---
title: Thermal Universal Functional in GEM
url: https://www.emergentmind.com/topics/thermal-universal-functional
type: topic
---

# Thermal Universal Functional in GEM

Searching arXiv for the specified paper to ground the article in the cited source.
In the context of the generalized elastic model (GEM), the expression “thermal universal functional” does not denote an explicitly defined free-energy functional, Gibbs functional, or equilibrium action. The pertinent universal object is instead a thermal scaling structure for equilibrium correlation functions: under thermal initial conditions, the GEM admits stationary two-point observables that collapse onto scaling forms controlled by a characteristic correlation time $\tau(r)$ and a universal scaling function $f(u)$ [1203.3297]. Within this framework, the same model accommodates polymers, membranes, surfaces, fluctuating interfaces, and related systems, while non-thermal preparation preserves key exponents but alters amplitudes, stationarity, ageing behavior, and finite-time scaling sectors [1203.3297].

## 1. Generalized elastic model and stochastic dynamics

The GEM studied in [1203.3297] is a linear stochastic field theory for a $D$-component field $h_j(\mathbf{x},t)$ defined on a $d$-dimensional substrate $\mathbf{x}\in\mathbb{R}^d$. Its dynamics are governed by
$$
\frac{\partial}{\partial t} h_j(\mathbf{x},t) = \int d^d x'\,\Lambda(\mathbf{x}-\mathbf{x}') \frac{\partial^z}{\partial |\mathbf{x}'|^z} h_j(\mathbf{x}',t) +\eta_j(\mathbf{x},t),
$$
where $\Lambda(\mathbf{r})$ is the hydrodynamic or friction kernel, $\frac{\partial^z}{\partial |\mathbf{x}|^z}\equiv -(-\nabla^2)^{z/2}$ is the fractional Laplacian, and $\eta_j(\mathbf{x},t)$ is thermal Gaussian noise [1203.3297].

For long-ranged hydrodynamic interactions, the kernel is
$$
\Lambda(\mathbf{r}) = \frac{1}{|\mathbf{r}|^\alpha},
$$
whereas for local, screened hydrodynamics one has
$$
\Lambda(\mathbf{r})=\delta(\mathbf{r}).
$$
The stochastic forcing satisfies the fluctuation-dissipation relation
$$
\langle \eta_j(\mathbf{x},t)\eta_k(\mathbf{x}',t')\rangle = 2k_B T\,\Lambda(\mathbf{x}-\mathbf{x}')\,\delta_{jk}\,\delta(t-t').
$$
The analysis is restricted to rough systems satisfying $z>d$, and a central exponent combination is
$$
\gamma = 2(z+\alpha-d),
$$
with the local case treated formally by setting $\alpha=d$ together with $\Lambda=\delta$, while noting that this is not to be interpreted as a strict limit [1203.3297].

This setup is the basis for the paper’s universality statement. The stochastic equation itself is model-generic, but the equilibrium and non-equilibrium distinctions emerge through the preparation protocol and the resulting covariance structure rather than through a separate thermodynamic functional.

## 2. Thermal preparation and the meaning of universality

Thermal initial conditions mean that the system has reached stationarity at $t=-\infty$ and that observation begins at $t=0$ [1203.3297]. This implies time-translation invariance, stationary correlations depending only on $t-t'$, and compatibility with fluctuation-dissipation because the noise is thermal and matched to $\Lambda$ [1203.3297]. Physically, this corresponds to a polymer already relaxed in its coil, a membrane already equilibrated, or a rough interface already at thermal equilibrium.

By contrast, non-thermal initial conditions are imposed as
$$
h(\mathbf{x},0)=0 \qquad \forall \mathbf{x},
$$
representing a flat or unrelaxed initial state [1203.3297]. The paper lists as examples a membrane from flat initial condition, interface growth on a flat substrate, equally spaced particles in single-file diffusion, and a polymer after forced translocation [1203.3297]. This preparation breaks stationarity and generates explicit dependence on $t$ and $t'$ separately, rather than only on their difference.

The term “thermal universal functional” is therefore best understood as a shorthand for the universal thermal scaling function and correlation kernel that characterize the stationary GEM. The paper explicitly states that it does not provide an equilibrium free-energy functional $\mathcal{H}[h]$, a Gibbs weight $P[h]\propto e^{-\beta \mathcal H[h]}$, or a Martin–Siggia–Rose/Janssen–De Dominicis action identified as a universal thermal functional [1203.3297]. A plausible implication is that universality here is operational and covariance-based rather than variational.

