Abstract
A reduced transport model previously proposed for gyrokinetic turbulence simulations has been shown to reproduce turbulent transport over a wide range of parameters. While this model successfully captures simulation results, the origin of its functional form has remained largely empirical. In this work, we investigate a phenomenological representation of turbulent transport based on a minimal set of physical constraints. By separating turbulence-driven transport enhancement from zonal-flow-induced regulation, a general class of admissible transport functions is constructed. Representative regulation functions constructed within a physically motivated class of admissible forms are shown to share a common asymptotic structure in the weak-regulation regime relevant to gyrokinetic simulations. Based on this observation, a lowest-order rational representation is identified as a minimal realization of the phenomenological constraints. The resulting expression naturally includes the transport models previously proposed as special cases. This work provides a phenomenological and scaling-based foundation for reduced turbulent transport models, thereby clarifying the relation between previously proposed empirical models and a minimal rational realization of the phenomenological constraints and establishing a first step toward a systematic classification of admissible transport models. Such a classification can provide objective guidelines for selecting appropriate reduced models across different physical transport regimes.
Introduction
Turbulent transport driven by microinstabilities plays a central role in determining the confinement properties of magnetically confined plasmas. Considerable efforts have therefore been devoted to development of transport models for predictive simulations, including quasi-linear approaches such as TGLF [1] and QuaLiKiz [2]. As zonal-flows generated through nonlinear interactions regulate turbulent transport [3], understanding the interplay between turbulence and zonal-flows is a critical issue in transport modeling. In order to characterize this interplay in a reduced form, our previous studies [4–6] introduced measures of turbulence intensity T and zonal-flow intensity Z, defined as
|
T
=
1
2
∑
k
x
,
k
y
≠
0
⟨
|
ϕ
~
k
x
,
k
y
|
2
⟩
,
| (1) |
|
Z
=
1
2
∑
k
x
⟨
|
ϕ
~
k
x
,
k
y
=
0
|
2
⟩
.
| (2) |
Here, ϕ~ is the normalized electrostatic potential fluctuation defined as ϕ~=ϕ/(Tiρi/eR0), the average along the field line is denoted by ⟨⋯⟩, and (kx,ky) represent the radial and poloidal wavenumbers. Hereafter, T and Z are regarded as generic measures of turbulence and zonal-flow activity, respectively. Using these measures, the normalized ion heat transport coefficient χ/χGB was represented by the following physically motivated form:
|
χ
χ
GB
=
F
(
T
,
Z
)
≡
C
1
T
α
C
2
+
Z
β
/
T
.
| (3) |
Here, χGB is the gyro-Bohm diffusivity, and C1,C2 and α are positive fitting parameters. The usefulness of Eq. (3) has been demonstrated in previous studies. When combined with additional linear physics quantities, such as the linear instability growth rate and linear zonal-flow response properties (e.g., residual levels or decay times of the response function), it served as a core component of a reduced transport model that successfully reproduced gyrokinetic transport levels over a broad parameter range. This functional form has subsequently been extended to applications involving solution trajectories extracted from time-series data [6, 7], as well as to discussions of the model extension for broader plasma parameter regimes [8]. Furthermore, a parameter optimization based on the integrated distribution of fitting errors suggested β≃ 0.5 as a robust scaling exponent for the zonal-flow contribution in the solution trajectory analysis for the LHD plasmas [6]. Although Eq. (3) successfully reproduces gyrokinetic simulation results, its specific functional form was originally introduced based on physical considerations rather than derived from a systematic set of constraints. It therefore remains unclear whether Eq. (3) represents one empirical fitting formula among many possible alternatives or whether it can be regarded as a minimal representation arising from more general physical requirements. Rather than proposing another empirical fitting function, the purpose of the present work is to identify a minimal functional form consistent with the essential phenomenology of turbulence enhancement and zonal-flow regulation, thereby revealing the essential structure governing turbulent transport regulation.
