Replica-symmetry breaking, coefficients, and detectability in Parisi's analysis of microscopic random energy-momentum non-conservation

Giorgio Parisi's 1984 Physics Letters B note 'On energy-momentum conservation' introduced a replica-field-theory analysis of a hypothetical random violation of energy-momentum conservation at the microscopic level (distances much smaller than $10^{-13}$ cm). Its central prediction is that the macroscopic strength of the induced translational-invariance violations scales as $l^{d_c-4}$, where $l$ is the disorder correlation length and $d_c$ is the engineering dimension of the lowest-dimension operator coupling different replicas ($d_c=4$ in scalar $\phi^4$ theory, giving order-one effects; $d_c=6$ with fermions, giving $O(l^2)$ effects; $d_c=8$ in pure gauge theory, giving $O(l^4)$ suppression). The analysis left open three linked issues that together form the present problem: whether replica symmetry breaking invalidates or qualitatively changes the replica-symmetric scaling argument (in particular whether violations can be parametrically larger), what the actual coefficients and renormalization-group logarithmic corrections to the $l^{d_c-4}$ law are, and whether the $O(l^2)$ class of violations could be experimentally detected — through which observables or rare high-momentum-transfer processes, and how the framework maps onto experimental contexts such as candidate proton-decay events.

Background

In relativistic quantum field theory, conservation of energy and momentum is not an independent axiom: it is the consequence of translation invariance of the action, enforced as a Ward identity (an identity among correlation functions forced by a continuous symmetry). Any mechanism that makes the microscopic structure of space-time inhomogeneous at very short distances should therefore show up, in principle, as small violations of four-momentum conservation in particle reactions. In the late 1970s and early 1980s this became a concrete speculative question: Hawking, Page and Pope argued that virtual quantum-gravitational bubbles — 'space-time foam' — could scatter particles so as to mimic non-conservation of energy and momentum at the Planck scale, and Friedberg and Lee's 'discrete quantum mechanics', in which time is a dynamical discrete variable, likewise entails a tiny microscopic non-conservation of energy. Parisi had also proposed, in 'Randomness as a source of massless particles', that microscopic randomness could have macroscopic field-theoretic consequences.

Parisi's 1984 note asked a sharp version of the question: if a quantum field $\phi$ propagates in a fixed but spatially random background $B$ with correlation length $l\ll 10^{-13}$ cm — a concrete model of microscopic random energy-momentum non-conservation — how large are the induced violations at macroscopic distances? Two naive expectations conflict: violations should die out as $l\to 0$, because long-distance physics cannot resolve very-short-distance disorder; or they could grow, because the background can exchange momenta of order $1/l$ with the field. The observable is the disorder-averaged quantity $\Gamma(\omega,\Delta)$, constructed as follows: in a fixed background $B$ the two-point function $G(x,y\mid B)$ depends separately on the center coordinate $x$ and the relative separation $\Delta$; writing $y=x+\Delta$ and Fourier transforming in $x$ gives $\tilde G(\omega,\Delta\mid B)$, and the disorder average $\Gamma(\omega,\Delta)=\overline{|\tilde G(\omega,\Delta\mid B)|^{2}}$ (the overbar denoting the average over the probability distribution of $B$) vanishes identically for $\omega\neq 0$ if translation invariance is exact and otherwise measures residual center-coordinate dependence after averaging — that is, non-conservation of total momentum in correlators. To compute it, the paper uses the replica trick, the standard device of disordered-systems statistical physics: introduce $n$ identical copies $\phi_i$ of the field coupled to the same quenched background (quenched meaning fixed for each disorder realization and averaged only at the level of observables), average over $B$ — moments of the disorder distribution become local inter-replica couplings — and continue to $n\to 0$, assuming the permutation symmetry $P_n$ among replicas remains unbroken. Under this assumption the disorder-averaged two-point function equals the replica two-point function, and the variance-type combination $\Gamma^{(4)}$ that builds $\Gamma(\omega,\Delta)$ equals the connected four-point coefficient $G^{(4)}_{1}$ of the replica theory. Integrating out the short-distance background then yields a local effective action for the replicas organized by operator dimension, schematically $\exp(-l^{-2}S_{2}+S_{4}+l^{2}S_{6}+\ldots)$ in four space-time dimensions, where $S_{d}$ collects all local operators of engineering dimension $d$ (the classical mass dimension of the operator). The strength of the induced energy-momentum non-conservation is controlled by the lowest-dimension operator $S_{d_c}$ that couples different replicas and scales as $l^{d_c-4}$, up to logarithmic corrections. Enumerating the operators: in scalar $\phi^4$ theory the replica-singlet quartic $(\sum_i \phi_i^2)^2$ has $d_c=4$, so violations remain of order one as $l\to 0$; with fermions, four-fermion replica operators such as $(\sum_i \bar\psi_i \psi_i)^2$ have $d_c=6$, giving $O(l^2)$ violations — unless chiral and charge-conjugation symmetries forbid them (as for a purely gravitational background coupling), in which case the leading operator again has $d_c=8$; in pure gauge theory, gauge invariance acting independently on each replica pushes the lowest replica-coupling operator, for example $(\sum_i F^{(i)}_{\mu\nu}F^{(i)}_{\mu\nu})^2$, to $d_c=8$, giving $O(l^4)$ suppression.

