Preferred Basis from Relational Invariants

provisional

Overview

This derivation confronts a foundational puzzle in quantum mechanics: what determines which measurement outcomes are possible in a given experiment?

Quantum mechanics tells us how to compute probabilities via the Born rule, but the rule only works once you know the measurement basis — the set of possible outcomes. A spin can be measured along any axis, and each choice gives different probabilities. The theory itself does not say which axis is “the” measurement axis. This is the preferred basis problem, and it sits at the heart of the measurement problem.

The approach. The framework resolves this by tying the measurement basis to the specific physical interaction between observer and system:

• Every measurement interaction generates a conserved quantity — a relational invariant — that encodes what the observer learns about the system.
• This conserved quantity, being a self-adjoint operator, has a unique set of eigenstates (by the spectral theorem).
• The measurement outcomes are exactly these eigenstates: the states in which the conserved quantity takes a definite value.

The result. The measurement basis is not a property of the system alone but of the observer-system interaction. Different experimental setups generate different conserved quantities, hence different bases. Complementarity (the impossibility of simultaneously measuring position and momentum, for example) arises because different interactions can generate non-commuting conserved quantities. Environmental decoherence is recovered as the macroscopic limit where many interactions collectively select a unique classical basis.

Why this matters. The preferred basis problem is one of the core obstacles to a complete interpretation of quantum mechanics. This derivation resolves it structurally: the basis is determined by the physics of the interaction, not by an ad hoc choice or an appeal to consciousness. The framework also sharpens decoherence — standard decoherence describes approximate suppression of interference, while here the basis selection is exact.

An honest caveat. The structural correspondence between relational invariants and conserved quantities of the Hamiltonian is forced by the continuous-discrete duality — there is nothing else in the continuous layer for relational invariants to map to. What remains postulated is the explicit construction: the map from a specific interaction configuration to the specific Hamiltonian that governs it.

Note on status. This derivation is provisional because it contains preferred-basis S1 (interaction-invariant correspondence). If that postulate is promoted to a theorem, this derivation would be upgraded to rigorous.

Statement

Theorem. The preferred basis for measurement outcomes is the eigenbasis of the relational invariant generated in the Type III interaction between observer and system. This basis is the unique one in which the relational invariant takes definite values — the coherence-stable states. Different interaction configurations generate different relational invariants, hence different measurement bases. This resolves the basis problem structurally, with environmental decoherence recovered as the macroscopic limit.

Derivation

Structural Postulate

Structural Postulate S1 (Interaction–invariant correspondence). Every Type III interaction between an observer $\mathcal{O}$ and a system $S$ generates a relational invariant $I_{\mathcal{O}S}$ whose operator representation $\hat{I}_{\mathcal{O}S}$ on $\mathcal{H}_S$ is the conserved quantity associated with the interaction Hamiltonian $H_{\text{int}}$ via Noether’s theorem. Formally, $[\hat{I}_{\mathcal{O}S}, H_{\text{int}}] = 0$, and the interaction dynamics drives the joint state toward the eigenbasis of $\hat{I}_{\mathcal{O}S}$.

Remark (Tightened content). The Continuous-Discrete Duality (Proposition 3.2) forces the structural part of this correspondence: a relational invariant in the discrete layer is conserved (Axiom 1) and generated by an interaction. In the continuous layer’s Lagrangian dynamics (Coherence Lagrangian, Theorem 6.0), the only conserved quantities associated with interactions are Noether charges of the Hamiltonian. The compatibility condition between layers leaves no other candidate for what a relational invariant could correspond to — the structural correspondence is forced by the duality.

What remains postulated is the explicit construction: the map from a specific physical interaction configuration to the specific Hamiltonian $H_{\text{int}}$ that governs it. The Noether correspondence (invariant $\leftrightarrow$ symmetry of $H_{\text{int}}$) is rigorous once the Hamiltonian is identified (Relational Invariants, Theorem 3.2). The open content of S1 is therefore: “which Hamiltonian governs which interaction?” — a question about the specific dynamics, not about the structural correspondence.