## 3. Universal thermal scaling function and correlation time

The paper’s main universal object is the scaling form of the thermal two-point, two-time increment correlation:
$$
\langle \delta_t h(\mathbf{x},t)\,\delta_{t'} h(\mathbf{x}',t')\rangle_{\rm th} = \frac{k_B T}{(2\pi)^{d/2} |\mathbf{x}-\mathbf{x}'|^{z-d} }
\left[ f\!\left(\frac{t}{\tau(r)}\right) + f\!\left(\frac{t'}{\tau(r)}\right) - f\!\left(\frac{|t-t'|}{\tau(r)}\right) \right],
$$
with $r=|\mathbf{x}-\mathbf{x}'|$ and
$$
f[u] = \int_0^\infty dy\, y^{d/2-z} J_{d/2-1}(y) \left(1-e^{-y^{\gamma/2}u}\right).
$$
The associated correlation time is
$$
\tau(r)=\frac{r^{\gamma/2}}{A},
$$
and equivalently the correlation length is
$$
\xi(t)=(At)^{2/\gamma}
$$
[1203.3297].

This pair, $\tau(r)$ and $f(u)$, is the closest object in the paper to a “thermal universal functional.” Once $d$, $z$, $\alpha$, and the scale $A$ are fixed, the equilibrium space-time covariance is determined by this scaling function and its asymptotic structure [1203.3297]. The universality is not tied to a single microscopic realization; rather, it organizes the crossover between short-time/large-distance and long-time/short-distance regimes across the class of systems represented by the GEM.

The asymptotic behavior of $f(u)$ encodes the thermal scaling sectors. For long-ranged hydrodynamics and small $u\ll 1$,
$$
f[u]\sim 2^{\,z-d/2}\, \frac{\Gamma(z/2)}{\Gamma((d-z)/2)}\,u.
$$
For local hydrodynamics and generic $z\neq 2m$,
$$
f[u]\sim 2^{\,z+d/2-2} \frac{z}{\pi} \sin\!\left(\frac{z\pi}{2}\right) \Gamma(z/2)\Gamma((z+d)/2)\, u^2.
$$
For local hydrodynamics with $z=2m$,
$$
f[u]\sim u^{\,1+d/2-z}\,e^{-u^{-2/(z-d)}},
$$
up to constants. In the large-$u$ regime,
$$
f[u]\sim \frac{2^{1-d/2}}{z-d} \frac{\Gamma(1-\beta)}{\Gamma(d/2)} u^\beta,
$$
with
$$
\beta=\frac{2(z-d)}{\gamma}=\frac{z-d}{z+\alpha-d}, \qquad 0<\beta<1
$$
[1203.3297].

A central consequence is that when $t,t'\gg \tau(r)$, the spatial dependence drops out and the two-point thermal increment correlation approaches the one-point form. This reflects the condition $\xi(t)\gtrsim r$, under which the system is correlated over distance $r$ [1203.3297].

## 4. Stationary covariance, structure factor, and tracer dynamics

At the same spatial point, the thermal two-time covariance is
$$
\langle \delta_t h(\mathbf{x},t)\,\delta_{t'} h(\mathbf{x},t')\rangle_{\rm th} = K\left[t^\beta + t'^\beta - |t-t'|^\beta\right],
$$
where
$$
\delta_t h(\mathbf{x},t)=h(\mathbf{x},t)-h(\mathbf{x},0),
$$
and
$$
K= \frac{2k_B T\,\pi^{d/2}}{(2\pi)^d} \frac{A^\beta\,\Gamma(1-\beta)}{(z-d)\Gamma(d/2)}.
$$
The paper identifies this as exactly the covariance structure of a fractional Brownian motion in time with Hurst index
$$
H_t=\frac{\beta}{2}
$$
for the tagged probe [1203.3297]. The process is Gaussian and stationary in increments.

The corresponding one-point mean-square displacement under thermal preparation is
$$
\langle [h(\mathbf{x},t)-h(\mathbf{x},0)]^2\rangle_{\rm th} = 2K t^\beta.
$$
The structure factor at equal time is
$$
\langle h(\mathbf{q},t)h(\mathbf{q}',t)\rangle_{\rm th} = k_B T(2\pi)^d \frac{1}{|\mathbf{q}|^z} \delta(\mathbf{q}+\mathbf{q}').
$$
This equilibrium covariance kernel is another candidate for what might informally be called the thermal universal object, although it is a correlation kernel rather than a free-energy functional [1203.3297].

