Slow-Roll Inflation

1. Introduction — Why Slow-Roll?

The simplest physical picture of inflation is a ball rolling slowly down a gentle hillside. The "ball" is the inflaton field $\phi$, and the "hill" is the potential energy function $V(\phi)$. While the ball creeps along the nearly flat portion of the potential, the vacuum energy $V(\phi)$ dominates the energy budget of the universe. Because $V$ is approximately constant, the Friedmann equation gives a nearly constant Hubble parameter:

and the scale factor grows quasi-exponentially: $a(t) \propto e^{Ht}$. The essential requirement is that the inflaton's kinetic energy remain negligible compared to its potential energy throughout inflation. This demands two conditions:

• ●Flatness of the potential: The slope $V'(\phi)$ must be small enough that the field does not pick up too much kinetic energy.
• ●Persistence: The curvature $V''(\phi)$ must also be small so that the flatness condition is not rapidly violated — inflation must last long enough to solve the horizon and flatness problems (at least $N \sim 50{-}60$ e-folds).

The slow-roll approximation formalizes these requirements via dimensionless parameters. When they are much less than unity, the second-order Klein-Gordon equation becomes first-order and all observables are expressed directly in terms of the potential and its derivatives.

2. Derivation: Potential Slow-Roll Parameters

We start from the full equations governing an FRW universe with a single minimally-coupled scalar field:

The slow-roll approximation makes two assumptions:

Under SR1, the Friedmann equation simplifies to $H^2 \approx V/(3M_P^2)$. Under SR2, the Klein-Gordon equation becomes:

We can now solve for $\dot{\phi}$:

Defining $\epsilon_V$

Condition SR1 requires $\frac{1}{2}\dot{\phi}^2 \ll V$. Substituting the slow-roll expression for $\dot{\phi}$:

The condition $\frac{1}{2}\dot{\phi}^2 \ll V$ then becomes:

This motivates the definition of the first potential slow-roll parameter:

When $\epsilon_V \ll 1$, the kinetic energy is negligible compared to the potential energy, and inflation proceeds. Moreover, inflation ends when $\epsilon_V = 1$ — at this point the kinetic and potential energies become comparable, and the accelerated expansion ceases. More precisely, the condition for accelerated expansion ($\ddot{a} > 0$) is equivalent to $\epsilon_H < 1$, and in the slow-roll regime $\epsilon_H \approx \epsilon_V$.

Defining $\eta_V$

Condition SR2 requires $|\ddot{\phi}| \ll |3H\dot{\phi}|$. To see what this implies about the potential, differentiate the slow-roll equation $3H\dot{\phi} \approx -V'$with respect to time:

Dropping the $\ddot{\phi}$ term (by assumption SR2) and the $\dot{H}$ term (which is of the same order as $\epsilon_V$ corrections), we find:

Thus $|\ddot{\phi}| \ll |H\dot{\phi}|$ requires both $\epsilon_V \ll 1$ and:

The parameter $\eta_V$ measures the curvature of the potential relative to its height. While $\epsilon_V \ll 1$ ensures that inflation is occurring at a given moment,$|\eta_V| \ll 1$ ensures that the flatness condition persists — that is, inflation lasts long enough. If $|\eta_V|$ were of order unity, the potential would curve sharply and the field would quickly roll into a steep region, ending inflation prematurely.

3. Derivation: Hubble Slow-Roll Parameters

An alternative and in many ways more fundamental set of slow-roll parameters is defined directly in terms of the Hubble parameter and its time derivatives, without reference to a specific potential. These are called the Hubble slow-roll parameters (or Hubble flow parameters).

