---
title: Mukhanov Parametrization in Cosmology
url: https://www.emergentmind.com/topics/mukhanov-parametrization
type: topic
---

# Mukhanov Parametrization in Cosmology

“Mukhanov parametrization” does not denote a single universally fixed construction. In current arXiv usage, it designates several related but technically distinct parametrization schemes associated with V. Mukhanov’s perturbation formalism and its later extensions: direct parametrizations of the Mukhanov–Sasaki mode equation, the inflationary equation-of-state ansatz \(1+\omega(N)=\beta/(N+1)^\alpha\), canonical gauge-invariant reformulations in loop quantum cosmology and quantum cosmology, and, in a separate black-hole context, the Bekenstein–Mukhanov linear area spectrum \(A=aN\ell_p^2\) [2601.04760] [2507.05648] [1609.07125]. Related literature also uses the name in the Markov–Mukhanov modification of the Einstein–Hilbert action, where the central object is an energy-density-dependent coupling \(\chi(\varepsilon)\) [2510.14416].

## 1. Range of meanings

In the literature covered here, the expression is used in several non-identical senses. The common thread is the replacement of a detailed microscopic model by a compact parametrization of the relevant dynamics.

| Usage | Defining object | Representative papers |
|---|---|---|
| Perturbation-equation parametrization | Modifications of \(v_k''+\left(k^2-z''/z\right)v_k=0\) | [2601.04760], [1012.4811] |
| Inflationary equation-of-state parametrization | \(1+\omega(N)=\beta/(N+1)^\alpha\) | [2412.16703], [2507.05648] |
| Canonical gauge-invariant scalar mode | \(v=z\mathcal R\), with \(z=a\varphi'/\mathcal H\), and LQC-modified analogues | [1111.7192], [1503.03907] |
| Black-hole area quantization | \(A=aN\ell_p^2\), or \(A_{\rm cl}=\tilde a\,\ell_p^2 N\) | [1609.07125] |
| Energy-density coupling in modified gravity | \(\chi(\varepsilon)\) in the Markov–Mukhanov action | [2510.14416] |

A recurrent source of confusion is that some papers use “Mukhanov parametrization” for the background equation of state of inflation, whereas others use it for perturbation dynamics, initial states, or even black-hole area spectra. The technical content therefore depends entirely on context.

## 2. Parametrization at the level of the Mukhanov–Sasaki equation

The standard single-field canonical setup introduces the comoving curvature perturbation \(\zeta\) and the canonical Mukhanov–Sasaki variable
\[
v \equiv z\,\zeta,\qquad z \equiv a\,\frac{\dot\phi}{H},
\]
with Fourier modes satisfying
\[
v_k''+\left(k^2-\frac{z''}{z}\right)v_k=0.
\]
The Bunch–Davies condition is imposed by requiring
\[
v_k(\eta)\xrightarrow{-k\eta\to\infty}\frac{1}{\sqrt{2k}}e^{-ik\eta},
\]
and the scalar power spectrum is
\[
\mathcal P_\zeta(k)=\lim_{-k\eta\to0}\frac{k^3}{2\pi^2}\left|\frac{v_k}{z}\right|^2.
\]
In this line of work, “Mukhanov parametrization” means parametrizing departures from the standard slow-roll form directly in the mode equation, for example through a modified effective mass term, modified dispersion relation, or extra time-dependent operators [2601.04760].

A particularly explicit realization is the generalized Sasaki–Mukhanov equation
\[
v_k''+\left[k^2-\frac{\nu^2-\frac14}{\eta^2}+\frac{f}{\eta}\right]v_k=0,
\]
with \(f\) a constant parameter and \(\lambda=f/(2k)\). This converts the mode equation into Whittaker form and admits the exact normalized solution
\[
v_k(\eta)=\frac{1}{\sqrt{2k}}\frac{1}{e^{\pi\lambda/2}}\,W_{-i\lambda,\nu}(2ik\eta).
\]
The primordial spectrum can then be written as
\[
\mathcal P_\zeta(k)=\mathcal P_\zeta^{\rm(std)}(k)\,\mathcal C(\lambda),
\qquad
\mathcal P_\zeta^{\rm(std)}(k)=A_s\left(\frac{k}{k_\star}\right)^{n_s-1},
\]
with
\[
\mathcal C(\lambda)=\frac{e^{\pi\lambda}\sinh(\pi\lambda)}{\pi\lambda(1+\lambda^2)}\,\mathcal F(\epsilon,\delta).
\]
Because \(\lambda=f/(2k)\), the correction is scale-selective and is strongest at low \(k\); negative \(f\) suppresses large-scale power, while positive \(f\) enhances it. A Planck 2018 plus ACT DR6 Bayesian analysis gives
\[
f=(-0.52\pm0.51)\times10^{-4}\quad(68\%\ {\rm C.L.}),
\qquad |f|\lesssim10^{-4}\quad(95\%\ {\rm C.L.}),
\]
with a slight but non-decisive preference for negative \(f\) and a modest improvement of the low-\(\ell\) fit [2601.04760].

