New · Paper 6

Research Repository · Boolean Function Space Theory

Boolean Function Space Theory

Gate-Generated Function Spaces, Structural Classification of the 65,536 Four-Variable Boolean Functions, and Karnaugh Map Geometry of Function Classes

Repository: 6BooleanFunctionSpaceTheory  ·  Author: Choi Jonghun (최종훈), Inha University · Computer Science  ·  Status: New — Initial Exploration (2026)

Abstract

Overview

This repository investigates the mathematical structure of function spaces generated by single primitive Boolean gates, and explores how these spaces can be recognized, classified, and visualized through Karnaugh maps.

Unlike conventional digital logic courses that focus on gate design and circuit minimization, this repository treats Boolean function spaces as mathematical objects with structural properties — analogous to how the Structure Recognition Research Program studies structure spaces in Karnaugh maps.

The four-variable Boolean function space contains exactly 216 = 65,536 distinct functions. The central question is: what is the geometry of this space when viewed through the lens of gate-generated function classes?

Core Research Questions (Q1–Q4)

Q1 — Universal Gate Demonstration

Demonstrate that NAND alone can generate all 65,536 four-variable Boolean functions. Demonstrate the same for NOR.

This is an application of Post's Functional Completeness Theorem. The research value is in the Karnaugh map visualization of the generation process.

Q2 — Gate-Generated Function Space Characterization ★ Primary Research Target

For each primitive gate, determine:

  • The size of the generated function space
  • The characterization of the generated class
  • The visual patterns these functions form on Karnaugh maps
GateGenerated SpaceKnown Class
XORAffine (linear) functionsWell-characterized
ANDSubset of monotone functionsPartially known
NANDEntire Boolean space (65,536)Post's theorem
NOREntire Boolean space (65,536)Post's theorem

The unexplored question: what do these function classes look like on Karnaugh maps? This is where this research makes its contribution.

Q3 — Constructibility Framework

Given a target function (e.g., 3-input AND), determine which single gate types can generate it and which cannot. Generalize this into a constructibility framework — equivalent to asking: is f contained in the clone generated by G?

Target: 3-input ANDConstructible?
NAND-generated spaceYes
NOR-generated spaceYes
AND-generated spaceYes
XOR-generated spaceNo (AND is not affine)

Q4 — Function Space Structure ★ Foundational Question

For any primitive gate G, define and characterize Space(G):

  • Closure properties (composition, negation, duality)
  • Visual characteristics on Karnaugh maps
  • Structural invariants shared by all functions in Space(G)
  • Position in Post's lattice of clones

Q1–Q3 are all special cases of Q4. This is the foundational question of the repository.

Q4 (Function Space Structure) ↓ Q1 (Universality: Space(NAND) = Space(NOR) = All functions) Q2 (Characterization: Space(XOR) = Affine, Space(AND) = Monotone subset, ...) Q3 (Constructibility: f ∈ Space(G)?)

The Structure of Boolean Function Spaces

Post's Functional Completeness Theorem

Emil Post's theorem (1941) characterizes exactly which sets of Boolean functions are functionally complete — capable of generating all Boolean functions through composition. A gate G generates all Boolean functions if and only if its one-element set {G} is not contained in any of Post's five maximal clones:

Maximal CloneDescriptionNAND ∈?XOR ∈?
T₀ (truth-preserving)f(0,...,0) = 0NoYes
T₁ (falsity-preserving)f(1,...,1) = 1NoYes
M (monotone)x ≤ y ⟹ f(x) ≤ f(y)NoNo
S (self-dual)f(¬x) = ¬f(x)NoYes
L (affine/linear)f = a₀ ⊕ a₁x₁ ⊕ ... ⊕ aₙxₙNoYes

NAND is not in any of these five maximal clones → NAND generates all functions. XOR is in T₀, T₁, S, and L → XOR generates only the affine (linear) functions.

