Phenomenological Renormalization Group (PRG)

• Phenomenological Renormalization Group is a method that defines RG steps by matching finite-size observables or empirical data rather than relying on a detailed microscopic Hamiltonian.
• It applies techniques such as Nightingale-style finite-strip correlation matching in lattice models and covariance-based pairing in neuronal systems to extract critical exponents.
• The approach validates scaling by cross-checking multiple observables like mass gaps, correlation lengths, and non-Gaussian distributions to robustly diagnose criticality.

Searching arXiv for recent and foundational papers on phenomenological renormalization group. Phenomenological renormalization group (PRG) denotes a class of renormalization procedures in which the RG step is defined operationally from finite-size observables or empirical correlation structure rather than from a fully specified microscopic Hamiltonian. In the lattice-model setting, PRG refers to Nightingale-style matching of finite-strip correlation lengths or mass gaps, possibly combined with cluster decimation, to construct recursion relations in coupling space and extract critical data (Borisenko et al., 2013). In the neural-data setting, PRG is a model-agnostic, data-driven coarse-graining of activity variables based on correlation-preserving pair merges in direct space or low-variance mode elimination in “momentum” space, with the central question being whether the resulting distributions flow toward a non-Gaussian critical fixed point (Nicoletti et al., 2020, Nascimento et al., 16 Jun 2025).

1. Conceptual basis

The RG logic underlying PRG is the standard one: a block transformation at scale factor $b$ maps microscopic variables $\sigma$ into coarse variables $\sigma'=\mathcal{R}[\sigma;b]$, and correspondingly maps couplings $K$ into renormalized couplings $K'=R_b(K)$. Critical points satisfy the fixed-point condition $K^*=R_b(K^*)$, and the linearization around $K^*$ determines relevant and irrelevant directions. PRG retains this fixed-point viewpoint while replacing an exact microscopic derivation of $R_b$ with phenomenological matching conditions imposed on observables that can be computed from finite systems or directly from data (Nascimento et al., 16 Jun 2025).

In the finite-size formulation, the relevant observables are strip correlation lengths, second-moment lengths, transfer-matrix eigenvalues, and free-energy-related quantities. In the neuronal formulation, the observables are covariance matrices, cluster variances, silence probabilities, covariance spectra, autocorrelations, and the full distribution of PCA-based coarse variables. The common principle is that scale invariance is inferred from the flow of observables under repeated coarse-graining rather than from explicit knowledge of all microscopic interactions. PRG is therefore phenomenological in the precise sense that it defines RG steps by matching finite-resolution descriptions of the same system (Borisenko et al., 2013).

A related point is methodological rather than definitional: PRG provides an alternative to explicit avalanche constructions. In the avalanche paradigm one measures $P(S)\sim S^{-\tau}f(S/S_c)$ and $P(T)\sim T^{-\alpha}g(T/T_c)$, with hyperscaling relations such as $\sigma$0; PRG instead probes scale invariance by asking how observables transform under coarse-graining, without necessarily constructing avalanches at all (Nascimento et al., 16 Jun 2025).

2. Finite-size matching in lattice and gauge models

Borisenko et al. formulate a modified phenomenological RG by combining Nightingale’s PRG with the cluster-decimation approximation (CDA). The starting point is a general $\sigma$1 spin model on a hypercubic lattice $\sigma$2, with partition function

$\sigma$3

where

$\sigma$4

The two-point correlator in representation $\sigma$5 is

$\sigma$6

Embedding the model on an $\sigma$7 strip with $\sigma$8 and transfer matrix $\sigma$9, one denotes by $\sigma'=\mathcal{R}[\sigma;b]$0 the largest eigenvalue and by $\sigma'=\mathcal{R}[\sigma;b]$1 the leading eigenvalue in sector $\sigma'=\mathcal{R}[\sigma;b]$2, so that

$\sigma'=\mathcal{R}[\sigma;b]$3

$\sigma'=\mathcal{R}[\sigma;b]$4

and the mass gap is

$\sigma'=\mathcal{R}[\sigma;b]$5

These strip quantities furnish exact input for the CDA-based RG step (Borisenko et al., 2013).

