---
title: "Nonlinear Tree-Based Numeration Systems: A Consolidated Synthesis"
slug: "nonlinear-tree-based-numeration-systems-a-consolidated-synthesis"
url: https://papers.qnfo.org/papers/nonlinear-tree-based-numeration-systems-a-consolidated-synthesis
authors: ["Rowan Brad Quni-Gudzinas"]
version: "2.0.0"
published: "2026-06-29"
updated: "2026-11-02"
license: "QNFO-ULA-2.1"
status: "published"
quality_label: "High-quality scholarly research"
quality_basis: "QNFO publisher classification for open-access, versioned research with a public revision history and reader correction channel; this is not a claim of independent peer review."
peer_review_status: "Not independently peer-reviewed"
quality_evidence_url: "https://qnfo.org/about"
quality_tags: ["High-quality scholarly research", "open-access scholarly work", "versioned research publication"]
---

# Nonlinear Tree-Based Numeration Systems: A Consolidated Synthesis

## Abstract

Positional notation — the ordinary decimal and binary systems that underlie all computation — is natively an ultrametric tree. The digit string is a path from root to leaf; the base is a branching factor; the distance between two numbers is the depth of their lowest common ancestor. This insight, formalized through the $b$-adic valuation $d(x,y)=b^{-v_b(x-y)}$, unifies five research threads — distance-dependent weighting (Rofa 2013), hereditary binary compression (Tarau 2013), rational base representation trees (Akiyama et al. 2018), automata-theoretic formalism (Charlier & Kreczman 2025, Lecomte & Rigo 2001), and historical/biological parallels (Chrisomalis 2010) — and opens extensions into tropical geometry, perfectoid spaces, quantum computing, and interface design. This document synthesizes the full investigation.

## 1. The Core Theorem: Positional Notation as an Ultrametric Tree

**Theorem (Native Ultrametric of Positional Notation).** `[established]` In any integer base $b \ge 2$, the distance between two integers measured by the highest power of $b$ dividing their difference is exactly the $b$-adic valuation:

$$d(x, y) = b^{-v_b(x-y)}$$

where $v_b(n)$ is the exponent of the highest power of $b$ dividing $n$. This metric satisfies the strong triangle inequality:

$$d(x, z) \le \max\{d(x, y), d(y, z)\}$$

**Proof (sketch).** Two numbers share $k$ trailing digits in base $b$ iff $b^k \mid (x-y)$ but $b^{k+1} \nmid (x-y)$, giving $d(x,y)=b^{-k}$. For any three integers, if $x$ shares $k_1$ trailing digits with $y$ and $y$ shares $k_2$ with $z$, then $x$ and $z$ share at least $\min(k_1, k_2)$ trailing digits. Hence $d(x,z) \le b^{-\min(k_1,k_2)} = \max(d(x,y), d(y,z))$. $\square$

**The Tree Structure:**
- Level 0 (rightmost digit) branches $b$ ways per digit $a_0 \in \{0,\ldots,b-1\}$
- Each integer corresponds to a unique root-to-leaf path
- Ultrametric distance = depth of lowest common ancestor = count of trailing matching digits

**The Inversion:** The Archimedean "number line" is a projection. The native geometry is the discrete ultrametric tree — a Cantor set, clopen balls, all triangles isosceles.

**Ostrowski's Theorem** `[established]` states that $\mathbb{Q}$ has exactly two types of completions: the Archimedean $\mathbb{R}$ and the $p$-adic $\mathbb{Q}_p$ for each prime $p$. Both are valid. Only one is canonized.

---

## 2. Five Research Threads — Deep Dive

### Thread 1: Distance-Dependent Weighting (Rofa, 2013)

**Core innovation:** A digit's value depends on its vertex-to-root distance in the tree, not on its string position. This activates the digit "0" — it is no longer a passive placeholder. "012" and "12" yield distinct numerical values.

**Generalized valuation:**
$$val(\text{path}) = f(\text{depth of node}, \text{digit value})$$

where $f$ need not factor as $digit \times base^{position}$.

**Implications:** Redundancy (multiple representations for the same integer), applications in error detection/correction, and a new class of computational arithmetic.

### Thread 2: Hereditary Binary Arithmetic (Tarau, 2013)

**Problem:** Hyper-large numbers (Knuth up-arrow scale) cannot even have their *length* written in flat notation.

**Solution:** Recursively compress binary strings via run-length encoding, then encode the run lengths in the same compressed format. This yields a **tree of trees** where each node compresses the binary expansion of its subtree.