General Representation
To address the above purpose, we first introduce a general representation of the normalized heat diffusivity χ/χGB, consistently with the transport quantity employed in our previous reduced models [4, 5, 8], whose functional form is the primary subject of the present analysis. Motivated by the phenomenology of turbulence–zonal-flow interactions, we decompose the transport process into turbulence-driven enhancement and zonal-flow regulation. The former determines the overall transport level, whereas the latter acts as a regulating mechanism. Based on this decomposition, we introduce a general class of phenomenological representations for the transport coefficient. With a positive constant A, it is expressed as
|
χ
χ
GB
=
F
(
gen
)
(
T
,
Z
)
≡
A
T
α
G
(
u
)
,
| (4) |
where the factor Tα with a power exponent α > 0 represents the transport enhancement by turbulence, and G(u) is the regulation function. This form automatically satisfies the physically required condition that the transport vanishes in the absence of turbulence, χ → 0 as T → 0. Since zonal-flow regulation is expected to depend on the relative strength of zonal-flows with respect to turbulence rather than on their absolute amplitudes, and since the requirement of scale consistency–namely that the transport representation preserves its self-similar scaling relation under a simultaneous rescaling of T and Z, as explicitly formulated below–implies that the regulation function can depend only on a scale-invariant combination of T and Z, the function G(u) is assumed to depend on a dimensionless scaling variable
where p > 0 and q > 0 are scaling exponents. The condition p > 0 reflects the regulating role of zonal-flows, whereas q > 0 accounts for the tendency of stronger turbulence to mitigate the relative effectiveness of zonal-flow shearing due to enhanced turbulent decorrelation. The variable u characterizes the relative strength of zonal-flows in a scale-independent manner. Indeed, under the simultaneous rescaling
|
T
→
λ
T
and
Z
→
λ
q
/
p
Z
,
| (6) |
the quantity u remains invariant. Consequently, this scale consistency leads to the self-similar scaling behavior of the transport coefficient,
|
F
(
gen
)
(
λ
T
,
λ
q
/
p
Z
)
=
λ
α
F
(
gen
)
(
T
,
Z
)
.
| (7) |
This formulation clarifies that while the overall transport scale is governed by the turbulence-dependent prefactor Tα, the regulation efficiency is uniquely determined by the relative balance between turbulence and zonal-flows through G(u), whose functional form will be systematically examined in the following sections.
Phenomenological Constraints
In order to identify the admissible class of regulation functions, we impose the following phenomenological constraints. For clarity, each condition is first stated in terms of the general transport coefficient χ/χGB=F(gen) in Eq. (4), followed by its equivalent representation in terms of G(u).
-
(i)
Turbulence-dominated limit:
-
The reference state is defined by the absence of zonal-flow regulation or turbulence-dominated limit. In this limit, the transport coefficient must recover the pure turbulence scaling. Since u=ZpT−q → 0 as Z → 0 (or equivalently as T→∞), without loss of generality, we can impose the normalization for the transport coefficient such that F(gen)(T,Z=0)=ATα. This can yield the normalization condition for G(u) as follows,
which fixes the reference state corresponding to vanishing zonal-flow regulation. Note that this constraint is imposed as a mathematical boundary condition to restrict the functional form of G(u), independent of the actual dynamical trajectory of the plasma state in realistic regimes.
-
(ii)
Monotonic suppression by zonal-flows:
-
Increasing zonal-flow intensity reduces turbulent transport, namely, ∂χ/∂Z=∂F(gen)(T,Z)/∂Z<0. Since u is an increasing function of Z, this is equivalently expressed as
|
d
G
(
u
)
d
u
<
0
.
| (9) |
-
(iii)
Strong zonal-flow limit:
-
In the limit of dominant zonal-flow regulation, turbulent transport is expected to be asymptotically suppressed, i.e., limZ→∞F(gen)(T,Z)=0. This corresponds to the asymptotic condition,
|
lim
u
→
∞
G
(
u
)
=
0
.
| (10) |
These constraints (8)–(10) establish a direct one-to-one correspondence between physically motivated requirements on the transport coefficient χ and conditions on the regulation function G(u). The formulation in terms of G(u) therefore constitutes a reduced representation of the above phenomenological constraints on χ rather than an additional modeling assumption.
Candidate Functional Forms
The constraints derived in the previous discussion (i)–(iii) define a restricted class of admissible regulation functions for G(u). Here we examine representative functional forms belonging to this class in order to elucidate their common structural properties and to identify a minimal representation of turbulent transport.