The decimation step relates the original lattice $\sigma'=\mathcal{R}[\sigma;b]$6 to a lattice $\sigma'=\mathcal{R}[\sigma;b]$7 of spacing $\sigma'=\mathcal{R}[\sigma;b]$8 through new couplings $\sigma'=\mathcal{R}[\sigma;b]$9:

$K$0

$K$1

Within CDA, the prefactors are taken from exact strip data, in particular

$K$2

with analogous expressions for $K$3. The phenomenological matching condition is the preservation of the mass gap, up to the scale factor $K$4:

$K$5

for each representation $K$6. An equivalent formulation preserves the second-moment correlation length $K$7 through

$K$8

The resulting nonlinear system defines the renormalized couplings and hence an RG flow in coupling-constant space (Borisenko et al., 2013).

Fixed points $K$9 satisfy

$K'=R_b(K)$0

Linearization gives a Jacobian $K'=R_b(K)$1, whose leading eigenvalue $K'=R_b(K)$2 yields the correlation-length exponent

$K'=R_b(K)$3

Because the free-energy prefactor $K'=R_b(K)$4 is retained along the flow, the method also permits computation of the bulk free energy and specific heat (Borisenko et al., 2013).

3. Data-driven coarse-graining in neuronal systems

In neuronal applications, PRG is defined directly on activity variables. The raw variables are denoted $K'=R_b(K)$5, $K'=R_b(K)$6, with equal-time covariance matrix

$K'=R_b(K)$7

and normalized correlation matrix

$K'=R_b(K)$8

The direct-space construction iteratively forms disjoint pairs by selecting the maximal off-diagonal correlation $K'=R_b(K)$9, removing that pair, and repeating until $K^*=R_b(K^*)$0 pairs have been formed. For each selected pair, the coarse variable is defined by simple summation,

$K^*=R_b(K^*)$1

After $K^*=R_b(K^*)$2 steps, each $K^*=R_b(K^*)$3 is the sum of $K^*=R_b(K^*)$4 original variables, and the covariance is recomputed after each blocking step. Because all $K^*=R_b(K^*)$5, coarse blocks can vanish, which makes the silence probability a natural observable (Nicoletti et al., 2020).

The complementary momentum-space construction starts from the eigendecomposition of the original covariance matrix,

$K^*=R_b(K^*)$6

One forms the rank-$K^*=R_b(K^*)$7 projector

$K^*=R_b(K^*)$8

and defines coarse variables

$K^*=R_b(K^*)$9

where $K^*$0 enforces $K^*$1. As $K^*$2 decreases, low-variance high-rank modes are discarded, mimicking a momentum-space RG cutoff. The same PCA-based construction is used in the later neuronal study, there written in terms of binary variables $K^*$3 and coarse variables $K^*$4 with a normalization constant $K^*$5 chosen so that $K^*$6 (Nicoletti et al., 2020, Nascimento et al., 16 Jun 2025).

A basic interpretive premise of this formulation is that no assumption of Gaussianity is built into the procedure. For weakly correlated variables, a Gaussian, central-limit fixed point is expected. Persistent non-Gaussian tails under repeated coarse-graining are taken as evidence for a nontrivial interacting fixed point rather than a trivial Gaussian one (Nicoletti et al., 2020).

4. Scaling observables and fixed-point diagnostics

The neuronal PRG literature emphasizes several scaling relations expected at a nontrivial critical fixed point. The first is variance scaling:

$K^*$7

The second is the silence probability,

$K^*$8

The third is covariance-spectrum scaling: if at criticality in $K^*$9 dimensions $R_b$0, then the ordered covariance eigenvalues obey

$R_b$1

The fourth is dynamical scaling of the mean autocorrelation

$R_b$2

from which a characteristic time $R_b$3 is extracted, with

$R_b$4

These observables operationalize the fixed-point question in direct space and dynamics (Nicoletti et al., 2020).