This mode-equation usage includes other parametrization strategies as well. One paper replaces the exact Mukhanov potential \(f(\tau)=z''/z\) or \(a''/a\) by analytically solvable fitting functions and imposes boundary conditions at a finite conformal time rather than at \(\tau\to-\infty\) [1012.4811]. Another paper reparametrizes the evolution as a sequence of analytic segments in e-fold time \(N\), using the variables
\[
\mathcal S \equiv z\sqrt{aH}\,\mathcal R,\qquad \mathcal Q \equiv a\sqrt{aH}\,\mathcal T,
\]
so that solving the Mukhanov–Sasaki equation reduces to transfer-matrix multiplication [1809.11095]. In all such cases, the parametrized object is not the inflaton potential itself but the effective frequency or effective mass entering the perturbation equation.

## 3. Equation-of-state parametrization of inflation

A second major usage specifies inflation through the equation-of-state parameter as a function of the number of e-folds before the end of inflation:
\[
1+\omega(N)=\frac{\beta}{(N+1)^\alpha},
\qquad \alpha>0,\ \beta>0.
\]
Here \(N=0\) at the end of inflation and \(N\sim50\text{–}60\) at CMB scales. In Hamilton–Jacobi language, the exact relation
\[
\epsilon_H=\frac32(1+\omega)
\]
implies
\[
\epsilon_H(N)=\frac32\frac{\beta}{(N+1)^\alpha}.
\]
This determines the background evolution through
\[
dN=\frac{1}{\epsilon_H}\frac{dH}{H},
\]
leading to
\[
H(N)=
\begin{cases}
H_0(1+N)^{\frac{3\beta}{2}}, & \alpha=1,\\[4pt]
H_0\exp\!\left[\dfrac{3\beta}{2(1-\alpha)}(1+N)^{1-\alpha}\right], & \alpha\neq1.
\end{cases}
\]
The scalar field trajectory follows from
\[
\frac{d\phi}{dN}=\sqrt{2\epsilon_H(N)}\,M_P=\sqrt{3\beta}\,M_P\,(N+1)^{-\alpha/2},
\]
so the parametrization is equivalent to a reconstruction of \(H(\phi)\) and \(V(\phi)\) rather than a genuinely model-free description [2507.05648].

To first order in slow roll, the central observables are
\[
n_s \simeq 1 - 3(1+\omega) + \frac{d}{dN}\ln(1+\omega),
\qquad
r \simeq 24(1+\omega),
\]
hence
\[
n_s \simeq 1 - 3\frac{\beta}{(1+N)^\alpha} - \frac{\alpha}{1+N},
\qquad
r \simeq \frac{24\beta}{(1+N)^\alpha}.
\]
Eliminating \(\beta\) yields
\[
\frac{r}{8}\simeq (1-n_s)-\frac{\alpha}{1+N}.
\]
Within this parametrization, \(\alpha\) is primarily constrained by the scalar spectral index, whereas \(\beta\) is strongly tied to the tensor amplitude [2507.05648].

Reanalyses using Planck 2018 and the bound \(r<0.032\) find that the current observational values of \(n_s\) and \(r\) can efficiently constrain the parameters. One summary result is
\[
1.50<\alpha\leq2.20,
\]
with the precise allowed \(\beta\) depending on \(N\) and on the tensor bound [2412.16703]. A later ACT-DR6 study sharpened the comparison by combining ACT-DR6, Planck-2018, DESI-Y1, and forecasts from LiteBIRD and CMB-S4. In that analysis, Planck plus \(r<0.032\) gives for \(N=50\)
\[
1.5759\lesssim\alpha\lesssim1.8003,
\]
and for \(N=60\)
\[
1.8849\lesssim\alpha\lesssim2.1533,
\]
while ACT+Planck+DESI with \(r<0.032\) narrows the ranges to
\[
1.1373\lesssim\alpha\lesssim1.2801\quad (N=50),
\qquad
1.3603\lesssim\alpha\lesssim1.5311\quad (N=60)
\]
[2507.05648].