In Fourier representation, thermal preparation uses transforms in both space and time, yielding
$$
h(\mathbf{q},\omega)=\frac{\eta(\mathbf{q},\omega)}{-i\omega + A|\mathbf{q}|^{z+\alpha-d}},
$$
with thermal noise covariance
$$
\langle \eta_j(\mathbf{q},\omega)\eta_k(\mathbf{q}',\omega')\rangle = 2k_BT(2\pi)^{d+1}\delta_{jk}\, A|\mathbf{q}|^{\alpha-d}\, \delta(\mathbf{q}+\mathbf{q}')\delta(\omega+\omega')
$$
and field covariance given in the paper as Eq. (3) [1203.3297]. These Fourier-space expressions provide the stationary basis from which the universal thermal scaling form is derived.

## 5. Spatial morphology, roughening, and the transition at \( z=d+2 \)

The thermal one-time connected increment
$$
\langle [h(\mathbf{x},t)-h(\mathbf{0},t)] [h(\mathbf{x}',t)-h(\mathbf{0},t)]\rangle_{\rm th}
$$
is expressed in terms of a spatial kernel $l_{\rm th}[\mathbf{x}]$:
$$
\langle \delta_x h(\mathbf{x},t)\,\delta_{x'} h(\mathbf{x}',t)\rangle_{\rm th} = \frac{k_B T}{(2\pi)^{d/2}}
\left\{ l_{\rm th}[\mathbf{x}] + l_{\rm th}[\mathbf{x}'] - l_{\rm th}[\mathbf{x}-\mathbf{x}'] \right\},
$$
with $l_{\rm th}[\mathbf{x}]$ given explicitly in Eq. (20) of [1203.3297]. The analysis reveals a morphological transition at
$$
z=d+2.
$$

For $z<d+2$, the system belongs to the Family–Vicsek class, and spatial increments behave like fractional Brownian motion in space with Hurst exponent
$$
H_x=\frac{z-d}{2}.
$$
For $z\ge d+2$, an infrared divergence requires a system-size cutoff, yielding effectively
$$
l_{\rm th}[\mathbf{x}] \sim L^{z-d-2}|\mathbf{x}|^2,
$$
so that the root-mean-square height difference is linear in distance [1203.3297].

The same transition organizes roughening behavior. For a system of size $L$, after saturation time $\tau(L)$ the global width obeys
$$
W^2(L,t)\sim L^{2\chi}, \qquad \chi=\frac{z-d}{2}.
$$
For intermediate scales $\tau(l)\ll t\ll \tau(L)$, the local width behaves as
$$
w^2(l,t)\sim l^{2\chi_{\rm loc}}.
$$
Using the non-thermal scaling function $g_{\rm nth}$, the paper obtains $\chi_{\rm loc}=\chi$ for Family–Vicsek systems with $z<d+2$, and $\chi_{\rm loc}=1$ for super-rough systems with $z\ge d+2$ [1203.3297]. The examples given are flexible Zimm polymers as Family–Vicsek systems, and fluid membranes and semiflexible polymers as super-rough systems.

This suggests that the thermal universal scaling structure does not merely organize temporal covariance; it also encodes the crossover between ordinary roughening and super-roughening through the same exponent set $(d,z,\alpha)$.

## 6. Non-thermal preparation, ageing, and approach to equilibrium

For non-thermal initial conditions, the system starts from $h(\mathbf{x},0)=0$, and the natural representation is Fourier in space and Laplace in time [1203.3297]. The solution is
$$
h(\mathbf{q},s)=\frac{\eta(\mathbf{q},s)}{s+A|\mathbf{q}|^{z+\alpha-d}},
$$
leading to the covariance reported as Eq. (13) in the paper.

The real-space two-time increment correlation becomes
$$
\langle \delta_t h(\mathbf{x},t)\,\delta_{t'} h(\mathbf{x}',t')\rangle_{\rm nth} =
\frac{k_B T}{(2\pi)^{d/2} |\mathbf{x}-\mathbf{x}'|^{z-d}}
\left[ f\!\left(\frac{t+t'}{\tau(r)}\right) - f\!\left(\frac{|t-t'|}{\tau(r)}\right) \right].
$$
At the same point,
$$
\langle \delta_t h(\mathbf{x},t)\,\delta_{t'} h(\mathbf{x},t')\rangle_{\rm nth} =
K\left[(t+t')^\beta - |t-t'|^\beta\right].
$$
The exponent $\beta$ is unchanged, but the functional dependence differs from the thermal case by replacing $t^\beta+t'^\beta$ with $(t+t')^\beta$ [1203.3297]. The process is therefore non-stationary and ageing.

