This paper provides a complete proof that the Born rule probability formula P = cos²θ is the unique correlation function satisfying natural structural constraints on a self-similar binary tree. The proof requires no quantum mechanical assumptions — only boundary conditions, monotonicity, a composition rule, and ultrametric structure. The key insight is that the composition constraint, when properly transformed, reveals itself to be the cosine addition identity in disguise.
Relation to Gleason's theorem. Gleason's theorem (1957) establishes that the Born rule is the unique probability measure on Hilbert spaces of dimension d ≥ 3. The restriction to d ≥ 3 is essential and well known: in dimension 2 (the qubit case), non-Born frame functions exist that satisfy all of Gleason's constraints. The uniqueness argument does not close because two-dimensional spaces lack enough orthogonal directions to generate the necessary interlocking constraints.
A binary tree — two branches at each node — maps directly to the d = 2 case where Gleason fails. Every branching event is a two-way split. The local structure at each node is two-dimensional.
Yet this proof succeeds where Gleason's does not. The ingredient that closes the gap is self-similar recursion. Where Gleason requires interlocking constraints from three or more orthogonal directions at a single depth, this proof generates an infinite tower of consistency conditions across depths. The same correlation function must satisfy the composition rule at every scale simultaneously. "Enough dimensions" is replaced by "enough scales." Depth does the constraining work that width cannot.
This framing makes the result independently significant: it is not merely a Born rule derivation for a niche framework, but a resolution of a known limitation in a foundational theorem. The Born rule emerges uniquely even in dimension 2, provided the structure is self-similar.
Definition A.1 (Binary branching tree). Let T be an infinite binary tree where each node branches into exactly two children. We index nodes by their depth d ≥ 0 from an arbitrarily chosen root.
Definition A.2 (Branching distance). For two leaves at depth n, the branching distance d(a,b) is twice the number of edges from either leaf to their lowest common ancestor. Equivalently, if the paths from root to a and b first diverge at depth k, then d(a,b) = 2(n - k). The maximum branching distance at depth n is 2n.
Definition A.3 (Normalized distance). For a tree of depth n, the normalized distance ratio is r = d/(2n), giving r ∈ [0,1].
Definition A.4 (Ultrametric space). A metric space (X, d) is ultrametric if it satisfies the strong triangle inequality: d(a,c) ≤ max(d(a,b), d(b,c)) for all a, b, c ∈ X. A self-similar tree with branching distance as defined in A.2 is the canonical model of an ultrametric space.
Definition A.5 (Correlation function). A correlation function g: [0,1] → [0,1] assigns to each normalized distance r a value g(r) representing the correlation (or transition probability) between branches at that separation.
We impose three constraints on g. A fourth (the composition rule) will be derived as a lemma:
Axiom A1 (Boundary conditions).
Axiom A2 (Monotonicity). For r₁ < r₂, we have g(r₁) ≥ g(r₂). Correlation does not increase with distance.
Lemma A3 (Composition). For r₁, r₂ ≥ 0 with r₁ + r₂ ≤ 1:
$g(r_1 + r_2) = \left(\sqrt{g(r_1)}\sqrt{g(r_2)} - \sqrt{1-g(r_1)}\sqrt{1-g(r_2)}\right)^2$
This formula is not assumed — it is derived from two requirements: geometric consistency and the nature of branching.
Axiom A4 (Ultrametric structure). The correlation function is defined on an ultrametric space: the distance function satisfies d(a,c) ≤ max(d(a,b), d(b,c)) for all points a, b, c.
Lemma A4.1 (Horizontal uniformity). In an ultrametric space, if d(x,y) < r then B(x,r) = B(y,r). All balls of the same radius are isometric. Therefore the arrangement of children at any node is metrically indistinguishable from any other node at the same depth.
Lemma A4.2 (Vertical self-similarity). Distance is determined by ancestor depth. The strong triangle inequality forces all triangles isosceles with the two long sides equal. Therefore the composition formula depends only on depth separations, never on branching number.
