Revisiting the Mazur bound and the Suzuki equality
Introduction
Time dependent equilibrium auto-correlation functions of physical observables play an important role in understanding dynamical properties of a system. In particular they tell us about the ergodicity of a given Hamiltonian system. In a seminal paper [1], Mazur discussed the long time average of such auto-correlation functions in the context of ergodicity. A rigorous lower bound was obtained for this quantity. In the classical description, we consider systems with phase space degrees of freedom and is some function of the phase space variables and describes some physical observable. The systems dynamics is described by a Hamiltonian and let us assume that apart from there are other conserved quantities, not necessarily independent ones. We denote the set of conserved quantities as with where denotes a thermal average over the Gibbs distribution with . Without loss of generality, one can assume that for all . We define the correlation matrix, with elements and consider the quantitywhich we will refer to as the Mazur bound. Let us also define the long time average of the temporal auto-correlation function of the observable :where the average is again over the equilibrium Gibbs distribution. Mazur proved that
In [1] it was shown that this result could provide insights on the ergodicity of the variable or its absence. We briefly discuss the notion of ergodicity as indicated in the behavior of the correlation function. Let denote the microcanonical average of the observable. Then ergodicity implies and in the thermodynamic limit this leads to the equality,where . Allowing for the presence of extra conservation laws, the notion of sub-ergodicity was discussed in [2], which in modern terms relates to the idea of generalized Gibbs ensembles.
The Mazur bound can be proven for both classical and quantum systems as already noted in the original paper. For the case of quantum systems, an exact Mazur-type equality was derived by Suzuki [3], which hold when one includes a “sufficient” number of constants of motion. For example, in a quantum system with a Hilbert space of finite dimension a trivial complete set of constants of motion are the energy projection operators where with denotes the energy eigenstates. Then (see later) it is easy to show that one obtains the equality in the Mazur relation in Eq. (3). Two natural questions that one could ask are: (i) instead of the projection operators, is it possible to obtain the equality with a smaller number of “local” constants of motion and (ii) does this result have a classical analogue. One of the aims of the present work is to discuss these questions and provide illustrative examples that clarify some subtle issues.
An important application of the Mazur relation has been in the context of transport properties of integrable systems [4], [6], [7], [8], [9], [10]. The auto-correlation functions involving currents corresponding to conserved quantities are related to transport coefficients via the Green-Kubo formulas. In particular, the asymptotic long time saturation value of the auto-correlation, gives the so-called Drude weight which is the strength of the zero-frequency component of the conductivity and implies ballistic transport. It was pointed out in [4] that the Mazur bound can be used to prove the presence of a finite Drude weight for integrable systems. This was used to prove ballistic transport in one-dimensional systems such as the quantum spin- XXZ chain [5] and the Toda lattice [6]. For classical integrable systems with degrees of freedom, the set of exactly independent constants of motion, which we denote by is in some sense special and in fact their existence defines integrability (e.g one can construct action-angle variables). One might expect that this set should be sufficient to saturate the Mazur bound. However, the numerical study in [11] found that one needs to include bilinear combinations of the conserved quantities, of the form in order to approach the equality. The study was restricted to systems of sizes and an important question is whether the contribution of the bilinear terms vanishes in the thermodynamic limit. On the other hand, the situation is even more complicated in the quantum case since the notion of quantum integrability is not so well defined and a basic question is on the choice of the set of constants required to saturate the Mazur bound. These aspects will also be discussed in this paper.
The plan of the paper is as follows: In Section 2 we outline the proofs of the Mazur inequality and the Suzuki equality. In Section 3 we describe a procedure which leads to the classical analogue of the Suzuki equality. As illustrative examples, we then provide in Section 4 explicit results, both numerical and analytical, on the application of the Mazur-Suzuki results in different physical systems. These include few body systems such as an oscillator and a two coupled spin system, as well as many body systems such as systems described by quadratic Hamiltonians, and finally the Toda chain. We conclude with a discussion in Section 5.
Section snippets
Proof of the Mazur and Suzuki relations
The Mazur bound: We start the discussion for a classical system. Consider an observable whose time evolution is given by and let denotes an average over the canonical distribution . Using stationarity and time reversal invariance one can show, via the Wiener-Khinchine theorem, that the correlation can be expressed in terms of the power spectral density of the signal. Thus one has
Classical analogue of Suzuki equality
Consider a classical Hamiltonian system with positional and momentum degrees of freedom, and having independent conserved quantities . Then the infinite time averagewhere is the initial condition, is by definition a conserved quantity. Let us also define the average of in a “generalized” microcanonical ensemble aswhere are the constrained values of the constants of motion.
Examples
In this section we discuss a number of examples to illustrate and clarify the Mazur-Suzuki bounds and their applications.
Discussion
We examined the relation between the Mazur inequality and the Suzuki equality. In particular we asked as to when, for classical systems, the Mazur inequality become an equality. In that case the time averaged autocorrelation, of an observable would be exactly equal to the Mazur bound, . A crucial point is the choice of conserved quantities to be included while constructing the . In general, a classical system with coordinate degrees of freedom will have a small number of independent
Credit Author Statment
All authors contributed equally to the project.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
We thank Sriram Shastry and Peter Young for very useful comments and suggestions. A.D. acknowledges support of the Department of Atomic Energy, Government of India, under project no.12-R& D-TFR-5.10-1100. K.S. was supported by Grants-in-Aid for Scientific Research (JP16H02211, JP19H05603, JP19H05791).
References (21)
Non-ergodicity of phase functions in certain systems
Physica
(1969)Ergodicity, constants of motion, and bounds for susceptibilities
Physica
(1971)Note on ergodic functions
Physica
(1971)- et al.
Transport and conservation laws
Phys Rev B
(1997) Open XXZ spin chain: nonequilibrium steady state and a strict bound on ballistic transport
Phys Rev Lett
(2011)Ballistic transport in classical and quantum integrable systems
J Low Temp Phys
(2002)- et al.
Conservation laws, integrability, and transport in one-dimensional quantum systems
Phys Rev B
(2011) - et al.
Thermodyamic bounds on Drude weights in terms of almost-conserved quantities
Commun Math Phys
(2013) - et al.
Microscopic origin of ideal conductivity in integrable quantum models
PRL
(2017) - et al.
Drude weight for the Lieb-Liniger Bose gas
SciPost Physics
(2017)