The whole chain rests on the replica-symmetric saddle and on engineering-dimension power counting. Replica symmetry breaking (RSB) is the phenomenon, discovered by Parisi in spin-glass theory, that the permutation symmetry among replicas can be spontaneously broken in the $n\to 0$ limit; wherever it occurs, observables are typically non-self-averaging (sample-to-sample fluctuations of the quantity survive even as the system grows), so that variance-type quantities such as $\Gamma(\omega,\Delta)$ require the broken-symmetry saddle for their correct evaluation. Already in 1987 Sourlas showed that the long-distance correlation functions of the random-field model and of the Ising spin glass are not self-averaging even at high temperature — a warning that replica-symmetric reasoning about long-distance disorder fluctuations is not automatically reliable. Within Parisi's framework itself, however, whether RSB occurs for random-translational-violation backgrounds, whether it changes the $l^{d_c-4}$ scaling (possibly producing larger violations), what the coefficients and renormalization-group logarithmic corrections are, and whether the experimentally most interesting $O(l^2)$ effects could actually be detected, were all left open.

Problem Statement

Within the replica framework of Parisi (1984) for a random microscopic violation of energy-momentum conservation — a quantum field theory coupled to a quenched random background $B$ with correlation length $l\ll 10^{-13}$ cm, whose macroscopic effect is measured by the disorder-averaged translational-violation correlator $\Gamma(\omega,\Delta)$ and is predicted by the replica-symmetric analysis to scale as $l^{d_c-4}$, with $d_c$ the engineering dimension of the lowest-dimension operator coupling different replicas — determine, as one coherent objective, the actual macroscopic size and detectability of the induced violations of translational invariance. Specifically: (a) analyze whether replica symmetry breaking occurs in the replica effective action for this class of random backgrounds, whether the identification of the disorder-averaged translational-violation observable with replica connected correlators survives in the broken-symmetry phase, and whether the scaling of the violations is qualitatively changed — this last question including whether replica symmetry breaking can produce parametrically larger violations than the replica-symmetric estimates (order one in scalar $\phi^4$ theory; $O(l^2)$ with fermions when chiral and charge-conjugation symmetries allow dimension-six four-fermion replica operators; $O(l^4)$ in pure gauge theory); (b) compute the coefficients and the renormalization-group logarithmic corrections to the $l^{d_c-4}$ law for the scalar, fermionic, and gauge cases — including at least the leading closed-form coefficient of the $l^{d_c-4}$ term together with the leading logarithmic correction, expressed through the anomalous dimension of the lowest-dimension replica-coupling operator — including a definitive determination of the leading replica-coupling operator dimensions and their anomalous dimensions; and (c) establish whether the $O(l^2)$ class of violations could be experimentally detected in a specified experimental setting, identify the observables or rare high-momentum-transfer processes that would most sensitively probe such effects, and work out the mapping to at least one named experimental context. These three components jointly serve the single objective of converting the source's engineering-level replica estimate into a controlled, quantitative statement about the observable macroscopic consequences of microscopic random energy-momentum non-conservation; answers must engage this stated framework (quenched random short-distance breaking of translational invariance, disorder-averaged observables), not a different model of energy-momentum non-conservation. The three components (a), (b), and (c) form one linked objective; part (b) is conditional on part (a), and the coefficients and logarithmic corrections are to be computed for whatever scaling law part (a) establishes.

References

1. G. Parisi, On energy-momentum conservation, Physics Letters B 136 (1984) 392-394. DOI: 10.1016/0370-2693(84)92026-4, https://doi.org/10.1016/0370-2693(84)92026-4
2. N. Sourlas, Long-Distance Behaviour of Correlation Functions in Disordered Systems, Europhysics Letters 3 (1987) 1007-1011. DOI: 10.1209/0295-5075/3/9/008, https://doi.org/10.1209/0295-5075/3/9/008
3. K. Enqvist, D. V. Nanopoulos, Proton decay, Progress in Particle and Nuclear Physics 16 (1986) 1-33. DOI: 10.1016/0146-6410(86)90001-3, https://doi.org/10.1016/0146-6410(86)90001-3
4. R. Friedberg, T. D. Lee, Discrete quantum mechanics, Nuclear Physics B 225 (1983) 1-52. DOI: 10.1016/0550-3213(83)90011-1, https://doi.org/10.1016/0550-3213(83)90011-1
5. S. W. Hawking, D. N. Page, C. N. Pope, Quantum Gravitational Bubbles, Nuclear Physics B 170 (1980) 283-306. DOI: 10.1016/0550-3213(80)90151-0, https://doi.org/10.1016/0550-3213(80)90151-0
6. G. Parisi, Randomness as a source of massless particles, in: Unification of the Fundamental Particle Interactions (Erice 1980 proceedings), Plenum, New York, 1980, pp. 583-585, INSPIRE record 161077, https://inspirehep.net/literature/161077
7. T. Jacobson, Unitarity in discrete quantum mechanics, preprint UCSB-TH-11-1985 (1985, unpublished), INSPIRE record 219946, https://inspirehep.net/literature/219946
8. D. V. Nanopoulos, Beyond the standard model, in: Astronomy, Cosmology and Fundamental Physics (Bologna 1988 proceedings), pp. 355-382, INSPIRE record 272979, https://inspirehep.net/literature/272979