The momentum-space literature adds a distributional diagnostic. One monitors the full distribution of the PCA coarse variables and, in particular, the kurtosis

$R_b$5

At a true non-Gaussian fixed point, $R_b$6 remains distinct from a Gaussian under many coarse-graining steps, so $R_b$7 remains distinct from the Gaussian value $R_b$8. This criterion is presented as especially sensitive in neuronal models: momentum-space PRG asks whether non-Gaussian structure survives removal of a large fraction of modes, whereas a trivial or weakly correlated regime flows rapidly toward Gaussianity (Nascimento et al., 16 Jun 2025).

Taken together, these diagnostics motivate a multi-observable reading of PRG outputs. Variance growth, silence scaling, eigenvalue spectra, autocorrelation times, and the flow of coarse-grained distributions probe different aspects of the same candidate fixed point. A central practical claim in the literature is that no single observable suffices on its own (Nicoletti et al., 2020).

5. Benchmark systems and quantitative behavior

The systematic benchmark in the contact process considers $R_b$9 binary sites $P(S)\sim S^{-\tau}f(S/S_c)$0 on either a $P(S)\sim S^{-\tau}f(S/S_c)$1D lattice with $P(S)\sim S^{-\tau}f(S/S_c)$2 or a small-world network. Each occupied site empties at rate $P(S)\sim S^{-\tau}f(S/S_c)$3, and each empty site with $P(S)\sim S^{-\tau}f(S/S_c)$4 occupied neighbors becomes occupied at rate $P(S)\sim S^{-\tau}f(S/S_c)$5. The model has an absorbing state with all $P(S)\sim S^{-\tau}f(S/S_c)$6, order parameter $P(S)\sim S^{-\tau}f(S/S_c)$7, and control parameter $P(S)\sim S^{-\tau}f(S/S_c)$8, with transitions at $P(S)\sim S^{-\tau}f(S/S_c)$9 in $P(T)\sim T^{-\alpha}g(T/T_c)$0D and $P(T)\sim T^{-\alpha}g(T/T_c)$1 on the small-world network. At $P(T)\sim T^{-\alpha}g(T/T_c)$2, the variance exponent is reported as $P(T)\sim T^{-\alpha}g(T/T_c)$3, the silence exponent as $P(T)\sim T^{-\alpha}g(T/T_c)$4, the covariance-spectrum exponent as $P(T)\sim T^{-\alpha}g(T/T_c)$5 consistent with $P(T)\sim T^{-\alpha}g(T/T_c)$6 in directed percolation, and the dynamical exponent as $P(T)\sim T^{-\alpha}g(T/T_c)$7. For $P(T)\sim T^{-\alpha}g(T/T_c)$8, $P(T)\sim T^{-\alpha}g(T/T_c)$9 relaxes toward $\sigma$00, the silence exponent is smaller and decays faster, the spectrum flattens, and the autocorrelation becomes exponential rather than scaling. In momentum space, $\sigma$01 flows toward Gaussian for $\sigma$02, while at $\sigma$03 it retains clear non-Gaussian tails even when keeping only $\sigma$04 of modes (Nicoletti et al., 2020).

The same study highlights near-critical ambiguity. At $\sigma$05, variance and $\sigma$06 still show apparent power laws, with exponents intermediate between critical and deep super-critical values, but the covariance spectrum and $\sigma$07 do not exhibit clean scaling, and the momentum-space distribution rapidly becomes Gaussian. This is presented as evidence that some PRG observables can remain deceptive at distances of $\sigma$08 from the critical point, whereas spectrum and momentum-space diagnostics are more discriminating (Nicoletti et al., 2020).