A standard misconception in this literature is the label “model independent.” The Hamilton–Jacobi reconstruction makes explicit that choosing \(\omega(N)\) is equivalent to choosing a corresponding \(H(\phi)\) and \(V(\phi)\). The practical merit of the parametrization is not model independence in a strict sense but a compact phenomenological classification of inflationary backgrounds.

## 4. Canonical gauge invariants, loop quantum cosmology, and quantum backgrounds

A third major line identifies the Mukhanov variable as the unique gauge-invariant scalar degree of freedom within canonical perturbation theory. In a Hamiltonian treatment of scalar perturbations of FLRW spacetime, one introduces canonical transformations so that the first-order Hamiltonian and diffeomorphism constraints act trivially on a new variable \(Q\). The resulting gauge-invariant combination is
\[
Q=B_0\left(\delta\varphi+\frac{\varphi'}{\mathcal H}\psi\right),
\]
and the choice \(B_0=a(\eta)\) yields the canonical Mukhanov variable
\[
v=a\left(\delta\varphi+\frac{\varphi'}{\mathcal H}\psi\right)=z\mathcal R,
\qquad
z=\frac{a\varphi'}{\mathcal H}.
\]
Its dynamics is governed by
\[
v_k''+\left(k^2-\frac{z''}{z}\right)v_k=0
\]
in the classical limit [1111.7192].

Loop quantum cosmology modifies this structure while preserving its form. For inverse-volume corrections, the gauge-invariant variable becomes
\[
Q=\delta\varphi+\frac{\varphi'}{\mathcal H}\frac{1+f_1}{1+f}\,\psi,
\]
and the perturbations satisfy
\[
v_k''+\left(s^2k^2-\frac{z''}{z}\right)v_k=0,
\]
with a modified effective sound speed \(s^2\) and modified \(z\). For holonomy corrections,
\[
Q=\delta\varphi+\frac{K^{[2]}}{\mathcal H}\psi,
\qquad
z=\frac{a\varphi'}{K^{[2]}},
\]
and the equation again has Mukhanov–Sasaki form, now with the main modification entering the effective mass term through \(z''/z\) [1111.7192].

This canonical perspective was generalized further to arbitrary scalar potentials and arbitrary spacelike hypersurfaces without using the background classical equations of motion. After Faddeev–Jackiw reduction and canonical redefinitions, the scalar sector is written in terms of a single gauge-invariant variable \(\zeta\), with action
\[
S^{(2)}=\frac12\int d^4x\left[z^2\dot\zeta^2-z^2\zeta D^2\zeta\right],
\qquad
v\equiv z\zeta,
\]
so that the standard Mukhanov–Sasaki equation follows, but now in a form suitable for quantum backgrounds [1305.4664].

Hybrid quantum cosmology and hybrid LQC push this further by combining a quantized homogeneous geometry with a Fock quantization of the Mukhanov–Sasaki modes. In this setting the zero-mode Hamiltonian constraint is corrected by a quadratic perturbative contribution identifiable as the Mukhanov–Sasaki Hamiltonian, and a Born–Oppenheimer ansatz yields an approximate Schrödinger equation for perturbations whose coefficients are expectation values on the quantum background state [1503.03907]. Closely related work in hybrid LQC derives effective equations for the Mukhanov–Sasaki variables that include quantum contributions but retain the same ultraviolet limit as the classical equations; with alternate factor ordering, the resulting dynamics becomes similar to the dressed metric approach, apart from field scaling and quantization-prescription issues [1407.0998].