Step 1: The Basis Problem

Definition 1.1 (Basis ambiguity). In a Hilbert space $\mathcal{H}_S$ of dimension $d$ (Born Rule, Theorem 7.1), any state $|\psi\rangle \in \mathcal{H}_S$ can be expanded in infinitely many orthonormal bases. The Born rule $P(k) = |\langle k | \psi \rangle|^2$ gives probabilities given a basis $\{|k\rangle\}$, but does not select the basis.

Proposition 1.2 (Logical priority). The basis selection problem is logically prior to the probability problem. Without specifying the measurement basis, the Born rule is incomplete — it gives probabilities for every possible set of outcomes, but does not determine which set is physically realized in a given measurement.

Step 2: Relational Invariants and Self-Adjoint Operators

Definition 2.1. Let $\mathcal{O}$ (the observer/measurer) and $S$ (the system) undergo a Type III interaction (Three Interaction Types). The interaction generates a relational invariant $I_{\mathcal{O}S}$ (Relational Invariants).

Proposition 2.2 (Operator structure). The relational invariant $I_{\mathcal{O}S}$, restricted to the system’s Hilbert space $\mathcal{H}_S$ for a fixed observer state $\sigma_\mathcal{O} \in \Sigma_\mathcal{O}$, defines a self-adjoint operator on $\mathcal{H}_S$:

$\hat{I}_{\mathcal{O}S} : \mathcal{H}_S \to \mathcal{H}_S$

Proof. The relational invariant $I_{\mathcal{O}S}: \Sigma_\mathcal{O} \times \Sigma_S \to \mathbb{R}$ is real-valued (it encodes a coherence content, which is real by Coherence Conservation, Definition 1.1). For fixed $\sigma_\mathcal{O}$, the map $I_{\mathcal{O}S}(\sigma_\mathcal{O}, \cdot): \Sigma_S \to \mathbb{R}$ is a real-valued function on the system’s state space.

In the Hilbert space formulation (Theorem 7.1 of Born Rule), real-valued observables correspond to self-adjoint (Hermitian) operators. The coherence content is preserved under unitary evolution (coherence conservation), so $\hat{I}_{\mathcal{O}S}$ commutes with the joint unitary dynamics — it is a constant of the motion. By the spectral theorem for self-adjoint operators on finite-dimensional Hilbert spaces, $\hat{I}_{\mathcal{O}S}$ has a complete orthonormal eigenbasis. $\square$

Step 3: The Eigenbasis Is the Measurement Basis

Theorem 3.1 (Basis selection). The measurement basis for the $\mathcal{O}$-$S$ interaction is the eigenbasis of $\hat{I}_{\mathcal{O}S}$:

$\hat{I}_{\mathcal{O}S} |k\rangle = \lambda_k |k\rangle, \quad k = 1, \ldots, d$

The states $\{|k\rangle\}$ are the possible measurement outcomes; the eigenvalues $\{\lambda_k\}$ are the measured values.

Proof. The relational invariant $I_{\mathcal{O}S}$ is a conserved quantity (by Coherence Conservation, Axiom 1): once generated, it maintains a definite value. The post-measurement state must therefore be one in which $\hat{I}_{\mathcal{O}S}$ takes a definite value — an eigenstate.

A superposition $|\psi\rangle = \sum_k \psi_k |k\rangle$ with $\lambda_j \neq \lambda_k$ for some $j, k$ does not have a definite value of $\hat{I}_{\mathcal{O}S}$:

$\hat{I}_{\mathcal{O}S}|\psi\rangle = \sum_k \lambda_k \psi_k |k\rangle \neq \lambda |\psi\rangle \text{ for any single } \lambda$

Therefore the post-measurement state must be an eigenstate $|k\rangle$, and the measurement outcomes are the eigenstates. $\square$

Step 4: Coherence Stability

Definition 4.1. A state $|\phi\rangle \in \mathcal{H}_S$ is coherence-stable with respect to $\hat{I}_{\mathcal{O}S}$ if:

$\hat{I}_{\mathcal{O}S}|\phi\rangle = \lambda |\phi\rangle \quad \text{for some } \lambda \in \mathbb{R}$

Proposition 4.2. The eigenbasis $\{|k\rangle\}$ is the unique basis of coherence-stable states.