Consider three points on the tree at normalized distances 0, r₁, and r₁+r₂ from a reference point. Their pairwise correlations g(r₁), g(r₂), and g(r₁+r₂) must be realizable as squared overlaps of unit vectors — that is, there must exist unit vectors whose pairwise inner products squared reproduce these values. This is equivalent to requiring that the associated Gram matrix be positive semidefinite. This is not a quantum mechanical assumption; it is the geometric requirement that transition probabilities be mutually consistent.
The PSD requirement constrains g(r₁+r₂) to a closed interval determined by g(r₁) and g(r₂):
$\left(\sqrt{g(r_1)}\sqrt{g(r_2)} - \sqrt{1-g(r_1)}\sqrt{1-g(r_2)}\right)^2 \leq g(r_1+r_2) \leq \left(\sqrt{g(r_1)}\sqrt{g(r_2)} + \sqrt{1-g(r_1)}\sqrt{1-g(r_2)}\right)^2$
The lower bound corresponds to minimum correlation between the endpoints: the intermediate point removes as much coherence as possible. The upper bound corresponds to maximum preservation of correlation: the intermediate point reinforces coherence.
On a tree, the intermediate point is a branch — a lowest common ancestor. A branching event is a complete distinction: the path went left or right. This complete distinction removes correlation maximally; the branch point contributes nothing beyond the binary choice itself. There is no mechanism at a branch point to reinforce correlation, because the branch point carries no degrees of freedom beyond the distinction it records.
The PSD requirement constrains g(r₁+r₂) to a closed interval. Lemma A.3.2 establishes that only the lower bound is consistent with boundary conditions.
On a self-similar tree, this argument is reinforced: any hypothetical structure at a branch point that could shift g(r₁+r₂) above the lower bound would constitute additional branching — which would appear as explicit nodes at the next depth. At every resolution, the branch points are bare distinctions, and the lower bound is achieved.
Lemma. The upper bound composition rule and all convex combinations thereof are inconsistent with Axioms A1. Only the lower bound satisfies boundary conditions under self-similarity.
Proof.
The upper bound composition rule states:
$g(r_1 + r_2) = \left(\sqrt{g(r_1)}\sqrt{g(r_2)} + \sqrt{1-g(r_1)}\sqrt{1-g(r_2)}\right)^2$
Evaluate at r₁ = r₂ = r:
$g(2r) = \left(\sqrt{g(r)} \cdot \sqrt{g(r)} + \sqrt{1-g(r)} \cdot \sqrt{1-g(r)}\right)^2 = \left(g(r) + (1 - g(r))\right)^2 = 1$
This holds for any value of g(r). Therefore g(2r) = 1 for all r ∈ [0, 1/2].
Setting r = 1/2: g(1) = 1.
But A1 requires g(1) = 0. Contradiction. The upper bound is impossible.
For convex combinations λ·(lower) + (1-λ)·(upper) with 0 < λ < 1: the mixed composition at r = r₂ = 1/2 yields g(1) = λ·0 + (1-λ)·1 = 1-λ > 0, violating g(1) = 0.
Only λ = 1 (pure lower bound) satisfies the boundary conditions.
Alternatively, the h-substitution proof (more general, shows the functional equation structure):
Proof (h-substitution).
Define h(r) = 2g(r) - 1 and φ(r) such that h(r) = cos(φ(r)). Under the upper bound, the half-angle identities yield:
$g(r_1+r_2) = \cos^2\left(\frac{\varphi(r_1) - \varphi(r_2)}{2}\right)$
Comparing with g(r₁+r₂) = cos²(φ(r₁+r₂)/2), we obtain:
$\varphi(r_1 + r_2) = |\varphi(r_1) - \varphi(r_2)|$
This is anti-additive. Setting r₁ = r₂ = r: φ(2r) = 0 for all r.
But A1 requires g(1) = 0, hence cos²(φ(1)/2) = 0, hence φ(1) = π.
Since φ(1) = φ(2·½) = 0 under the upper bound, we have 0 = π. Contradiction.
Theorem A.1 (Uniqueness of the Born rule).
$g(r) = \cos^2\left(\frac{\pi r}{2}\right)$
is the unique solution.
The proof proceeds in five steps.
Step 1: The h-substitution.