A further extension introduces Brown–Kuchař or Gaussian dust as reference fields and constructs perturbation theory directly on the reduced phase space of Dirac observables. In that framework the relational Mukhanov variable is
\[
Q=\delta\Phi+Z\left(\psi-\frac{\Delta}{3}E\right),
\qquad
Z=2\lambda_\varphi\frac{\overline\Pi_\Phi}{A\mathcal P},
\]
and its equation of motion takes the standard Mukhanov–Sasaki form plus dust-source terms. Those dust contributions disappear if the dust energy and momentum density as well as their perturbations vanish, and numerically they decay rapidly during inflation when the initial dust content is small [2003.13729].

## 5. Initial-state, numerical, and effective-mass parametrizations

Several papers use “Mukhanov parametrization” for the choice of initial state or for a controlled numerical or effective-mass reexpression of the Mukhanov–Sasaki dynamics. One line replaces the asymptotic Bunch–Davies prescription by finite-time boundary data. In a model with analytically solvable background, the Mukhanov potential \(f(\tau)=z''/z\) is fitted by
\[
f_{\rm fit}(\tau)=m(\tau-p)^2+h,
\]
and the mode equation
\[
v_k''+\big(k^2-f(\tau)\big)v_k=0
\]
is solved with boundary conditions imposed at a finite conformal time \(\tau=p\),
\[
v_k(\tau)\approx
a\,\frac{e^{-i\omega_*\tau}}{\sqrt{2\omega_*}}
+b\,\frac{e^{i\omega_*\tau}}{\sqrt{2\omega_*}},
\qquad |a|^2-|b|^2=1.
\]
In that sense, both the effective potential and the initial state are parametrized directly at the Mukhanov–Sasaki level [1012.4811].

A conceptually different proposal chooses the vacuum by minimizing the renormalized stress–energy tensor rather than diagonalizing the Hamiltonian. Writing the Mukhanov–Sasaki modes as \(\chi_k\), the resulting initial conditions are
\[
|\chi_k|^2=\frac{1}{2k},
\qquad
\chi_k'=\left(-ik+\frac{z'}{z}\right)\chi_k,
\]
to be compared with the Hamiltonian-diagonalizing conditions
\[
|\chi_k|^2=\frac{1}{2\omega_k},
\qquad
\chi_k'=-i\omega_k\chi_k,
\qquad
\omega_k^2=k^2-\frac{z''}{z}.
\]
This defines a distinct parametrization of the vacuum state inside the Mukhanov–Sasaki formalism, especially relevant in rapidly changing backgrounds such as a kinetically dominated universe [1607.04148].

The Lewis–Riesenfeld-invariant approach studies the Mukhanov–Sasaki Hamiltonian as an infinite collection of time-dependent oscillators and introduces the Ermakov equation
\[
\ddot\xi+\omega^2(t)\xi-\omega_0^2\xi^{-3}=0
\]
to construct an invariant and the associated time-dependent canonical transformation. The corresponding mode functions can be written in the polar form
\[
v_{\mathbf k}(\eta)=N_{\mathbf k}\,\xi_{\mathbf k}(\eta)\,
\exp\!\left\{-i\,\omega_{\mathbf k}^{(0)}
\int^\eta \frac{d\tau}{\xi_{\mathbf k}^2(\tau)}\right\}.
\]
A central result is that a solution of the Ermakov equation yields a full solution of the differential equation defining adiabatic vacua, without truncation at finite adiabatic order. The same analysis shows that not every finite-dimensional canonical transformation generalizes unitarily to Fock space; the Shale–Stinespring condition constrains the admissible Bogoliubov maps [1812.11122].

On the numerical side, the Mukhanov–Sasaki system can be recast in e-fold time using
\[
\mathcal S=z\sqrt{aH}\,\mathcal R,
\qquad
\mathcal Q=a\sqrt{aH}\,\mathcal T,
\]
so that the scalar and tensor equations become
\[
\frac{d^2\mathcal S_k}{dN^2}
+\left[\Omega_{\mathcal S}(N)+\frac{k^2}{a^2H^2}\right]\mathcal S_k=0,
\qquad
\frac{d^2\mathcal Q_k}{dN^2}
+\left[\Omega_{\mathcal Q}(N)+\frac{k^2}{a^2H^2}\right]\mathcal Q_k=0.
\]
Approximating the time-dependent frequency \(\omega_k^2(N)\) by linear or exponential segments allows the evolution to be computed by repeated \(2\times2\) matrix multiplication rather than direct ODE integration, with large speed gains at intermediate and high wavenumber [1809.11095].