**Data type:**
```
data Tree = Zero | One | Cons Tree Tree
```

**Result:** Numbers requiring $10^{100}$ decimal digits compress to a tree of depth ~20. Addition, multiplication, and exponentiation remain computable *directly on the compressed hereditary form* — complexity shifts from $O(\log N)$ to $O(\text{depth of hereditary tree})$, which can be exponentially smaller for hyper-large numbers.

### Thread 3: Rational Base Representation Trees (Akiyama et al., 2008/2018)

**Surprising fact:** In integer bases, not every rational has a finite expansion ($1/3 = 0.\overline{3}$). In rational base $p/q$ (coprime, $p>q>1$), **every integer has a unique finite expansion** — the opposite behavior.

**Conversion algorithm:**
$$N = \frac{p}{q} \cdot Q + r, \quad 0 \le r < p/q$$

which is equivalent to $Nq = pQ + rq$. Iterating produces the finite expansion.

**The fractal tree:** Unlike integer bases where every subtree is isomorphic, rational base trees vary by modular residue class. Some subtrees appear infinitely often; others finitely often. The language of representations is typically **not regular** — a fundamentally more complex object.

**Connection:** Rational bases are a special case of **beta-expansions** (Parry, 1960), linking to symbolic dynamics and subshifts of finite type.

### Thread 4: Automata-Theoretic Formalism (Lecomte & Rigo, 2001; Charlier & Kreczman, 2025)

**Abstract numeration system:** A regular language $L$ over $\Sigma$, ordered lexicographically, where the $n$-th word represents integer $n$.

**Key result (Charlier & Kreczman, 2025):** The set of valid representations $\mathcal{R}(U)$ for a weight sequence $U_n$ is **regular** iff $U_n$ satisfies a linear recurrence with integer coefficients — equivalently, the base is a **Pisot number** (algebraic integer $>1$ with all conjugates $<1$ in modulus).

**Why regularity matters:**
- Normalization computable by a finite transducer
- Addition recognizable by a finite automaton (digit-by-digit, 3-track)
- Properties decidable in linear time

**Buchi's theorem** `[established]`: The first-order theory of $\langle\mathbb{N}, +\rangle$ (Presburger arithmetic) is decidable *because* addition is automaton-recognizable. The open question: which nonlinear weight sequences preserve recognizability?

**Dumont-Thomas connection:** Substitutions $\sigma: \Sigma^* \to \Sigma^*$ generate infinite words; prefix trees form numeration systems in non-standard bases (the dominant eigenvalue of the substitution's incidence matrix).

### Thread 5: Historical Context & Biological Parallels

**Chrisomalis (2010)** catalogued 100+ numerical notation systems across 5,500 years. The transition to positional notation was convergent evolution — multiple civilizations independently built tree structures.

| System | Native Tree Feature | Suppressed By |
|--------|-------------------|---------------|
| Babylonian sexagesimal | Floating radix point = undetermined tree depth | Decimal point |
| Mayan calendar numerals | Explicit vertical stacking = visible tree | Linear timeline |
| Modern binary/decimal | bit/digit strings are tree paths | "Number line" projection |

**Three historical erasures:**
1. **ZFC set theory:** Numbers constructed from $\emptyset$ and $\in$ — no tree structure encoded
2. **Dedekind cuts:** $\mathbb{R}$ as connected continuum canonized; $\mathbb{Q}_p$ marginalized as "exotic"
3. **Digital computation:** Hardware enforces tree structure (bit-trees) but software presents linear abstraction

**Laws of Form (Spencer-Brown, 1969):** An alternative foundation where "drawing a distinction" is primitive. A positional numeral is nested marks — a tree. The base is grouping $b$ distinctions into a higher-order distinction. "10" is a **re-entry** of the form into its own boundary.

**DNA parallel:** The genetic code uses a 4-letter alphabet in codons (triplets) — structurally a 4-ary tree of depth 3 with 64 leaves mapping to ~20 amino acids. The code is degenerate and non-power positional — structurally analogous to a redundancy-allowing nonlinear tree-based numeration system.

