Osmotic-Velocity Scaling at Symmetry-Breaking Transitions in Multi-Asset Markets

1. Introduction

Spontaneous symmetry breaking (SSB) describes the situation in which the ground state of a system does not share the full symmetry of the underlying Hamiltonian. In quantum field theory the canonical example is the Higgs mechanism (Higgs, 1964; Goldstone et al., 1962); in condensed matter the same mathematics governs phase transitions including ferromagnetism and superconductivity (Anderson, 1958). In finance, the analogue is the alignment of asset returns along a dominant principal direction during certain market regimes: the cross-section correlations approach unity, and the system’s “ground state” (the dominant factor) breaks the rotational symmetry of the asset universe.

The condensate magnitude |Φ|² measures the degree of this alignment. In a symmetric phase (low correlation, |Φ|² → 0), assets behave nearly independently. In a fully broken phase (|Φ|² → 1), assets are maximally aligned: a single factor explains most variance.

The osmotic velocity u of Nelson’s (1966) decomposition is the time-antisymmetric component of the drift field; in the financial context it measures the information gradient encoded in the cross-sectional probability density. In equilibrium quantum mechanics (Valentini, 1991), u is conjugate to the position-density gradient u = (ℏ/2m) ∇ρ/ρ. In finance, aggregate u measures the strength of information flow that is not directly expressible as price drift—crowded positioning, anticipatory order flow, and density-gradient noise.

The relationship between aggregate u and the SSB condensate |Φ|² is theoretically ambiguous. Two readings are available:

The two readings are physically distinct: Reading A treats |Φ|² as a stabilising order parameter; Reading B treats it as a destabilising market-coordination indicator. Which is correct is an empirical question.

2. Specifications Tested

2.1 Aggregate single-series

Specification 1 (S1) aggregates osmotic velocity across the universe and regresses on the condensate magnitude:

where u_i(t) is the asset-i osmotic velocity at time t, computed as the difference between forward and backward EWM-smoothed drift estimators per Andreev & Howden (2026a) §2.1. The SSB condensate is the mean-field average of inverse-correlation matrix entries:

2.2 Per-asset panel with fixed effects

Specification 2 (S2) treats each asset as a panel unit, includes asset-level fixed effects, and clusters standard errors at the asset level:

This specification controls for asset-specific osmotic-velocity levels and tests whether the within-asset relationship between |u| and |Φ|² is consistent across the universe.

2.3 Alternative order parameter: normalised spectral gap

Specification 3 (S3) replaces |Φ|² with the normalised spectral gap of the correlation matrix:

where λ₁ ≥ λ₂ ≥ ... ≥ λ_N are the eigenvalues of the correlation matrix Σ(t). The spectral gap is large when one principal direction dominates and small when eigenvalues are nearly degenerate (symmetric phase). Substituting Δλ for |Φ|² tests robustness to the specific definition of the order parameter.

2.4 Stress-window-only restriction

Specification 4 (S4) restricts the regression to days within identified stress windows (rolling-21-day realised volatility above 80th percentile). This is the regime in which the condensate is most active and tests whether the ν > 0 relationship is driven by stress regimes (which would be consistent with Reading B’s crowded-trade mechanism) or whether it persists across all regimes.

3. Empirical Results

3.1 Sample and computation

Universe: SPY, QQQ, IWM, TLT, GLD. Sample: 2005-01-03 to 2024-12-31, daily closes; N adjusted-business-day observations = 4979 after computing 60-day rolling windows.

The osmotic velocity for asset i at day t is computed as

The condensate magnitude |Φ|²(t) is the mean of the inverse correlation matrix entries, computed on a 60-day rolling window and normalised so its sample range maps to [0, 1] across the full N.

3.2 Headline results

All four return positive slope coefficients in the range +0.50 to +0.61, with absolute t-statistics well above the Harvey–Liu–Zhu (2016) threshold of 3.0. The spread of point estimates across specifications is narrow relative to either the slope magnitude or the Harvey threshold, indicating the result is not specification-fragile.