A later analysis examines two neuronal models with known critical points: an excitable cellular automaton in the mean-field directed-percolation class on $\sigma$09 sites with branching ratio $\sigma$10 and $\sigma$11, and a stochastic excitatory–inhibitory integrate-and-fire network with inhibition strength $\sigma$12 and $\sigma$13. In both cases, $\sigma$14 time series are recorded from $\sigma$15 randomly subsampled neurons, binned adaptively via $\sigma$16, and analyzed with PRG. The momentum-space kurtosis remains near the Gaussian baseline $\sigma$17 away from criticality but exhibits a sharp peak at $\sigma$18 or $\sigma$19; surrogate data obtained by shuffling ISIs stays near $\sigma$20 throughout. Real-space coarse-graining yields $\sigma$21, with $\sigma$22 away from criticality, $\sigma$23 at extreme saturation, and nontrivial $\sigma$24 only within a narrow window around the true critical point. The reported conclusion is that PRG detects scaling only in a very narrow range around criticality under proper preprocessing (Nascimento et al., 16 Jun 2025).

In the lattice-model tradition, Borisenko et al. report quantitatively accurate flows for several systems. For the $\sigma$25d $\sigma$26 spin model, four fixed lines are found, including the standard Potts line $\sigma$27, where the critical coupling and exponent converge toward the exact values $\sigma$28 and $\sigma$29, with deviations $\sigma$30 for the sequence of steps $\sigma$31. Using $\sigma$32 matching on $\sigma$33 clusters, the predicted $\sigma$34 and $\sigma$35 for the $\sigma$36d Ising model approach the exact values $\sigma$37 and $\sigma$38 as $\sigma$39 grows to $\sigma$40, with similarly good accuracy for $\sigma$41. For $\sigma$42d $\sigma$43 lattice gauge models, the method predicts $\sigma$44 within a few percent of large-scale Monte Carlo and yields $\sigma$45–$\sigma$46, consistent with known universality classes (Borisenko et al., 2013).

6. Limitations, artifacts, and interpretive cautions

A recurrent theme in the modern PRG literature is that scaling signatures are necessary but not sufficient conditions for criticality. In the contact-process benchmark, no single PRG observable suffices: real-space variance and silence probability can show apparent power laws even when the system is not exactly critical, whereas covariance spectra, autocorrelation scaling, and momentum-space non-Gaussianity provide stricter cross-checks. The same study reports that the qualitative RG flow does not depend on the presence of long-range interactions, indicating that PRG is agnostic to network topology at that level, but this topological robustness should not be conflated with proof of criticality (Nicoletti et al., 2020).

A distinct source of ambiguity is preprocessing. In neuronal spike trains, binarization over time windows $\sigma$47 induces an average firing density $\sigma$48 that depends strongly on $\sigma$49. If $\sigma$50 is too small, most bins are empty and spurious correlations arise among all zeros, producing artificially high kurtosis; if $\sigma$51 is too large, almost every bin is nonempty, again producing large kurtosis. The proposed remedy is adaptive binning,

$\sigma$52

with $\sigma$53, which keeps $\sigma$54 roughly constant, for example around $\sigma$55, across subcritical and supercritical regimes. Under fixed rather than adaptive $\sigma$56, spurious large kurtoses are found deep in subcritical or supercritical regimes (Nascimento et al., 16 Jun 2025).

The literature also records more subtle failure modes. Large system size $\sigma$57 and a wide hierarchy of block sizes $\sigma$58 are required to resolve power laws and to separate Gaussian from non-Gaussian flows. For asynchronous updates, uncorrelated time points should be subsampled to avoid artificial temporal correlations. Moreover, simple extrinsic-noise models can produce non-Gaussian momentum-space tails without true interactions; in the formulation summarized for the contact-process study, direct-space pairing is presented as a way to catch such cases. These caveats place PRG in the category of a stringent but indirect inferential framework: it can rule out triviality more effectively when multiple observables agree, but it does not by itself replace a full dynamical or microscopic theory (Nicoletti et al., 2020).

A plausible implication is that the strongest use of PRG lies in comparative diagnosis rather than in isolated exponent fitting. When direct-space, momentum-space, spectral, and dynamical observables all point to the same non-Gaussian scale-invariant regime, the case for a critical fixed point is materially stronger than when only one observable displays an approximate power law. That interpretive standard is common to both the finite-size lattice formulation and the neural-data formulation, despite their different operational definitions of the RG step (Borisenko et al., 2013, Nascimento et al., 16 Jun 2025).