Thermodynamic Geometry Through Second Order Phase Transitions

Omer M. Basri, Oren Raz · 2025

Read Thermodynamic Geometry Through Second Order Phase Transitions by Omer M. Basri, Oren Raz (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-01936.txt.

Introduction

Driving a physical system slowly and gently wastes the least energy when the path follows the geometry of its control settings. Omer Basri and Oren Raz study in 2025 what happens to that geometric picture near a phase transition, where a substance changes state, such as water turning to steam.

Away from transitions, a Riemannian metric on the control space measures wasted energy, and the best driving paths are its geodesics, the straightest routes. Near a second-order transition the metric blows up and those routes may stop existing. The authors work through this breakdown, which matters for designing efficient processes in systems that must cross a transition.

Catalog record

CREATOROmer M. Basri, Oren Raz
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-01936

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Thermodynamic Geometry Through Second Order Phase Transitions Preprint: APS/123-QED Omer M. Basri, Oren Raz Abstract A common approach to quantify excess dissipation in slowly driven thermodynamic processes is through the use of a Riemannian metric on the space of control parameters, where optimal driving protocols follow geodesics. Near phase transitions, this geometric picture breaks down as the metric diverges and geodesics may cease to exist. Using Widom scaling, we analyze this framework for several universality classes and show that in some cases the thermodynamic length across the phase transition remains finite. We then demonstrate a numerical approach for computing minimal paths in such systems. We show that, in some regimes, the shortest path crosses the phase transition – even when alternative paths confined to a single phase exist. Introduction Since the earliest developments of equilibrium statistical mechanics to modern nonequilibrium statistical mechanics there have been attempts to cast the theory in a geometric form 10 ; 25 ; 41 ; 9 ; 13 ; 19 ; 22 ; 37 ; 5 ; 26 ; 20 ; 43 . One of the most useful approaches is to regard the excess dissipation as a Riemannian metric. This idea was introduced already in the 80’s 29 ; 28 , following purely geometrical concepts suggested already in the 70’s 25 ; 41 . It was based on two assumptions: (i) endoreversibility—the system remains in internal equilibrium throughout the process, though not necessarily in equilibrium with the environment; (ii) the dissipation is only a function of the system’s mean response time. The latter assumption restricts the formalism’s applicability to systems with nearly constant relaxation time. Despite these limitations, the framework has proven valuable for identifying minimal-entropy-production protocols in processes such as chemical reactions and distillation 1 ; 2 . The two limiting assumptions were lifted by using a closely related metric, which only assumes linear response theory to be applicable 32 ; 44 ; 24 ; 31 ; 4 ; 23 . In this formulation, the system’s relaxation dynamics are incorporated into the metric, hence removing the dependence on a constant response-time and eliminating the endoreversible assumption. This generalized framework extends thermodynamic geometry to a much broader class of systems—including those with widely varying time scales and microscopic systems such as biomolecular or single-molecule systems—and has found applications in optimizing computational and physical processes 23 . A similar framework was suggested for quantum systems 30 . A particularly challenging class of systems exhibiting widely separated time scales are those undergoing phase transitions. The application of thermodynamic geometry to such systems has drawn significant attention 14 ; 6 , including efforts to identify optimal driving protocols across critical regions (e.g., the 2D ferromagnetic Ising model 24 ). In many cases of interest, e.g., antiferromagnets 38 , crossing the phase transition cannot be avoided as the two phases are separated by a phase transition surface. Yet, the divergence of thermodynamic length near phase transitions has received limited study. Notably, a divergence in the metric tensor does not necessarily imply a divergence in thermodynamic length. However, even if the length remains finite, other geometric quantities such as the curvature may diverge, signaling the critical behavior. In this paper, we analyze the behavior of the thermodynamic metric near second-order phase transitions. Using the Widom scaling and the dynamical scaling hypothesis, we show that in some class of models, the divergence of the metric