Proof. By the spectral theorem, the eigenstates of a self-adjoint operator form a complete orthonormal basis, and they are the only states with definite eigenvalues. Any non-eigenstate has a variance $(\Delta I)^2 = \langle \hat{I}^2 \rangle - \langle \hat{I} \rangle^2 > 0$, meaning the relational invariant does not take a definite value. $\square$

Proposition 4.3 (Uniqueness of basis). For non-degenerate $\hat{I}_{\mathcal{O}S}$ (all $\lambda_k$ distinct), the measurement basis is unique up to phase.

Proof. A non-degenerate self-adjoint operator has a unique eigenbasis (up to individual phase factors $|k\rangle \mapsto e^{i\phi_k}|k\rangle$, which do not affect the Born probabilities $|\langle k|\psi\rangle|^2$). $\square$

Step 5: Different Interactions, Different Bases

Theorem 5.1 (Interaction-dependence of basis). The measurement basis is a property of the Type III interaction, not of the system alone. Different interactions generate different relational invariants, hence different eigenbases.

Proof. Let $\hat{I}^{(A)}_{\mathcal{O}S}$ and $\hat{I}^{(B)}_{\mathcal{O}S}$ be the relational invariants generated by two different Type III interactions between the same observer $\mathcal{O}$ and system $S$. If the interactions differ (different interaction Hamiltonians), then in general $\hat{I}^{(A)} \neq \hat{I}^{(B)}$, and their eigenbases $\{|a_k\rangle\}$ and $\{|b_k\rangle\}$ differ.

Example. A Stern-Gerlach apparatus oriented along $\hat{z}$ generates $\hat{I}^{(z)}_{\mathcal{O}S} \propto \hat{S}_z$, with eigenbasis $\{|\uparrow_z\rangle, |\downarrow_z\rangle\}$. The same apparatus rotated to $\hat{x}$ generates $\hat{I}^{(x)}_{\mathcal{O}S} \propto \hat{S}_x$, with eigenbasis $\{|\uparrow_x\rangle, |\downarrow_x\rangle\}$. The physical configuration of the observer determines which relational invariant is generated. $\square$

Step 6: Complementarity

Definition 6.1. Two observables $\hat{I}^{(A)}$ and $\hat{I}^{(B)}$ are complementary if they do not commute: $[\hat{I}^{(A)}, \hat{I}^{(B)}] \neq 0$.

Theorem 6.2 (Complementarity from relational structure). Complementary observables arise when different Type III interactions generate non-commuting relational invariants. No state can be simultaneously an eigenstate of both.

Proof. If $[\hat{I}^{(A)}, \hat{I}^{(B)}] \neq 0$, then by a standard result in linear algebra, there is no common eigenbasis. The system cannot simultaneously have definite values for both relational invariants.

This non-commutativity is structural: different Type III interactions probe different “directions” in the joint state space $\Sigma_\mathcal{O} \times \Sigma_S$. The relational invariants they generate act on different subspaces, and these subspaces are generically non-commuting. $\square$

Corollary 6.3 (Uncertainty from complementarity). For complementary observables $\hat{I}^{(A)}, \hat{I}^{(B)}$ with $[\hat{I}^{(A)}, \hat{I}^{(B)}] = i\hat{C}$:

$\Delta I^{(A)} \cdot \Delta I^{(B)} \geq \frac{1}{2}|\langle \hat{C} \rangle|$

This is the Robertson uncertainty relation — a consequence of non-commuting relational invariants.

Step 7: Recovery of Decoherence

Proposition 7.1 (Structural decoherence). The standard decoherence program (Zurek’s pointer basis) is recovered as the macroscopic limit of the relational-invariant mechanism.