Define h: [0,1] → [-1,1] by
$h(r) = 2g(r) - 1$
This maps the unit interval [0,1] to [-1,1]. Since h takes values in [-1,1], we can write
$h(r) = \cos(\varphi(r))$
for some function φ: [0,1] → [0,π]. At this stage φ is defined pointwise: for each r, cos is surjective onto [-1,1] from [0,π], so φ(r) exists, and monotonicity of g (A2) forces φ to be non-decreasing, which makes the choice unique. Continuity of φ is not assumed here — it will follow from Step 3, where additivity and monotonicity together force linearity. The boundary conditions give φ(0) = 0 and φ(1) = π.
Step 2: Rewriting the composition formula.
We use the half-angle identities. Since g(r) = (1 + h(r))/2 = (1 + cos φ(r))/2:
$\sqrt{g(r)} = \sqrt{\frac{1 + \cos\varphi(r)}{2}} = \cos\left(\frac{\varphi(r)}{2}\right)$
$\sqrt{1 - g(r)} = \sqrt{\frac{1 - \cos\varphi(r)}{2}} = \sin\left(\frac{\varphi(r)}{2}\right)$
Substituting into Lemma A3:
$g(r_1 + r_2) = \left(\cos\frac{\varphi_1}{2}\cos\frac{\varphi_2}{2} - \sin\frac{\varphi_1}{2}\sin\frac{\varphi_2}{2}\right)^2$
where φ₁ = φ(r₁) and φ₂ = φ(r₂).
By the cosine addition formula, the expression in parentheses is cos((φ₁ + φ₂)/2). Therefore:
$g(r_1 + r_2) = \cos^2\left(\frac{\varphi_1 + \varphi_2}{2}\right)$
But we also have g(r₁ + r₂) = cos²(φ(r₁ + r₂)/2) by definition. Comparing:
$\cos^2\left(\frac{\varphi(r_1 + r_2)}{2}\right) = \cos^2\left(\frac{\varphi(r_1) + \varphi(r_2)}{2}\right)$
Since cos²(A) = cos²(B) admits A = B or A = π − B, we have two cases: φ(r₁+r₂) = φ(r₁) + φ(r₂) (additive), or φ(r₁+r₂) = 2π − φ(r₁) − φ(r₂) (anti-additive). Monotonicity (A2) eliminates the second. Set r₁ = r₂ = 1/4 under the anti-additive case: φ(1/2) = 2π − 2φ(1/4). Then r₁ = r₂ = 1/2: φ(1) = 2π − 2φ(1/2) = 2π − 2(2π − 2φ(1/4)) = −2π + 4φ(1/4). Since φ(1) = π, this gives φ(1/4) = 3π/4 and φ(1/2) = π/2. But monotonicity requires φ(1/4) ≤ φ(1/2), i.e. 3π/4 ≤ π/2. Contradiction. Therefore:
$\varphi(r_1 + r_2) = \varphi(r_1) + \varphi(r_2)$
Step 3: Additivity forces linearity.
The function θ(r) = φ(r)/2 maps [0,1] → [0, π/2], is non-decreasing (from A2), and satisfies θ(r₁ + r₂) = θ(r₁) + θ(r₂). Additivity gives θ(m/n) = m · θ(1/n) for positive integers m ≤ n. The boundary condition θ(1) = π/2 gives θ(1/n) = π/(2n). Therefore θ(m/n) = mπ/(2n) for all rationals in [0,1] — linearity on a dense subset. A monotone function agreeing with a continuous function on a dense set is that function (it has at most countably many discontinuities, and agreement on a dense set eliminates all of them). Therefore θ(r) = πr/2 on all of [0,1]. Continuity is a corollary, not an assumption.
Step 4: Determining the constant.
From Axiom A1, g(1) = 0. Therefore:
$\cos^2\left(\frac{\varphi(1)}{2}\right) = 0$
This requires φ(1)/2 = π/2, hence φ(1) = π.
Since φ(r) = 2θ(r) = πr, we have φ(1) = π. ✓
Step 5: Conclusion.