In loop quantum cosmology, effective-mass ambiguities can themselves be parametrized. Starting from the classical comoving-gauge mass
\[
m_{\rm CG}^2=-\frac{z_s''}{z_s},
\qquad
z_s=a\frac{\dot\phi}{H},
\]
and polymerizing the inverse Hubble factor through a function \(f(\rho)\), one obtains a new family of effective masses containing four correction terms,
\[
\delta a,\ \delta b,\ \delta c,\ \delta d,
\]
whose explicit forms depend on \(f(\rho)\), \(f_{,\rho}\), and \(f_{,\rho\rho}\). These effective masses are distinct from the dressed metric and hybrid choices; notably, the \(\delta d\) term can remain important even for a kinetic-dominated bounce [2310.18408].

## 6. Black-hole and modified-gravity usages

A distinct black-hole usage identifies the “Mukhanov parametrization” with the Bekenstein–Mukhanov area spectrum
\[
A=aN\ell_p^2.
\]
In loop quantum gravity this arises not from a uniformly spaced microscopic area operator but from the thermodynamic large-area behavior of quantum isolated horizons. The microscopic spectrum is
\[
\hat A\,|\{s_j\}\rangle
=
8\pi\gamma \ell_p^2
\sum_j s_j\sqrt{j(j+1)}\,|\{s_j\}\rangle,
\]
while the most probable puncture distribution leads, for a macrostate \(|k,N)\), to an entropy
\[
S=N\frac{A_{\rm cl}}{8\pi\gamma\ell_p^2}+N\sigma,
\]
and an equation of state implying
\[
A_{\rm cl}=\tilde a\,\ell_p^2\,N.
\]
In this derivation the Bekenstein–Mukhanov integer \(N\) is not ad hoc: it is the total number of spin-network punctures on the horizon. With \(\sigma=0\), \(\lambda_0\approx1.2\), and \(\gamma=\lambda_0/(2\pi)\), the leading entropy matches the Bekenstein–Hawking law, and the macroscopic area levels inherit the exponential degeneracy required by the original Bekenstein–Mukhanov argument [1609.07125].

The same paper notes that transitions \(N\to N-1\) lead to discrete area changes and hence a line spectrum for black-hole radiation, but the underlying multiplicity of microscopic transitions causes line broadening. A conceptual gain over the original ansatz is therefore the physical identification of the integer label and its entropy-based degeneracy structure [1609.07125].

A further, differently named construction is the Markov–Mukhanov modification of the Einstein–Hilbert action,
\[
S=\int d^4x\sqrt{-g}\left(\frac{R}{8\pi G_N}+2\chi(\varepsilon)\mathcal L_m\right),
\]
where the coupling \(\chi(\varepsilon)\) depends only on the matter energy density. In this usage, the central parametrization is a series
\[
\chi(\varepsilon)=\sum_{n=0}^{\infty}C_n\left(\frac{\varepsilon}{\varepsilon_c}\right)^n,
\]
or specific choices such as
\[
\chi(\varepsilon)=1-\frac{\varepsilon}{\varepsilon_c}
\qquad\text{or}\qquad
\chi(\varepsilon)=\frac{1}{1+\varepsilon/\varepsilon_c}.
\]
The same function determines a running Newton constant,
\[
G(\varepsilon)=G_N(\chi\varepsilon)_{,\varepsilon},
\]
a running cosmological constant,
\[
\Lambda(\varepsilon)=-8\pi G_N \varepsilon^2\chi_{,\varepsilon},
\]
and an effective equation of state. In the model with
\[
\chi(\varepsilon)=\frac{1}{1+\varepsilon/\varepsilon_c},
\]
the background approaches an asymptotically de Sitter phase at high density, and the paper argues that viable inflation requires a bare dark-energy equation of state very close but not equal to \(-1\), in agreement with the DESI ranges quoted there [2510.14416].

Across these disparate usages, the unifying feature is methodological rather than semantic uniformity: “Mukhanov parametrization” typically denotes a compact parametrization of otherwise complicated gravitational or cosmological dynamics, whether through \(z''/z\), \(1+\omega(N)\), a black-hole area label \(N\), or an effective coupling \(\chi(\varepsilon)\). The specific object being parametrized, however, changes substantially from one subfield to another.

Source: https://www.emergentmind.com/topics/mukhanov-parametrization