The non-thermal mean-square displacement is
$$
\langle [h(\mathbf{x},t)-h(\mathbf{x},0)]^2\rangle_{\rm nth} = K(2t)^\beta,
$$
which has the same exponent as the thermal result but a different amplitude [1203.3297]. The equal-time structure factor is
$$
\langle h(\mathbf{q},t)h(\mathbf{q}',t)\rangle_{\rm nth} = k_B T(2\pi)^d \frac{1}{|\mathbf{q}|^z} \left(1-e^{-2A|\mathbf{q}|^{\gamma/2}t}\right) \delta(\mathbf{q}+\mathbf{q}'),
$$
where the factor $1-e^{-2A|\mathbf{q}|^{\gamma/2}t}$ is the explicit memory of initial preparation [1203.3297]. At long times it relaxes to the thermal spectrum.

The waiting-time-dependent MSD makes the ageing structure explicit:
$$
\left\langle [h(\mathbf{x},t+t_0)-h(\mathbf{x},t_0)]^2\right\rangle_{\rm nth}
= K\Big[(2(t+t_0))^\beta + (2t_0)^\beta + 2t^\beta - 2(t+2t_0)^\beta\Big].
$$
Its asymptotic behavior is
$$
\left\langle [h(\mathbf{x},t+t_0)-h(\mathbf{x},t_0)]^2\right\rangle_{\rm nth} \sim
\begin{cases}
K(2t)^\beta, & t_0\ll t,\\[4pt]
2K t^\beta, & t_0\gg t.
\end{cases}
$$
Thus the system ages, but for $t_0\gg t$ the MSD becomes identical to the thermal MSD even though the overall state remains out of equilibrium [1203.3297].

The paper emphasizes that this ageing is not continuous-time-random-walk-like. For subdiffusive CTRW with waiting-time tail $\psi(t)\sim t^{-1-\mu}$, the aged MSD slows as $t_0$ increases, while in the non-thermal GEM the leading large-$t_0$ behavior is $2Kt^\beta$, independent of $t_0$ to leading order [1203.3297]. A plausible implication is that ageing in the GEM arises from deterministic relaxation of an initially prepared rough field under thermal noise, rather than from heavy-tailed renewal statistics.

## 7. Ergodicity, diagnostics, and the strict status of the term

The paper distinguishes thermal and non-thermal preparation through ergodic properties of the tracer-level time-averaged MSD:
$$
\overline{\delta_t^2 h(\mathbf{x},t)} = \frac{1}{T-t}\int_0^{T-t} dt_0\, [h(\mathbf{x},t+t_0)-h(\mathbf{x},t_0)]^2.
$$
For thermal preparation,
$$
\lim_{T\to\infty}\overline{\delta_t^2 h(\mathbf{x},t)} =
\langle [h(\mathbf{x},t)-h(\mathbf{x},0)]^2\rangle_{\rm th} = 2Kt^\beta.
$$
The tracer dynamics are therefore ergodic in the sense discussed in the paper [1203.3297].

For non-thermal preparation, the ensemble average of the time-averaged MSD is given by Eq. (30) of [1203.3297], which contains explicit dependence on the total trajectory length $T$. In the limit $t/T\to 0$,
$$
\left\langle \overline{\delta_t^2 h} \right\rangle_{\rm nth} \to 2Kt^\beta,
$$
so finite-lag time averages mimic equilibrium. When $t$ approaches $T$, however, the explicit $T$-dependence persists and reveals the non-equilibrium preparation [1203.3297].

This is the paper’s practical diagnostic: the finite-trajectory time-averaged MSD, especially its dependence on measurement time $T$, can distinguish whether the system is in or far from equilibrium, whereas the ordinary MSD can be misleading [1203.3297]. Experimentally, short lag-time tracer analysis may therefore suggest equilibrium even when the system retains memory of non-thermal preparation.

In strict usage, “thermal universal functional” is therefore a misnomer for this paper’s contribution. No universal thermodynamic functional is introduced. The precise universal object is the thermal scaling function
$$
f[u] = \int_0^\infty dy\, y^{d/2-z} J_{d/2-1}(y)\left(1-e^{-y^{\gamma/2}u}\right)
$$
together with the correlation time
$$
\tau(r)=\frac{r^{\gamma/2}}{A},
$$
which jointly determine the equilibrium covariance structure of the generalized elastic model [1203.3297]. The broader significance of this framework is that it unifies stationary equilibrium scaling, roughening, and tracer-level ergodic behavior within a single covariance-based description, while making clear how non-thermal preparation modifies amplitudes, stationarity, and ageing without changing the core exponent set.

Source: https://www.emergentmind.com/topics/thermal-universal-functional