A straightforward choice is an exponential regulation function, which represents a smooth reduction of transport with increasing zonal-flow regulation strength,
|
G
1
(
u
)
=
exp
(
−
b
u
)
,
| (11) |
where b > 0 is a free parameter controlling the suppression rate. Another simple candidate is the rational form,
|
G
2
(
u
)
=
1
1
+
b
u
,
| (12) |
where b > 0 determines the strength of zonal-flow regulation. A broader class of admissible functions is provided by the generalized rational form,
|
G
3
(
u
)
=
(
1
+
b
c
u
)
−
c
,
| (13) |
where b > 0 and c > 0 are free parameters. The generalized rational form G3(u) in Eq. (13) encompasses the simple rational model G2(u) in Eq. (12) as a special case (c = 1), while asymptotically approaching the exponential form G1(u) in Eq. (11) in the limit c→∞. By construction, these candidate functions satisfy the phenomenological constraints (i)–(iii), despite exhibiting distinct behaviors in their higher-order contributions. Their common asymptotic structure in specific regimes will be examined in the following paragraph.
Minimal Representation
The candidate functions introduced above satisfy the same phenomenological constraints, yet exhibit different higher-order behaviors. To identify the essential structure shared by representative admissible representations, we examine their asymptotic behavior in the weak-regulation regime u≪1. From Eqs. (11)–(13), we can obtain the expanded forms,
|
G
1
(
u
)
=
1
−
b
u
+
b
2
2
u
2
+
O
(
u
3
)
,
| (14) |
|
G
2
(
u
)
=
1
−
b
u
+
b
2
u
2
+
O
(
u
3
)
,
| (15) |
|
G
3
(
u
)
=
1
−
b
u
+
c
+
1
2
c
b
2
u
2
+
O
(
u
3
)
.
| (16) |
Despite their different higher-order terms, the representative admissible functions considered here share the same leading-order structure,
|
G
0
(
u
)
=
1
−
b
u
+
O
(
u
2
)
.
| (17) |
Figure 1 compares representative regulation functions satisfying the phenomenological constraints. Note that while varying b and c quantitatively modifies higher-order terms O(u2), it does not alter the common leading-order structure 1−bu in the weak-regulation regime (u≪1). Although discrepancies emerge at larger values of u, all functions collapse onto the same leading-order behavior in the weak-regulation regime. In Fig. 1, the representative value of u in a typical gyrokinetic simulation in LHD plasmas [6] is indicated, u ∼ 0.045, where (T,Z)∼(100,20) in the common normalization unit and (p,q)=(0.5,1.0) which corresponds to the original empirical functional form in Ref. [5]. Even in regimes with substantially enhanced zonal-flow generation reported in Ref. [5], u remains relatively small, e.g., u∼0.07 for (T,Z)∼(100,50), indicating that practical simulation results are largely confined within this weak-regulation regime. This observation suggests that the actual transport behavior is primarily governed by the common leading-order structure of Eq. (17), while higher-order corrections depend on the specific choice of the empirical functional form. Consequently, the phenomenological constraints do not uniquely dictate G(u), but rather restrict it to a class of functions sharing the same leading-order asymptotic behavior. Among the admissible candidates, the simple rational form

|
G
(
u
)
=
G
2
(
u
)
=
1
1
+
b
u
,
| (18) |
may be regarded as the lowest-order rational representation preserving the common asymptotic behavior of Eq. (17) G0(u)=1−bu+O(u2). In particular, it corresponds to the lowest-order [0/1] Padé approximant of the exponential regulation function G1(u)=exp(−bu), and thereby exactly reproduces the common leading-order structure shared by the representative candidate functions considered in this work. For this reason, without introducing any additional free parameter, Eq. (18) is adopted as a minimal rational realization of the phenomenological constraints. Substituting Eq. (18) into the general scaling representation introduced in Eq. (4) yields
|
F
(
gen
)
(
T
,
Z
)
=
A
T
α
1
+
b
Z
p
T
−
q
.