Defining $\epsilon_H$

The condition for inflation ($\ddot{a} > 0$) can be rewritten. From$a(t) = a_0 e^{\int H\, dt}$, we compute:

So $\ddot{a} > 0$ requires $\dot{H} + H^2 > 0$, i.e.,$-\dot{H}/H^2 < 1$. This motivates:

Inflation occurs when $\epsilon_H < 1$, and slow-roll inflation requires$\epsilon_H \ll 1$. Let us now express $\epsilon_H$ in terms of the inflaton field. From the Friedmann and Raychaudhuri equations:

The second equation uses $\dot{H} = -(1/2M_P^2)(\rho + p)$ with$\rho + p = \dot{\phi}^2$. Therefore:

This is exact — no approximation. It shows $\epsilon_H$ measures the ratio of kinetic to total energy.

Defining $\eta_H$

The second Hubble slow-roll parameter encodes the condition SR2 ($|\ddot{\phi}| \ll |3H\dot{\phi}|$):

When $|\eta_H| \ll 1$, the acceleration of the field is negligible compared to the Hubble friction term.

Relating Hubble and Potential Parameters

We now derive the connection between the two sets of parameters. In the slow-roll regime, using $H^2 \approx V/(3M_P^2)$ and $\dot{\phi} \approx -V'/(3H)$:

Substituting $H^2 \approx V/(3M_P^2)$ so that $H^4 \approx V^2/(9M_P^4)$:

For the second parameter, we differentiate $\dot{\phi} \approx -V'/(3H)$ with respect to time:

Dividing by $-H\dot{\phi}$:

Using $3H^2 \approx V/M_P^2$ for the first term and $\dot{H}/H^2 = -\epsilon_H \approx -\epsilon_V$ together with $V'/(3H\dot{\phi}) \approx -1$ for the second:

The two sets agree at leading order; the distinction matters at second order, relevant for next-to-leading corrections to $n_s$ and its running.

4. Derivation: Number of e-Folds

The number of e-folds measures how much the universe expands during inflation. It is defined as:

To express this as an integral over the field, we change variable from $t$to $\phi$. Since $dt = d\phi/\dot{\phi}$:

Note the sign: since the field rolls from large to small values (for typical potentials),$\dot{\phi} < 0$ and $\phi > \phi_\text{end}$, so we write with the convention that $N > 0$:

Applying the slow-roll approximation $3H\dot{\phi} \approx -V'$, so$H/|\dot{\phi}| \approx 3H^2/|V'| \approx V/(M_P^2|V'|)$:

This is the master formula for the number of e-folds in slow-roll inflation. Given a potential, one computes $\phi_\text{end}$ from $\epsilon_V(\phi_\text{end}) = 1$, and then evaluates the integral to find how many e-folds of inflation occur for a given initial field value.

Example 1: Quadratic Potential $V = \frac{1}{2}m^2\phi^2$

Here $V' = m^2\phi$, so $V/V' = \phi/2$. The slow-roll parameter is:

Inflation ends when $\epsilon_V = 1$, giving $\phi_\text{end} = \sqrt{2}\, M_P$. The e-fold integral is:

For large $\phi$ (i.e., $\phi \gg \phi_\text{end}$):

For $N = 60$, the field value at horizon exit is $\phi_* \approx 15.5\, M_P$, deep in the super-Planckian regime. This is the well-known Lyth bound implication for large-field models.

Example 2: Quartic Potential $V = \frac{\lambda}{4}\phi^4$

Here $V' = \lambda\phi^3$, so $V/V' = \phi/4$. The slow-roll parameter:

Inflation ends at $\phi_\text{end} = 2\sqrt{2}\,M_P$. The e-folds:

Example 3: Starobinsky Model $V = \Lambda^4\left(1 - e^{-\sqrt{2/3}\,\phi/M_P}\right)^2$

The Starobinsky potential arises from $R + R^2/(6M^2)$ gravity after a conformal transformation. Define $x = e^{-\sqrt{2/3}\,\phi/M_P}$ so $V = \Lambda^4(1-x)^2$. Then:

For $\phi \gg M_P$ (i.e., $x \ll 1$), $\epsilon_V \approx (4/3)x^2 \ll 1$, which is exponentially small. The e-fold integral gives:

So $x \approx 3/(4N)$, and $\epsilon_V \approx 3/(4N^2)$,$\eta_V \approx -1/N$. For $N = 55$ this gives $n_s \approx 1 - 2/N \approx 0.964$and $r \approx 12/N^2 \approx 0.004$ — in excellent agreement with Planck data.