4. Theoretical Reconciliation

4.1 The Goldstone-mode reading (Reading A)

The canonical treatment of Nelson’s osmotic velocity at a symmetry-breaking transition treats it as a Goldstone-mode order-parameter dispersion. In the symmetric phase, the order-parameter modes are gapless and their fluctuations dominate the density gradient; the osmotic velocity scale is large because the density ρ is highly non-uniform across modes. In the broken phase, the Goldstone mode is gapped (the longitudinal direction along the condensate is suppressed by the broken symmetry), and the osmotic velocity scale is small because the density is more uniform. Critical scaling theory predicts

with ν > 0 the standard critical exponent. This predicts a negative log-log slope between |u| and |Φ|². The empirics disconfirm this.

4.2 The crowded-trade reading (Reading B)

The alternative reading treats |Φ|² as a market-coordination signal rather than a thermodynamic order parameter. When directional consensus is high (|Φ|² large), the cross-sectional density ρ concentrates on a few preferred asset configurations — a crowded long/short pattern across the universe. The density-gradient norm ∇ρ/ρ rises with this concentration, and the osmotic velocity rises with it.

The empirics confirm this. The slope of approximately +0.5 to +0.6 is consistent with a moderate positive correlation between order-parameter strength and order-flow noise — not a critical scaling exponent in the thermodynamic sense, but an empirical regularity of financial-market microstructure under coordinated positioning.

4.3 The open question

The two readings differ in their interpretation of the financial “condensate.” In Reading A, |Φ|² is a stabilising thermodynamic order parameter for which large values indicate consensus and small values indicate volatility. In Reading B, |Φ|² is a destabilising market-coordination indicator for which large values indicate positioning fragility (crowded trades produce large density gradients and large osmotic flows).

5. Implications and Discussion

5.1 For the Bohmian-finance research programme

The empirical confirmation of Reading B suggests that financial markets do not behave like equilibrium quantum systems in their osmotic dynamics. This is consistent with the broader non-equilibrium framing of finance under the Valentini H-theorem (Andreev & Howden, 2026f): the market is ρ ≠ |ψ|², and the osmotic velocity is a measurable indicator of how far from equilibrium the system sits at a given moment.

5.2 For market microstructure

The result is also testable against existing literature on crowded trades (Khandani & Lo, 2007; Stein, 2009). In those studies, crowded positioning produces self-reinforcing liquidations during stress regimes. The +0.5–0.6 osmotic–SSB slope quantifies one measurable aspect of this dynamic: when factor alignment grows, the noise floor in cross-sectional positioning grows with it.

5.3 As a market-state indicator

Operationally, the consistent positive slope can be used to construct a regime-state indicator: when realised osmotic velocity outpaces the |Φ|²-implied scaling, the market is in an unusually noisy regime relative to its degree of coordination. We register this hypothesis but do not test it empirically in this paper.

6. Limitations and Future Work

(1) The five-asset universe is small. Extending to a wider universe (50 or 100 ETFs, or single-name equities) would refine the cluster-robust standard errors and test whether the result is universe-size-dependent.

(2) The 60-day rolling window is one of many reasonable choices. Robustness to window length should be reported in a follow-up; we have informally confirmed that 30-day and 90-day windows produce slope estimates within the same +0.5 range, but a formal grid analysis is deferred.

(3) The osmotic-velocity computation uses an EWM-smoothed forward/backward drift estimator; alternative non-parametric estimators (kernel-based; Lyons-signature-based) may yield different slope coefficients. We have not run the comparison.

(4) The theoretical reconciliation between Reading A and Reading B is the central open question. We invite contributions from quantitative-finance theorists familiar with both the condensed-matter Goldstone-mode treatment and financial-market microstructure.

7. Discussion

This paper reports an empirical regularity (consistent positive slope across four specifications, |t| > 12 in every case) and uses it to discriminate between two theoretical readings of an underlying physical framework. The result confirms that the osmotic velocity rises with the SSB condensate magnitude in financial markets — the opposite direction predicted by the canonical Goldstone-mode reading. The methodological finding is that the two readings are empirically distinguishable, and that the answer is non-trivial: financial markets do not obey the equilibrium-statistical-mechanics scaling prediction.

We deliberately leave the theoretical reconciliation open. The empirical finding is robust; the explanation is not. Honest disclosure of theoretical ambiguity is a feature of methods papers, and we follow that principle here.

References