| (19) |
Retaining the freedom of overall normalization and scaling coefficients, Eq. (19) can be rewritten in another form for the transport coefficient,
|
χ
χ
GB
=
C
1
T
α
C
2
+
Z
p
T
−
q
,
| (20) |
Equation (20) has the same structure as the transport model introduced in Eq. (3), which corresponds to the special case p=β and q = 1 which was originally proposed as an empirical expression capable of reproducing gyrokinetic simulation results. Furthermore, Eq. (20) can be rewritten as
|
χ
χ
GB
=
C
1
T
α
C
2
+
(
Z
ξ
/
T
)
ζ
,
| (21) |
through a redefinition of the exponents with ξ=p/q and ζ=q. This expression is identical in structure to the generalized transport model proposed in Ref. [8]. Therefore, the present analysis provides a phenomenological and scaling-based foundation for both the original empirical model and its subsequent generalization. In particular, the generalized form emerges naturally from the minimal phenomenological constraints imposed on turbulent transport, rather than being introduced solely as an ad-hoc empirical fitting ansatz.
Summary
In this work, we have investigated the functional form of turbulent transport coefficients regulated by zonal-flows from a phenomenological viewpoint. Starting from a general scaling representation, in which turbulent transport is expressed as the product of turbulence-driven amplification and zonal-flow-induced suppression, we introduced a dimensionless regulation variable and reformulated the problem in terms of a single regulation function. By imposing a minimal set of phenomenological constraints, including the turbulence-dominated limit, monotonic transport suppression by zonal-flows, and vanishing transport in the strong zonal-flow limit, we identified a restricted class of admissible regulation functions. Representative candidates, including exponential and rational forms, were shown to share a common leading-order asymptotic structure in the weak-regulation regime. Since the dimensionless regulation is typically small in gyrokinetic simulations, the transport behavior is largely governed by this common leading-order structure, while higher-order corrections play only a minor role. Based on this observation, a minimal representation of the regulation function was obtained, leading to a transport model of a rational form. The resulting expression was shown to be equivalent to transport models previously introduced on empirical grounds to reproduce gyrokinetic simulation results. The present results indicate that the previously proposed transport models can be interpreted as a minimal realization of the phenomenological constraints governing turbulence–zonal-flow interactions.
While the present work adopts the lowest-order Padé representation as a minimal rational realization, the phenomenological constraints themselves do not uniquely specify the regulation function. A systematic classification of admissible functional forms and their higher-order corrections or different onset behaviors constitutes an important direction for future work. In particular, exploring alternative regulation functions exhibiting different onset behaviors, such as G(u)=exp(−bun) with n > 1, may be relevant in specific transport regimes, although they deviate from the minimal leading-order structure considered here. Furthermore, higher-order effects may become important in weakly driven turbulence regimes, where u is no longer small, i.e., the turbulence level is low and zonal-flow regulation can become comparatively strong, potentially requiring corrections beyond the minimal representation considered here.
The authors are very grateful to Meta-hierarchy Dynamics unit at National Institute for Fusion Science (NIFS) for useful discussions. This work is supported in part by the Japan Society for the Promotion of Science (JSPS) KAKENHI Grant Numbers 24K00615 and 26K17120, the NIFS Collaborative Research Program, NIFS25KISM011 and NIFS26KISI005, and the Collaborative Research Program of Research Institute for Applied Mechanics, Kyushu University. One of the authors (M. Nunami) is supported by JST (Moonshot R&D Program) Japan Grant Number JPMJMS24A3 for the development of transport models.
References
- [1] G.M. Staebler et al., Phys. Plasmas 12, 102508 (2005).
- [2] C. Bourdelle et al., Plasma Phys. Control. Fusion 58, 014036 (2016).
- [3] P.H. Diamond et al., Plasma Phys. Control. Fusion 47, R35 (2005).
- [4] M. Nunami et al., Plasma Fusion Res. 8, 1203019 (2013).
- [5] M. Nunami et al., Phys. Plasmas 20, 092307 (2013).
- [6] M. Nunami et al., Plasma Phys. Control. Fusion 67, 065038 (2025).
- [7] K. Fujii and M. Nunami, Plasma Fusion Res. 17, 2403030 (2022).
- [8] T. Nakayama et al., Sci. Rep. 13, 2319 (2023).