes establishing nearly optimal bounds for determinant ranks, explicit polynomial-size expansions with combinatorial transparency, and nontrivial separations of structure vs. randomness in tensor ranks (Houston et al., 2023, Lampert et al., 8 Sep 2025). Open questions include:

• Determining precise partition ranks for the determinant function in higher $n$—the gap between known lower and upper bounds is substantial.
• Whether analogous concise expansions exist for permanent or immanant polynomials.
• Extensions of path-integral and DOOMB techniques to noncommutative, quantum, or higher-categorical settings.

A plausible implication is that further advances in explicit, low-rank expansions would yield new complexity theory lower bounds and improved algorithms for both algebraic computation and combinatorial optimization. The methodology and combinatorial insight of rank-one determinant expansions continue to shape research at the intersection of linear algebra, algebraic geometry, and theoretical computer science.