We have established that φ(r) = πr. Therefore:
$g(r) = \cos^2\left(\frac{\varphi(r)}{2}\right) = \cos^2\left(\frac{\pi r}{2}\right)$
This is the unique function satisfying Axioms A1, A2, and A4. ∎
The theorem predicts specific values at dyadic rationals r = k/2ⁿ, which can be verified by direct computation or numerical constraint propagation.
Corollary A.2. For dyadic rationals r = k/2ⁿ with 0 ≤ k ≤ 2ⁿ:
$g\left(\frac{k}{2^n}\right) = \cos^2\left(\frac{k\pi}{2^{n+1}}\right)$
| r | g(r) = cos²(πr/2) | Decimal |
|---|---|---|
| 0 | cos²(0) | 1.000 |
| 1/8 | cos²(π/16) | 0.962 |
| 1/4 | cos²(π/8) | 0.854 |
| 3/8 | cos²(3π/16) | 0.691 |
| 1/2 | cos²(π/4) | 0.500 |
| 5/8 | cos²(5π/16) | 0.309 |
| 3/4 | cos²(3π/8) | 0.146 |
| 7/8 | cos²(7π/16) | 0.038 |
| 1 | cos²(π/2) | 0.000 |
These values were independently discovered through numerical constraint propagation on finite-depth trees before the analytical proof was found.
The self-similarity of the tree is essential to the uniqueness result. On a self-similar tree, every node can serve as a local root, and the same correlation function must satisfy the constraints at every depth simultaneously.
In the language of ultrametric spaces: every point lies on the boundary. Precisely: in an ultrametric space, if d(x,y) < r then B(x,r) = B(y,r) — every point in a ball can serve as its center. Balls at any fixed radius partition the space, and all balls are both open and closed (clopen). Each ball at any scale is isometric to every other ball at that scale, so the local constraints are identical everywhere. There is no "interior" where constraints might relax. The boundary conditions propagate everywhere because everywhere is a boundary.
This ultrametric ball isometry is the geometric content of Lemma A4.1 — horizontal uniformity is not an additional constraint imposed on the tree but a consequence of what ultrametric structure already guarantees. Similarly, Lemma A4.2 derives vertical self-similarity: distance is depth-determined, the strong triangle inequality forces isosceles triangles, and the composition formula depends only on depth separations. Both properties follow from A4 rather than being assumed alongside it.
This is the geometric content of Lemma A3: it encodes how correlations compose when paths are concatenated, and it must hold at every scale. The cosine addition identity is the unique rule satisfying this self-consistency.
The proof reveals that the Born rule's squared-cosine form is not a free parameter of quantum mechanics but a forced consequence of:
The composition formula (Lemma A3) appears complex, but the h-substitution reveals it to be the cosine addition identity:
$\cos(\alpha + \beta) = \cos\alpha\cos\beta - \sin\alpha\sin\beta$
written in terms of squared amplitudes. The "interference" structure of quantum mechanics — the fact that amplitudes add before squaring — is precisely this trigonometric identity. The doubling formula a(2r) = 2a(r)² − 1 used throughout the proof is T₂(a(r)), the second Chebyshev polynomial of the first kind; the p-ary generalization (Section 10) is T_p(cos θ) = cos(pθ), the defining identity of Chebyshev polynomials.
No Hilbert space structure was assumed. No representation theory was invoked. The constraints are purely combinatorial properties of ultrametric spaces. The Born rule emerges from the geometry of branching.
The saturation of Lemma A3 — the fact that the composition achieves the lower bound of the PSD interval rather than any value above it — is settled by Lemma 3.2 on purely mathematical grounds: the upper bound and all convex combinations are self-contradictory under the boundary conditions. No physics is needed to exclude them. But the result has a striking physical interpretation. Any value of g(r₁+r₂) strictly above the lower bound would require the branch point to carry degrees of freedom beyond the binary distinction it records — precisely a hidden variable. Lemma 3.2 shows this is impossible by functional-equation analysis; Bell's theorem (1964) shows the same thing is impossible experimentally. The two arguments converge on the same exclusion from opposite directions: one from the structure of trees, the other from the statistics of entangled particles.
Bell's theorem therefore becomes a consequence rather than a premise. The correlation cos²θ is the unique function consistent with self-similar tree structure. Physical systems that violate Bell inequalities do so because they exhibit tree geometry — not because of nonlocality. Hidden variables would require the upper bound composition, which Lemma 3.2 proves is mathematically impossible. Bell violations are the signature of dendrogramic structure.