---

## 3. Extensions & Future Directions

### 3.1 Theoretical Extensions

| Extension | Core Question | Status |
|-----------|--------------|--------|
| **Weight sequence classification** | For which $U_n$ is arithmetic polynomial-time? Regular? | Partial: Pisot = regular; Salem = open |
| **Tropical geometry** | Max-plus algebra is natively ultrametric; what are tropical numeration systems? | Unexplored |
| **Perfectoid spaces (Scholze, 2012)** | Can $\mathcal{O}_K$ where $K$ is perfectoid serve as a numeration ground ring? | Speculative |
| **Bruhat-Tits buildings** | Generalized trees for $\mathrm{SL}_n(\mathbb{Q}_p)$ — building-based numeration for algebraic groups? | Open |
| **Non-commutative trees** | Free group Cayley graphs as non-commutative numeration | Unexplored |

### 3.2 Computational Frontiers

**Silent-Radix Cryptography:** The fact that a digit string cannot internally specify its base is a cryptographic primitive. A secret large base $b$ makes numbers ambiguous without it — reducing to an integer knapsack problem.

**Ultrametric Quantum Computing** `[speculative]`:
- Thesis I: Tree topology confines errors geometrically — $p_L \propto (p/p_{\text{th}})^{d_{\text{tree}}}$
- Thesis II: 3D-integrated transmon qubits in coaxial cavities via TSVs — manufacturable
- Thesis III: Prime-frequency spectral engineering for frequency-domain multiplexing
- Disconfirmed if: depth-2 tree (8 transmons) shows no improvement over 2D layout

**p-Adic Neural Networks:** Replace real weights with $p$-adic valued weights. Non-Archimedean metric enables exact computation (no floating-point rounding).

**Tree-Native Programming Language:** First-class numeration bases, active zero, base-polymorphic arithmetic, ultrametric distance as primitive.

### 3.3 The p-Adic Curriculum

| Age | Content |
|-----|---------|
| 5-6 | The tree as complete representation of $\mathbb{Z}_2$ |
| 6-7 | Expansion to base $p$; "how far apart in the tree?" |
| 7-8 | $p$-adic metric formalized: $|x|_2 = 2^{-\text{depth}}$; clopen sets |
| 8-9 | Sequences and convergence: "climbing the tree together" |
| 9-10 | Differentiation: difference quotients, strict differentiability |
| 10-11 | Integration: Volkenborn integral, Mahler basis, Fundamental Theorem |
| 11+ | Introduce $\mathbb{R}$ as the "other" completion of $\mathbb{Q}$ |

**Key principle:** Build ultrametric intuition before Archimedean. The line becomes a specialized tool, not the default geometry.

### 3.4 Interface Design: The Archimedeanized Tree

All interfaces degrade an ultrametric catalog (file system, knowledge graph) into an Archimedean line (scrollbar, feed, search results). Design principle: the tree is ground truth; lateral edges are shortcuts; both views simultaneously available.

---

## 4. Production Infrastructure: Ultrametric Engine

The QNFO ultrametric engine (`ask-qwav.q08.workers.dev`, v2.8) implements the core discovery principles in production:

### 4.1 Deployed State

| Metric | Value |
|--------|-------|
| Papers indexed | 109 |
| Vectors in Vectorize | 170 (BGE-M3, 1024-dim) |
| AI model | @cf/deepseek-ai/deepseek-r1-distill-qwen-32b |
| Embedding model | @cf/baai/bge-m3 |
| Total Cloudflare Workers | 28 |

### 4.2 Working Endpoints

| Endpoint | Purpose | Mathematical Principle |
|----------|---------|----------------------|
| `/api/search?q=X` | Semantic search via vector embeddings | Cosine similarity in 1024-dim embedding space — the ultrametric tree is implicit in the nearest-neighbor structure of the vector index |
| `/api/papers` | List all papers | 1D linear listing — the Archimedean projection of the tree |
| `/api/ask?q=X` | AI-powered Q&A over corpus | Retrieval-augmented generation using embedding search |
| `/api/stats` | Paper metadata statistics | Quantitative audit of the corpus |
| `/api/health` | System health | Operational monitoring |

