Finite-Size Scaling in Critical Phenomena

• Finite-Size Scaling (FSS) is a framework that explains how singularities at phase transitions become smooth features in finite systems via scaling with system size.
• It uses a universal scaling ansatz and functions to extract critical exponents and quantify rounding, shifting, and finite-volume effects in both classical and quantum systems.
• FSS is essential for understanding diverse transitions—including percolation, nonequilibrium, and quantum criticality—with corrections from finite-entanglement and dangerous irrelevant variables.

Finite-Size Scaling (FSS) is the universal framework describing how singularities associated with phase transitions in infinite (thermodynamic-limit) systems are replaced by smooth, rapidly sharpened features in finite systems, parameterized by some characteristic linear size L (or Hilbert-space dimension N in quantum models). FSS quantifies the rounding, shifting, and emergence of scaling laws for observables as a function of system size, and thereby provides rigorous machinery for extracting critical exponents, scaling functions, and universal quantities in both classical and quantum many-body systems, as well as for exactly characterizing transitions in non-equilibrium, percolative, glassy, and disordered systems.

1. Fundamental Principles and General Finite-Size Scaling Ansatz

The foundational assumption of FSS is that proximity to a continuous phase transition endows the system with a diverging correlation length, $\xi\sim|t|^{-\nu}$, where $t$ is a relevant distance-to-criticality parameter (e.g., $t = (T-T_c)/T_c$) and $\nu$ the correlation-length exponent. Upon confining the system to finite linear size $L$, correlation growth is truncated at $\xi\sim L$, producing well-defined scaling windows:

• For $|t|\gg L^{-1/\nu}$, behavior is bulk-like.
• For $|t| \lesssim L^{-1/\nu}$, FSS governs the crossover between critical and finite-volume dominated behavior.

The generic scaling form for an observable $Q(t,L)$ is

$$Q(t,L) = L^{Y_Q} \, \tilde Q(x), \qquad x = t L^{1/\nu},$$

where $t$0 is the FSS exponent (fixed by bulk scaling: $t$1 with $t$2), and $t$3 is a universal scaling function (Li et al., 2024, 0710.1038, Campostrini et al., 2014).

For field-driven or multi-parameter transitions (e.g., symmetry-breaking field $t$4, temperature $t$5), the scaling fields $t$6 capture RG-determined analytic relations among the bare couplings, yielding multi-variable scaling forms. At quantum critical points, time (or temperature) and spatial dimensions rescale distinctively, producing anisotropic (quantum) FSS governed by the dynamic exponent $t$7 (Campostrini et al., 2014).

2. Scaling in Classical and Quantum Criticality: Exponents, Scaling Functions, and Corrections

At bulk (thermodynamic) continuous transitions, the scaling window width, the rounding of singularities, and the pseudocritical shift all scale as $t$8 (or volume $t$9). Universal exponents control:

• The amplitude scaling (e.g., for susceptibility $t = (T-T_c)/T_c$0 at criticality).
• The crossover scaling variable $t = (T-T_c)/T_c$1.

Nonlinear scaling fields give analytic corrections (e.g., $t = (T-T_c)/T_c$2, $t = (T-T_c)/T_c$3); irrelevant RG fields give universal non-analytic corrections (power-law $t = (T-T_c)/T_c$4, boundary corrections $t = (T-T_c)/T_c$5).

Quantum phase transitions introduce finite-size and finite-temperature scaling with the identification $t = (T-T_c)/T_c$6. The leading FSS forms are modified by scaling, e.g.

$t = (T-T_c)/T_c$7

and scaling corrections from both bulk and boundary irrelevant operators (Campostrini et al., 2014). For systems above the upper critical dimension ($t = (T-T_c)/T_c$8), dangerous irrelevant variables require further modification (see Section 5).

3. Application Paradigms: Percolation, Nonequilibrium, Discontinuous, and Nonstandard Regimes

3.1 Percolation

In percolation, the standard FSS form applies:

$t = (T-T_c)/T_c$9

with $\nu$0, $\nu$1 from percolation universality. For explosive percolation, the presence of strong sample-to-sample fluctuations and lack of self-averaging requires refined FSS protocols, including alignment around realization-dependent pseudocritical points, and the introduction of new exponents (e.g., $\nu$2) for describing maximal cluster growth events (Zhu et al., 2017, Li et al., 2024).

3.2 Nonequilibrium and Discontinuous Transitions

Discontinuous (first-order) nonequilibrium and absorbing-state transitions possess FSS with volume scaling:

$\nu$3

with shifted pseudotransition points scaling as $\nu$4 and interval widths as $\nu$5 (Oliveira et al., 2018). In absorbing phase transitions studied in random $\nu$6-SAT, FSS for the order-parameter (unsatisfied clause density) is

$\nu$7

with $\nu$8 (finite-size “correlation volume” exponent) fixed by percolation universality or directed percolation (Lee et al., 2010).