Proof. When the observer $\mathcal{O}$ is a macroscopic apparatus with many internal degrees of freedom, the Type III interaction with $S$ generates a relational invariant $\hat{I}_{\mathcal{O}S}$ whose eigenbasis is determined by the apparatus’s macroscopic configuration (by S1). Simultaneously, $S$ interacts (via Type III) with environmental observers $\mathcal{E}_1, \mathcal{E}_2, \ldots$, generating relational invariants $\hat{I}_{\mathcal{E}_1 S}, \hat{I}_{\mathcal{E}_2 S}, \ldots$.

The effective measurement basis is the common eigenbasis of the set $\{\hat{I}_{\mathcal{O}S}, \hat{I}_{\mathcal{E}_1 S}, \hat{I}_{\mathcal{E}_2 S}, \ldots\}$. By Proposition 4.2, this is the set of states that are simultaneously coherence-stable under all interactions.

This recovers Zurek’s einselection criterion: the pointer states are those that commute with the system-environment interaction Hamiltonian $H_{SE}$. In the framework, $H_{SE}$ generates the environmental relational invariants $\hat{I}_{\mathcal{E}_j S}$, and the pointer basis is the joint eigenbasis. The macroscopic limit is obtained when the number of environmental interactions $|\{\mathcal{E}_j\}|$ is large: the joint eigenbasis converges to a unique basis (the intersection of all eigenspaces), which is the classically stable pointer basis.

The framework adds precision: the basis selection is exact (determined by the spectral decomposition of the relational invariant), whereas standard decoherence theory describes approximate suppression of off-diagonal density matrix elements. In the framework, the off-diagonal elements vanish exactly in the eigenbasis of $\hat{I}_{\mathcal{O}S}$ — decoherence is the statement that the environmental relational invariants $\hat{I}_{\mathcal{E}_j S}$ commute with $\hat{I}_{\mathcal{O}S}$ in the macroscopic limit. $\square$

Proposition 7.2 (Comparison with decoherence). The relationship between the two approaches:

Step 8: Degenerate Spectra

Proposition 8.1 (Degeneracy resolution). When $\hat{I}_{\mathcal{O}S}$ has degenerate eigenvalues ($\lambda_j = \lambda_k$ for $j \neq k$), the eigenbasis is not unique within the degenerate subspace. The degeneracy is resolved by higher-order relational invariants from the same interaction.

Proof. By S1, a Type III interaction between $\mathcal{O}$ and $S$ may generate multiple relational invariants $\hat{I}^{(1)}, \hat{I}^{(2)}, \ldots$ corresponding to different conserved quantities of $H_{\text{int}}$. Since these all commute with $H_{\text{int}}$, they commute with each other: $[\hat{I}^{(i)}, \hat{I}^{(j)}] = 0$ (conserved quantities of the same Hamiltonian commute if the symmetry group is abelian, or can be simultaneously diagonalized by passing to a Cartan subalgebra if non-abelian).

The measurement basis is the simultaneous eigenbasis of the maximal commuting set $\{\hat{I}^{(1)}, \hat{I}^{(2)}, \ldots\}$. This is a complete set of commuting observables (CSCO) in the standard sense. By the spectral theorem for commuting self-adjoint operators, a simultaneous eigenbasis exists and is unique (up to phase) when the CSCO labels all states distinctly. This resolves all degeneracies. $\square$

Consistency Model

Theorem 9.1. The Stern-Gerlach measurement of spin-1/2 provides a consistency model for all results of this derivation.

Verification. Take $\mathcal{H}_S = \mathbb{C}^2$ (spin-1/2), with a Stern-Gerlach apparatus oriented along $\hat{z}$.