at the phase transition does not imply divergence of dissipation when crossing the phase transition. Therefore, the thermodynamic distance between states in the different phases is finite. The class of models with this property includes several mean field models, as well as the Ising model in dimensions 3 and above. To demonstrate our findings, we apply the formalism to the mean field antiferromagnetic Ising model, where we determine the optimal protocols connecting two points. Surprisingly, we find that in some cases the optimal protocol between two points that are in the same phase nevertheless crosses to the other phase and back. Framework Here we briefly define the key metrics and scaling tools used in our study. The dissipation metric, first introduced in 32 , is defined as g_{\mu\nu}=\beta\int_{0}^{\infty}\langle\delta X_{\mu}(t)\,\delta X_{\nu}(0)\rangle\,dt=\beta\mathcal{T}_{\mu\nu}\frac{\partial^{2}\ln Z}{\partial\lambda^{\mu}\partial\lambda^{\nu}} (1) where \mathcal{T}_{\mu\nu} is the time integrated temporal response matrix, Z the partition function, and X_{\mu} the conjugate variable to \lambda^{\mu} – the natural variables of the free energy, e.g., \{X_{1},X_{2}\}=\{S,M\} for F(T,h) 1 1 1 We assume that the Hamiltonian of the system is linear with \lambda_{i} . . In the linear response regime, the dissipated availability which expresses the excess dissipation in performing the trajectory \bm{\lambda(t)} in parameter space, is given by \mathcal{A}=\int_{\bm{\lambda}(t)}g_{\mu\nu}\,\dot{\lambda}^{\mu}\dot{\lambda}^{\nu}\,dt. (2) It is related to the length associated with the thermodynamic metric, defined by \mathcal{L}=\int_{\bm{\lambda}(s)}\sqrt{g_{\mu\nu}\,\dot{\lambda}^{\mu}\dot{\lambda}^{\nu}}\,ds (3) through the inequality \mathcal{A}\geq\frac{\mathcal{L}^{2}}{T} (4) (here T is the protocol duration, and \lambda(s) parametrize the trajectory in parameter space). This bound is saturated for protocols with g_{\mu\nu}\,\dot{\lambda}^{\mu}\dot{\lambda}^{\nu}=const. at all time 29 . Near second order phase transitions, second derivatives of the free energy, as well as response times —and hence the metric defined in Eq.( 1 )— diverge. The asymptotic form of these divergences is determined by critical exponents, that are common to all models that belong to the same universality class. These critical exponents can be analyzed using Widom’s scaling 33 , which for Ising type models relates the free energy \varphi near the critical point (T_{c},H_{c}) to reduced variables, often chosen to be t=(T-T_{c})/T_{c} and h=(H-H_{c})/H_{c} . The Widom scaling then states that \varphi\propto t^{b}f\!\left(\frac{h}{t^{\Delta}}\right) (5) where f(x) is continuous and f(x)\sim x^{p} as x\to\infty 15 . The exponents b , \Delta , p , are enough to uniquely dictate the standard critical exponents through the hyperscaling relations, e.g., the heat capacity exponent \alpha defined through C_{H}(t)\sim|t|^{-\alpha} is given by \alpha=2-b . This approach captures divergences of the free energy second derivatives across phase transitions, but does not address relaxation times, which are critical for the dissipation metric in Eq.( 1 ). This can nevertheless be done using the dynamical scaling relations 11 , that relates the relaxation time \tau and correlation length \xi as \tau\propto\xi^{z} (6) and the free energy scales as \varphi\propto\xi^{-d} 15 ; 36 . Combining these yields \tau\propto\varphi^{-z/d}\propto t^{-bz/d}\,f^{-z/d}\!\left(\frac{h}{t^{\Delta}}\right) (7) linking dynamical and thermodynamic critical behavior. Note that since we use the correlation length in this analysis, it cannot be used for mean field models. With the Widom and dynamical scaling, it is thus possible to find the divergence of the metric near the phase transition of various equivalence classes. Diverging metric, however, does not necessarily imply a divergent thermodynamic length. Since it is always possible to saturate the equality in Eq.( 4 ), finite thermodynamic length guarantees the existence of a finite-dissipation protocol. In addition, \mathcal{L}\to\infty implies \mathcal{A}\to\infty for all finite time protocols. Consequently, the convergence of thermodynamic length determines whether finite-dissipation protocols exist under the formalism of thermodynamic geometry. If \dot{\mathcal{L}}(s)=\sqrt{g_{\mu\nu}\dot{\lambda}^{\mu}\!\left(s\right)\dot{\lambda}^{\nu}\!\left(s\right)}\propto s^{-a} near criticality (as s\rightarrow 0 ), the total length diverges only for a>1 . Note, however, that finite distance across a phase transition does not necessarily inform us about the metric itself, or the intrinsic geometry of the manifold. The metric might diverge for flat space 18 ; 35 , and the manifold might

Read the complete text