3.3 Finite-Entanglement and Hilbert-Space Truncation Scaling

In DMRG/MPS-based approaches to critical spin chains, FSS divides into two regimes: finite-size versus finite-entanglement scaling. Accurate extraction of CFT data, central charges, and scaling dimensions requires ensuring the system is in the FSS regime with matrix-dimension-dependent “entanglement correlation length” much larger than $\nu$9 (Pirvu et al., 2012). Quantum models with infinite-dimensional local Hilbert spaces (e.g. Quantum Rabi model) employ an FSS protocol in the controlled large-$L$0 limit (Hilbert space cutoff) instead of spatial linear size (Khalid et al., 2022).

4. Crossover and Non-Universal Regimes: Classification and Logarithmic Corrections

At and above the upper critical dimension ($L$1), FSS is modified:

• Logarithmic multiplicative corrections appear (e.g., Ising in $L$2: $L$3, $L$4, $L$5) (Li et al., 2024, Chu et al., 2016).
• Crossovers between different scaling windows are governed by rates with which the critical point is approached ($L$6, $L$7), leading to $L$8-dependent exponents in measurable observables (Li et al., 2024).

Percolation FSS is classified into three types based on the scaling of mean and variance of local pseudocritical points:

• Type I: fluctuations and mean scale as $L$9 (standard);
• Type II: mean $\xi\sim L$0, fluctuations slower (explosive type—broad observable mixing at $\xi\sim L$1);
• Type III: fluctuations $\xi\sim L$2, mean slower (high-$\xi\sim L$3, free boundaries—observable exponents at $\xi\sim L$4 differ from pseudocritical) (Li et al., 2024).

5. FSS Above the Upper Critical Dimension and Dangerous Irrelevant Variables

For $\xi\sim L$5, hyperscaling fails, and dangerous irrelevant variables (DIVs) require the introduction of a new exponent $\xi\sim L$6. The finite-size correlation length grows as $\xi\sim L$7, yielding modified scaling windows:

$\xi\sim L$8

and a generalized hyperscaling relation,

$\xi\sim L$9

valid for all $|t|\gg L^{-1/\nu}$0 (Kenna et al., 2024, Langheld et al., 2022, Wittmann et al., 2014). Two types of Fisher scaling relations pertain: standard on the correlation length scale, $|t|\gg L^{-1/\nu}$1; and $|t|\gg L^{-1/\nu}$2-FSS on the system scale, with effective anomalous dimension $|t|\gg L^{-1/\nu}$3.

Quantum and classical models in this regime (e.g. 5D Ising, long-range-transverse-field Ising chains) confirm the universality of these modified scaling forms across boundary conditions and geometries (Langheld et al., 2022).

6. Numerical and Algorithmic Implementation: Protocols and Applications

FSS analysis proceeds by:

• Measuring observable(s) for a grid of control parameters near expected criticality and a sequence of $|t|\gg L^{-1/\nu}$4;
• Identifying and scaling to the appropriate collapse variable $|t|\gg L^{-1/\nu}$5 or equivalent (e.g., $|t|\gg L^{-1/\nu}$6 for finite-element Hilbert-space scaling (Antillon et al., 2011));
• Plotting $|t|\gg L^{-1/\nu}$7 versus $|t|\gg L^{-1/\nu}$8 to identify scaling collapse;
• Estimating exponents and critical points via sequence extrapolation, data collapse, and finite-difference analysis (e.g., Hellmann–Feynman difference crossing constructions (Antillon et al., 2011), dynamic scaling for relaxational behavior (Choi et al., 2013), or crossing points of RG invariants (Campostrini et al., 2014));
• In percolation and disordered ensembles, rebinning events at realization-dependent local pseudocritical points to restore universal collapse when self-averaging fails (Zhu et al., 2017, Li et al., 2024).

For molecular simulations (e.g., critical transport), background subtraction and definition of effective finite-size critical points enable accurate extraction of critical exponents for dynamical observables (Midya et al., 2016). In applications to lattice QCD, FSS provides a diagnostic for critical endpoints and universality classes in the presence of strong corrections from finite volume, background, or irrelevance of certain coupling operators (Lacey, 2024, 0710.1038, Hasenfratz et al., 2013).

7. Horizons and Generalizations

The scope of FSS includes: (i) explicit extraction of critical parameters in ab initio quantum Hamiltonians via finite-element discretization (Antillon et al., 2011); (ii) determination of central charges and operator content in lattice models (Pirvu et al., 2012); (iii) scaling of entanglement entropies and identification of subleading universal corrections in conformal and nonconformal (including first-order) quantum/thermal transitions (Campostrini et al., 2014, Campostrini et al., 2014, Oliveira et al., 2018); (iv) modeling of nonequilibrium and absorbing-state transitions in complex optimization landscapes (Lee et al., 2010, Oliveira et al., 2018); (v) percolation, glassy, and discontinuous phenomena exhibiting crossover mixing or anomalous scaling regimes (Li et al., 2024, Cho et al., 2010, Zhu et al., 2017); and (vi) universality classification and testing of RG-based field theories via rigorous, quantitative finite-size–dependent exponents, scaling functions, and data-collapse protocols.

The continuous expansion of FSS frameworks to accommodate quantum devices, dynamically evolving networks, hybrid entanglement-statistics limits, and nonstandard universality classes constitutes a rapidly active front in critical phenomena research (Khalid et al., 2022, Midya et al., 2016, Chu et al., 2016).