The theorem can be verified computationally by exhaustive search over discretized function spaces. At each tree depth n, one discretizes the interior values g(2/2n), g(4/2n), ..., g((2n-2)/2n) and checks which candidates satisfy all constraints within numerical tolerance.
At depth 1 (2 leaves, distances {0, 2}): Only g = (1, 0) survives — trivially satisfied.
At depth 2 (4 leaves, distances {0, 2, 4}): The composition constraint at r₁ = r₂ = 1/2 forces g(1/2) = 0.5.
At depth 3 and beyond: The constraints propagate to fix all dyadic rational values, converging to cos²(πr/2).
Python implementations of this constraint propagation are available in the supplementary materials.
The theorem holds not just for binary trees (p=2) but for all p-ary self-similar trees, where p is any prime. This section provides a rigorous proof.
Theorem A.3 (Prime-independence of the Born rule). Let T be a self-similar p-ary tree for any prime p ≥ 2. The unique correlation function satisfying Axioms A1, A2, and A4 is
$g(r) = \cos^2\left(\frac{\pi r}{2}\right)$
independent of p.
At each branch point, p child directions are represented by unit vectors u₁, ..., uₚ forming a regular (p-1)-simplex:
The Gram matrix Gₚ = (1-αₚ)Iₚ + αₚJₚ is positive semidefinite with rank p-1.
Consider three points: root R, point B at distance r₁, point C at distance r₁+r₂.
Amplitude decomposition:
The perpendicular components w_B and w_C live in the span of child directions.
Inner product constraint: $\langle v_B, v_C \rangle = \sqrt{g_1} \sqrt{g_{sum}} + \sqrt{1-g_1} \sqrt{1-g_{sum}} \cdot \omega$
where ω = ⟨w_B, w_C⟩. The constraint |⟨v_B, v_C⟩|² = g₂ determines the relationship.
Lemma. For any p ≥ 2, the extremal values ω = ±1 are achievable.
Proof.
In both cases, the extremes ω = ±1 are achievable. ∎
The constraint equation (√g₁ √g_sum + √(1-g₁) √(1-g_sum) ω)² = g₂ has solutions parameterized by ω ∈ [-1, +1].
Lower bound (ω = -1, maximally distinguishing):
Solving with ω = -1 yields:
$g_{sum}^{min} = \left(\sqrt{g_1}\sqrt{g_2} - \sqrt{1-g_1}\sqrt{1-g_2}\right)^2$
Upper bound (ω = +1, aligned):
Solving with ω = +1 yields:
$g_{sum}^{max} = \left(\sqrt{g_1}\sqrt{g_2} + \sqrt{1-g_1}\sqrt{1-g_2}\right)^2$
These are exactly the bounds in Lemma A3.
p does affect:
p does NOT affect:
The Born rule is a topological consequence of self-similar hierarchical structure, not a metrical consequence of specific branching geometry.
The p-adic formalism in DHT uses ℤₚ for various primes. This theorem shows that all choices of p yield the same correlation function. The Born rule is universal across all prime-adic completions.
The mathematical scaffolding connects to the classical theory of harmonic analysis on trees. The Chebyshev recursion a(pr) = T_p(a(r)) that drives the proof is the tree-geometric form of the Godement equation for spherical functions on Gelfand pairs — the non-abelian generalization of d'Alembert's cosine functional equation f(x+y) + f(x−y) = 2f(x)f(y). The classification of spherical functions on regular trees (Cartier 1972, Figà-Talamanca & Nebbia 1991) constrains positive-definite radial functions to a one-parameter complementary series; what pins the solution to a single function is the Chebyshev composition semigroup property T_{mn} = T_m ∘ T_n, which fixes the eigenvalue to the unique value compatible with both boundary conditions simultaneously across all depths.
slots:
- slug: two-balls-objection-closures
context:
- Linking to the proof appendix
slots:
- slug: cabibbo-case-distillation
context:
- "Sibling \u2014 the formal proof appendix that receives axiom-set editing notes"