5. Derivation: Attractor Behavior & Hamilton-Jacobi Formulation

A remarkable property of slow-roll inflation is that it is an attractor: regardless of initial conditions (provided the field starts in the slow-roll region), different trajectories converge to the same solution. The Hamilton-Jacobi formalism makes this manifest.

The Hamilton-Jacobi Equation

Instead of treating $\phi(t)$ as the dynamical variable, we use the inflaton field value as the time variable and express everything in terms of $H(\phi)$. From $\dot{H} = H'(\phi)\dot{\phi}$ and$\dot{H} = -\dot{\phi}^2/(2M_P^2)$, we obtain:

This is an exact first-order equation for $\dot{\phi}$, replacing the second-order Klein-Gordon equation. Substituting into the Friedmann equation$H^2 = (\dot{\phi}^2/2 + V)/(3M_P^2)$:

Rearranging gives the Hamilton-Jacobi equation:

Given $V(\phi)$, this is a first-order ODE for $H(\phi)$. Different "initial conditions" correspond to different integration constants, i.e., different amounts of kinetic energy at a given field value.

Proving the Attractor Property

Let $H_0(\phi)$ be a particular solution and consider a perturbation$H(\phi) = H_0(\phi) + \delta H(\phi)$. Substituting into the Hamilton-Jacobi equation and linearizing:

This gives:

Using $\dot{\phi} = -2M_P^2 H_0'$ and the number of e-folds$dN = -Hd\phi/\dot{\phi} = H_0/(2M_P^2 H_0')d\phi$:

In terms of the slow-roll parameter, the Hubble parameter in the Hamilton-Jacobi formalism gives:

This provides a clean parametrization: the slow-roll hierarchy is simply$\epsilon_H \ll 1$, $\eta_H \ll 1$, and higher-order flow parameters are small. The Hamilton-Jacobi approach forms the basis for the inflationary flow equations used in Monte Carlo reconstructions of the inflaton potential.

6. Applications

Computing Observables for Specific Models

The slow-roll parameters connect directly to CMB observables through the relations:

where $\xi_V^2 = M_P^4 V' V'''/ V^2$ is the third slow-roll parameter governing the running of the spectral index.

Flow Equations

The inflationary flow equations, introduced by Kinney (2002), extend the slow-roll hierarchy to infinite order. Define parameters $ {}^\ell\lambda_H$ via:

with $ {}^0\lambda_H = \epsilon_H$ and $ {}^1\lambda_H = \eta_H$. The flow equations form a coupled system of ODEs in the number of e-folds $N$. These can be integrated numerically to explore the space of inflationary models without committing to a specific potential.

Monte Carlo Reconstruction

By randomly sampling initial values of the flow parameters and evolving them, one can generate a large ensemble of inflationary models and map out the allowed region in the $(n_s, r)$ plane. This "Monte Carlo reconstruction" approach (Easther & Kinney, 2003) provides a model-independent way to test the slow-roll paradigm against CMB data. The results confirm that the observed values of$n_s$ and the upper bound on $r$ are naturally accommodated by the simplest slow-roll models, particularly plateau-type potentials like the Starobinsky model.

7. Historical Context

The slow-roll paradigm emerged from early attempts to fix the "graceful exit" problem of Guth's original inflationary model (1981), which relied on a first-order phase transition that produced an unacceptable inhomogeneous universe.

8. Python Simulation: Slow-Roll Analysis

The simulation below computes and compares the slow-roll parameters, e-fold counts, and field trajectories for three inflationary potentials: the quadratic model ($m^2\phi^2/2$), the Starobinsky model, and natural inflation. For each model it plots $\epsilon_V(\phi)$,$\eta_V(\phi)$, $N(\phi)$, and compares the slow-roll approximate trajectory against the exact numerical solution of the full Klein-Gordon + Friedmann system.