Lattice of Clones

The set of all clones (sets of functions closed under composition and containing all projections) forms a lattice under set inclusion. The five maximal clones described above are the top elements below the full clone. Understanding where Space(G) sits in this lattice is a core research direction.

All Boolean functions (full clone) ↓ Maximal clones: T₀, T₁, M, S, L ↓ ... (infinite lattice below) ↓ Projections (minimal clone)

Karnaugh Map Geometry of Function Classes

A core hypothesis of this research: functions belonging to the same gate-generated space share recognizable geometric patterns on Karnaugh maps. This connects the algebraic structure of clones to the visual, spatial intuition developed in Papers 1–3.

Known Visual Invariants

Function ClassKarnaugh Map CharacteristicConnection to Previous Papers
Affine (XOR-generated) Checkerboard pattern in Gray Code arrangement; double-diagonal in standard binary arrangement Paper 1 (Repo 1): checkerboard invariance; Paper 2 (Repo 2): layer-based analysis
Monotone functions No "isolated 0 islands" surrounded by 1s; patterns monotone in coordinate ordering Open question: visual characterization unexplored
Self-dual functions Complement of pattern equals pattern under bitwise negation of inputs Open question: Karnaugh map geometry not systematically studied
NAND-generated (all functions) All 65,536 patterns — no restriction Paper 2 (Repo 2): classification of all 65,536 functions by symmetry

Research Direction

The unexplored territory: for each non-trivial gate G, what is the visual invariant shared by all functions in Space(G)? If such invariants exist, they would create a new classification system for the 65,536 four-variable Boolean functions — based not on syntactic form but on generative origin.

Insight: 65,536 Functions and Symmetry Classification

Key Insight from Research Notes

The four-variable Boolean function space contains exactly 216 = 65,536 distinct functions. In general, an n-variable Boolean function space contains 2(2n) functions.

However, these 65,536 functions are not all "structurally independent." Under axis-swap symmetry (F(A,B,C,D) = F(C,D,A,B)), functions form equivalence classes (orbits under the symmetry group).

  • Some functions are symmetric under AB ↔ CD axis swap
  • Some functions form pairs with their axis-swapped counterpart
  • Some functions form larger orbits under a full symmetry group

Research question: When the 65,536 four-variable Boolean functions are classified by axis-swap symmetry, how many distinct equivalence classes result? This is a problem in combinatorics and group theory, and represents a natural extension of Papers 1–3.

n (variables)Total functions
122 = 4
224 = 16
328 = 256
4216 = 65,536
5232 = 4,294,967,296

The classification problem — "how many distinct classes under symmetry G?" — is where the study of function spaces (this repository) intersects with the study of structural invariance (Papers 1–3).

Relationship to Papers 1–5

Paper 1 (KMap patterns) Paper 2 (Symmetric function classification) → Paper 6 (Function Space Theory) Paper 3 (Variable rearrangement invariance) ↓ Paper 4 (Structure Recognition Theory — theoretical framework) ↓ Paper 5 (Human-AI collaboration — methodological layer)

Follow-Up Questions (FQ-1 to FQ-20)

The following 20 questions were generated as natural extensions of the core research questions Q1–Q4. Detailed answers are being developed in FOLLOW_UP_ANSWERS.md.

A. Function Space Structure

B. Karnaugh Map Geometry

C. Constructibility and Complexity

D. Connection to Structure Recognition Theory

E. Human-AI Collaboration

F. Generalization

Potential Future Repositories

Analysis of FQ-1 to FQ-20 identifies three clusters that may warrant independent repositories:

  • Candidate 7: Post Lattice × Karnaugh Map Geometry Visualization (FQ-2, 6, 7) — high educational value
  • Candidate 8: Boolean Function Space Topology (FQ-4, 19, 20) — pure mathematics direction
  • Candidate 9: Structure Recognition Experiments / Computational Model (FQ-13, 14, 15, 16) — SRT + Human-AI intersection