### 4.3 Theory-to-Engine Mapping

| Mathematical Concept | Engine Implementation |
|----------------------|----------------------|
| **Ultrametric distance** $d(x,y)=b^{-v_b(x-y)}$ | Cosine similarity on 1024-dim BGE-M3 embeddings — an ultrametric in the limit of single-linkage clustering |
| **Tree structure** of positional notation | Vectorize index partitions the embedding space into a hierarchical tree (nearest-neighbor graph) |
| **$p$-adic cache TTL** $= 15s \times 2^{\text{ord}_2}$ | Documented in skill but not exposed as an endpoint in v2.8 |
| **3-phase discovery** (word → cluster → tree) | v2.8 uses vectorize-only search; cluster expansion and tree-based search not deployed |
| **Spectral analysis** (Tate, Amice, Intrinsic Amice) | Not deployed in this instance |
| **Bruhat-Tits building** | Not deployed — theoretical only |

### 4.4 Architecture

```
Workers (28 total)
├── ask-qwav             — ultrametric discovery engine (v2.8)
├── ultrametric-tree-api — tree API (separate worker)
├── graph-api            — knowledge graph (graph-api.qnfo.org)
├── qnfo-data-api        — unified data API (v2 endpoints)
├── living-paper-api     — living paper system
├── cms-api              — content management
└── 22 other operational workers
R2: qnfo/ namespace (canonical storage)
D1: qnfo-audit (18 tables, FTS5)
Vectorize: BGE-M3 embeddings, 170 vectors, 1024-dim
Pages: deep.qwav.tech, papers.qnfo.org
```

---

## 5. The Gap Analysis

| Gap | Description | Impact |
|-----|-------------|--------|
| **External literature search** | arXiv and Semantic Scholar APIs unreachable from this environment | Literature discovery limited to internal QNFO corpus (109 papers) |
| **Advanced engine features** | Spectral analysis, Bruhat-Tits, Berkovich, dendrogram endpoints not deployed | Engine operates as basic semantic search, not full ultrametric tree explorer |
| **Engine-skill drift** | Skill documents 27+ endpoints; v2.8 has 11 | Skill documentation needs updating |
| **DNA connection** | The genetic code's tree structure is noted as an analogy but not mathematically formalized | Open research question |
| **Empirical validation** | Quantum computing thesis, p-adic curriculum, silent-radix crypto — all `[speculative]` or `[my conjecture]` | Requires experimental verification |

---

## 6. Research Roadmap

| Timeframe | Milestone | Certainty |
|-----------|----------|-----------|
| **1-5 years** | Pilot p-adic curriculum vs. control; depth-2 tree quantum chip benchmarked | Feasible |
| **5-10 years** | Full weight sequence classification; tree-native programming language prototype | Likely |
| **10-20 years** | Tree-based cryptographic standards; DNA-numeration isomorphism proven/refuted | Speculative |
| **20-50 years** | Bruhat-Tits numeration for $n$-dimensional groups; reflective numeration systems | Highly speculative |

---

## 7. Key References

1. **Rofa (Quni-Gudzinas, 2013).** *On Number Representation.* arXiv:2005.10207. `[EXTERNAL-SOURCE]`
2. **Tarau, P. (2013).** *Arithmetic Algorithms for Hereditarily Binary Natural Numbers.* arXiv:1310.8277. `[WEB-SEARCH]`
3. **Akiyama, S., Frougny, C., & Sakarovitch, J. (2008).** On subtrees of the representation tree in rational base numeration systems. *Discrete Mathematics & Theoretical Computer Science.* `[WEB-SEARCH]`
4. **Charlier, E. & Kreczman, S. (2025).** Positional numeration systems and regular languages. `[WEB-SEARCH]`
5. **Lecomte, P. & Rigo, M. (2001).** Abstract numeration systems. In *Words, Languages & Combinatorics III.* `[LLM-INFERRED]`
6. **Chrisomalis, S. (2010).** *Numerical Notation: A Comparative History.* Cambridge. `[WEB-SEARCH]`
7. **Spencer-Brown, G. (1969).** *Laws of Form.* Allen & Unwin. `[established]`
8. **Holly, J. (2001).** Pictures of Ultrametric Spaces, the $p$-adic Numbers, and Valued Fields. *Amer. Math. Monthly*, 108(8), 721-728. `[established]`
9. **QNFO (2026).** *The Ultrametric Foundation.* deep.qwav.tech/papers/ultrametric-foundation/. `[EXTERNAL-SOURCE]`
10. **Scholze, P. (2012).** Perfectoid Spaces. *Publ. Math. IHES*, 116, 245-313. `[established]`

---

*Synthesized from: five-thread deep-dive, extensions & future directions analysis, live Cloudflare infrastructure audit (28 Workers, ask-qwav v2.8, 109 papers), and the Ultrametric Foundation thesis.*
