Chow’s EL-Numbers: Computability and Matroid Invariants

• Chow’s EL-numbers are defined through the exp-minus-log system, generating closed-form numbers from 1 using arithmetic, exp, and log operations.
• They are inherently computable, as every EML expression yields a number in the computable reals, excluding classical non-computable numbers like Chaitin’s Ω.
• In matroid theory, EL-numbers emerge as Euler–Lefschetz invariants in Chow rings, linking algebraic combinatorics with permutation statistics and Hilbert series.

Chow's EL numbers, depending on context, refer to two distinct but influential mathematical constructs: the "EL-numbers" defined by Timothy Chow in the context of closed-form constants under exponentiation and logarithm (appearing via the Exp-minus-Log system), and the "EL-numbers" (Euler–Lefschetz numbers) as combinatorial and Hilbert theoretic invariants arising in the study of the Chow rings of matroids. Both lines are prominent in modern algebraic combinatorics and computable analysis, yet their technical settings, algebraic presentations, and mathematical purposes differ markedly.

1. Chow’s EL-Numbers via Exp-Minus-Log (EML) Systems

Chow’s EL-numbers in the context of closed-form constants are defined as the smallest field extension of $\mathbb{Q}$ within $\mathbb{C}$ that contains $1$ and is closed under the operations $+, -, \times, \div$, and the usual analytic functions $\exp$ and $\log$ taken as the principal branch. Formally, the field $\mathbb{E}$ of EL-numbers is given as the intersection of all subfields $F \subseteq \mathbb{C}$ satisfying $x \in F \implies \exp(x) \in F$ and $x \in F \setminus \{0\} \implies \log(x) \in F$ (principal branch). Odrzywołek introduced the "Exp-minus-Log" (EML) system to provide a minimal syntactic generation for this closure, reducing all constructions to the constant $\mathbb{C}$0 and a binary function $\mathbb{C}$1; formally, the set $\mathbb{C}$2 of EML-expressions is generated by $\mathbb{C}$3 and, if $\mathbb{C}$4, then $\mathbb{C}$5. Each expression $\mathbb{C}$6 is assigned a value $\mathbb{C}$7 recursively by $\mathbb{C}$8 and $\mathbb{C}$9. The system thus builds, in effect, all closed-form numbers obtainable from $1$0 by iterated application of $1$1 in the principal branch.

The critical equivalence EML = EL is established by exhibiting that every operation in the field $1$2 is replicable within the EML system and, conversely, that the EML syntactic constructions are encompassed by the field-theoretic closure (Carney, 2 May 2026).

2. Main Structural and Inexpressibility Theorems

The canonical result proved in the EML system (and hence for EL-numbers) is that every number generated is computable: for any closed EML expression $1$3, the complex value $1$4 is computable in the sense of computable analysis (Theorem 2.1). Specifically, every such real number $1$5 (the set of real EL-numbers) lies in the set of computable reals, $1$6. This leads to significant inexpressibility results: most notably, Chaitin's halting probability $1$7—a canonical non-computable real—is not an EL-number (Theorem 2.2). The argument exploits the left-computable enumerability but non-computability of $1$8 and the closure of EL-numbers under only computable constructions.

Formally, EL-numbers cannot represent any truly non-computable real, including Martin-Löf random reals, limits of Specker sequences, or the real encoding the halting set. These results place the EL-closure strictly inside the computable reals: $1$9 (Carney, 2 May 2026).

3. Proof Techniques and Complexity Considerations

The computability of EL-numbers is established by induction on expression structure. The base case ([1]) is rational and thus computable. For the inductive step, if $+, -, \times, \div$0, and $+, -, \times, \div$1 are computable by induction hypothesis, then $+, -, \times, \div$2 is computable provided one navigates possible logarithmic branch cuts. Essential in this argument is a lemma ensuring that within any subexpression, either the value is on the negative real axis or is bounded away by a rational $+, -, \times, \div$3, preventing non-computable discontinuities. There is, however, a noted non-uniformity: no single algorithm produces, for every expression, effective approximations to log-branch cuts; uniform computability within subsystems is an open question.

Beyond computability, it remains open to classify the bit-complexity of evaluating EML-expressions, with the possibility of quasi-linear time in the expression length being conjectural (Carney, 2 May 2026).

4. Comparison to Chow’s EL-Numbers in Chow Rings of Matroids

A distinct EL-number theory emerges in the combinatorial theory of Chow rings of matroids (Hameister et al., 2018). There, the term "EL-numbers" designates Euler–Lefschetz numbers or invariants associated to the graded structure of the Chow ring $+, -, \times, \div$4 of a matroid $+, -, \times, \div$5. This Chow ring is a graded algebra defined by generators $+, -, \times, \div$6 for each non-minimal flat $+, -, \times, \div$7 of $+, -, \times, \div$8 and two classes of relations: quadratic ($+, -, \times, \div$9 for incomparable $\exp$0) and linear (for each $\exp$1, $\exp$2). The structure of $\exp$3 encodes Poincaré duality and supports highly explicit combinatorial Hilbert series and invariants (such as Charney–Davis quantities) via permutation statistics.

Hilbert series of $\exp$4 and related invariants in uniform and vector-space matroids are described in terms of excedance and major index statistics on $\exp$5, leading to expressions involving $\exp$6-Eulerian polynomials and $\exp$7-secant/tangent numbers. The coefficient structure of these series gives a detailed “EL” enumeration of monomial bases in the Chow ring, and symmetry and unimodality properties reflect deep geometric and combinatorial facts.

5. Implications, Expressive Limitations, and Open Problems

Chow’s EL-numbers (in the field-theoretic sense) define the boundaries of closed-form computation with exponentials and logarithms over the complex numbers. The field is countable and has Lebesgue measure zero in $\exp$8. Its expressive limitations are profound: no non-computable real number can be constructed via finite, closed-form expressions in $\exp$9, $\log$0, arithmetic, and rational constants.

Open problems include seeking subclasses of the EML system where uniform computability of evaluation is possible, characterization of complexity for approximating EL-numbers, and the possibility of extending the closed-form paradigm by adjoining further special functions (e.g., Gamma, Riemann zeta, or elliptic integrals) and examining whether such augmentation can bridge to families of non-computable numbers or requires truly higher-order logical frameworks.

A plausible implication is that closed-form systems, even maximally extended with all elementary analytic operations, are fundamentally limited to computable analysis. Further progress would demand either non-recursive operations or logic that can describe non-computable constructions.

6. Connections and Historical Significance

Chow’s EL-numbers are situated at the intersection of transcendental number theory, computable analysis, differential algebra, and algebraic combinatorics. The distinction between computable and non-computable closed forms, explicitly realized through inexpressibility theorems for objects like Chaitin’s $\log$1, anchors modern perspectives on the mathematical limits of symbolic computation. In matroid theory, the EL-invariants in Chow rings encode deep connections between geometric combinatorics and permutation statistics, as evidenced in comprehensive treatments by Adiprasito, Huh, Katz, Shareshian, Wachs, and others.

Both themes exemplify the tension between constructive, algebraically closed worlds and the uncountable, algorithmically unreachable continuum of real numbers—a central motivator for ongoing research in the structure and complexity of mathematical constants (Carney, 2 May 2026, Hameister et al., 2018).