• Self-adjoint operator (Proposition 2.2): The interaction Hamiltonian $H_{\text{int}} \propto \hat{S}_z \otimes \hat{p}_z$ generates the relational invariant $\hat{I}_{\mathcal{O}S} = \hat{S}_z$, which is self-adjoint with eigenvalues $\pm\hbar/2$. $\checkmark$
• Eigenbasis (Theorem 3.1): The measurement outcomes are $|\uparrow_z\rangle, |\downarrow_z\rangle$ — the eigenstates of $\hat{S}_z$. $\checkmark$
• Coherence-stability (Proposition 4.2): $|\uparrow_z\rangle$ is the unique state with definite $S_z = +\hbar/2$. $\checkmark$
• Different interactions (Theorem 5.1): Rotating the apparatus to $\hat{x}$ gives $\hat{I}^{(x)} = \hat{S}_x$ with different eigenbasis $|\uparrow_x\rangle, |\downarrow_x\rangle$. $\checkmark$
• Complementarity (Theorem 6.2): $[\hat{S}_z, \hat{S}_x] = i\hbar\hat{S}_y \neq 0$ — no simultaneous eigenstates. $\checkmark$
• Uncertainty (Corollary 6.3): $\Delta S_z \cdot \Delta S_x \geq \frac{\hbar}{2}|\langle S_y\rangle|$. $\checkmark$
• Decoherence (Proposition 7.1): In the macroscopic apparatus, environmental interactions (air molecules, photons) generate relational invariants commuting with $\hat{S}_z$, selecting the $\hat{z}$-basis as the pointer basis. $\checkmark$ $\square$

Rigor Assessment

Fully rigorous:

• Proposition 2.2: $\hat{I}_{\mathcal{O}S}$ is self-adjoint (real-valued + conserved → Hermitian operator; spectral theorem gives eigenbasis)
• Theorem 3.1: Eigenbasis is the measurement basis (definite value of conserved quantity → eigenstate)
• Propositions 4.2–4.3: Uniqueness of coherence-stable basis (standard spectral theory)
• Theorem 6.2, Corollary 6.3: Complementarity and uncertainty from non-commutativity (standard linear algebra and operator inequalities)
• Theorem 9.1: Consistency model verified on Stern-Gerlach measurement

Rigorous given axioms + S1:

• Theorem 5.1: Different interactions give different bases (different $H_{\text{int}}$ generates different $\hat{I}_{\mathcal{O}S}$ via S1)
• Proposition 7.1: Recovery of decoherence (joint eigenbasis of multiple relational invariants → Zurek pointer basis in macroscopic limit)
• Proposition 8.1: Degeneracy resolution via CSCO (simultaneous eigenbasis of commuting conserved quantities)

Structural postulate (tightened):

• S1 (Interaction–invariant correspondence): The structural correspondence (relational invariants map to conserved quantities) is forced by the continuous-discrete duality. The remaining postulated content is the explicit map from interaction configurations to specific Hamiltonians — the “which Hamiltonian?” question.

Open assumptions:

• Extension to continuous spectra (position, momentum) requires spectral measures — mathematically standard but not carried out explicitly.
• The convergence of the pointer basis in the macroscopic limit (Proposition 7.1) assumes that environmental relational invariants commute with $\hat{I}_{\mathcal{O}S}$ — rigorously true in the macroscopic limit but approximate for finite environments.

Assessment: The core result — the measurement basis is the eigenbasis of the relational invariant — is rigorous given S1 (interaction–invariant correspondence). The spectral theorem and coherence conservation do all the mathematical work. The recovery of decoherence and complementarity are structural consequences.

Open Gaps

1. Interaction Hamiltonian mapping: The explicit map $H_{\text{int}} \mapsto \hat{I}_{\mathcal{O}S}$ from the physical interaction to the relational invariant is needed for concrete predictions.
2. Contextuality: The Kochen-Specker theorem shows that quantum observables cannot all have simultaneous definite values. This should follow from the relational-invariant mechanism: each measurement context generates a specific $\hat{I}_{\mathcal{O}S}$, and values are context-dependent. Explicit formalization is needed.
3. Continuous observables: Extension to position, momentum, and other continuous-spectrum observables via spectral measures.
4. Weak measurements: For partial (weak) Type III interactions, the relational invariant is not fully generated, and the system is left in a superposition of eigenstates with small disturbance. This should connect to the weak measurement formalism Aharonov, Albert, Vaidman, 1988.