Thin-Wall Control and Beyond-Thin-Wall Corrections

The thin-wall approximation is controlled when the bubble radius is much larger than every scale over which the wall profile changes and when the vacuum-energy splitting is small compared with the barrier scale that fixes the wall. In that regime the bounce action separates into a positive surface term and a negative volume term. The approximation must be tested against the full radial solution; near-degeneracy by itself is not a substitute for that comparison.

Required background. Bounce solutions and false-vacuum boundary conditions supplies the radial action and the tilted double-well example. Effective field theory as a controlled expansion supplies the logic of an expansion parameter and a declared error order. EFT truncation errors and breakdown diagnostics supplies residual and convergence tests.

Helpful background. Bounce existence, symmetry, and multifield geometry explains why a wall trajectory chosen in a multifield potential must satisfy the full normal-force equation.

Surface tension and scale separation

Let $U_0(\phi)$ have degenerate minima $\phi_{\mathrm t}$ and $\phi_{\mathrm f}$ with common value $U_0^{\min}$. The planar wall satisfies

$$\frac{\mathrm d^2\phi_0}{\mathrm d\rho^2} = U_0'(\phi_0), \qquad \phi_0(-\infty)=\phi_{\mathrm t}, \qquad \phi_0(+\infty)=\phi_{\mathrm f}.$$

Multiplying by $\mathrm d\phi_0/\mathrm d\rho$ and using the asymptotic conditions gives

$$\frac12 \left(\frac{\mathrm d\phi_0}{\mathrm d\rho}\right)^2 = U_0(\phi_0)-U_0^{\min}.$$

The tension—the action per unit $(d-1)$-area—is therefore

$$\begin{aligned} \sigma_0 &= \int_{-\infty}^{\infty}\mathrm d\rho \left[ \frac12\left(\frac{\mathrm d\phi_0}{\mathrm d\rho}\right)^2 +U_0(\phi_0)-U_0^{\min} \right]\\ &= \int_{\phi_{\mathrm t}}^{\phi_{\mathrm f}} \mathrm d\phi\, \sqrt{2\left[U_0(\phi)-U_0^{\min}\right]}. \end{aligned}$$

For several canonical fields, the analogous expression is minimized over admissible paths $\gamma$ joining the degenerate vacua,

$$\sigma_0[\gamma] = \int_\gamma \mathrm ds\, \sqrt{2\left[U_0(\boldsymbol\phi)-U_0^{\min}\right]}.$$

This formula does not authorize an arbitrary one-dimensional path: the minimizing wall path and the curved finite-radius bounce must still be stationary in directions normal to $\gamma$.

Let $\ell$ denote a wall thickness extracted from the planar profile—for example, the inverse exponential scale or a fixed percentile width—and let

$$\epsilon = U(\phi_{\mathrm f})-U(\phi_{\mathrm t})>0.$$

Two independent small quantities commonly enter:

$$\eta_R=\frac{\ell}{R}, \qquad \eta_U= \frac{ \displaystyle\sup_{\text{wall}} \left\lvert [U(\phi)-U(\phi_{\mathrm f})] -[U_0(\phi)-U_0(\phi_{\mathrm f})] \right\rvert } {\text{barrier height of }U_0}.$$

Subtracting the false-vacuum value makes $\eta_U$ independent of arbitrary additive constants in either potential. Both parameters must be small. Their precise definitions and numerical thresholds belong to the calculation; a claim of thin-wall control should state them rather than only saying that the vacua are “nearly degenerate.”

Critical radius and leading action

Use the convention

$$\Omega_{d-1} = \frac{2\pi^{d/2}}{\Gamma(d/2)}$$

for the area of the unit $(d-1)$-sphere. A bubble of true vacuum with radius $R$ has surface area $\Omega_{d-1}R^{d-1}$ and enclosed $d$-volume $\Omega_{d-1}R^d/d$. At leading thin-wall order,

$$B_{\mathrm{tw}}(R) = \Omega_{d-1}\sigma_0R^{d-1} -\frac{\Omega_{d-1}}{d}\epsilon R^d.$$

The stationary radius follows from

$$\frac{\mathrm dB_{\mathrm{tw}}}{\mathrm dR} = \Omega_{d-1}R^{d-2} \left[(d-1)\sigma_0-\epsilon R\right] =0:$$ $$R_{\mathrm{tw}} = \frac{(d-1)\sigma_0}{\epsilon}.$$

Substitution gives

$$B_{\mathrm{tw}} = \frac{\Omega_{d-1}}{d} (d-1)^{d-1} \frac{\sigma_0^d}{\epsilon^{d-1}}.$$

The dimensions provide an immediate check. In $d$ spacetime dimensions, $[\sigma_0]=\text{mass}^{d-1}$ and $[\epsilon]=\text{mass}^d$, so $[R_{\mathrm{tw}}]=\text{mass}^{-1}$ and $B_{\mathrm{tw}}$ is dimensionless. For $d=4$, $\Omega_3=2\pi^2$ and

$$R_{\mathrm{tw}}=\frac{3\sigma_0}{\epsilon}, \qquad B_{\mathrm{tw}} = \frac{27\pi^2\sigma_0^4}{2\epsilon^3}.$$

These are stationary-bubble formulas, not a real-time critical-bubble evolution law. Their derivation and regime are the flat, zero-temperature Euclidean bounce Coleman 1977, §IV, pp. 2932–2934.

First application: the tilted quartic

For

$$U(\phi) = \frac{\lambda}{4}(\phi^2-v^2)^2 +\frac{3\epsilon}{4} \left(\frac{\phi}{v}-\frac{\phi^3}{3v^3}\right),$$

the two stationary vacua remain at $\phi_{\mathrm t}=-v$ and $\phi_{\mathrm f}=+v$, with energy difference $\epsilon$. The degenerate wall data are

$$\phi_0(\rho) = v\tanh\!\left(\sqrt{\frac{\lambda}{2}}\,v\rho\right), \qquad \ell := \frac{\sqrt2}{\sqrt\lambda\,v}, \qquad \sigma_0 = \frac{2\sqrt2}{3}\sqrt\lambda\,v^3.$$

In four dimensions,

$$R_{\mathrm{tw}} = \frac{2\sqrt2\,\sqrt\lambda\,v^3}{\epsilon}, \qquad \frac{\ell}{R_{\mathrm{tw}}} = \frac{\epsilon}{2\lambda v^4},$$

and

$$B_{\mathrm{tw}} = \frac{32\pi^2}{3} \frac{\lambda^2v^{12}}{\epsilon^3}.$$

Thus the false vacuum can be locally stable throughout $0<\epsilon/(\lambda v^4)<4/3$, while the thin-wall approximation requires the much stronger hierarchy

$$\frac{\epsilon}{\lambda v^4}\ll1.$$

Local metastability and thin-wall control are different statements.

To test the approximation, solve the full radial equation for a sequence of decreasing tilts. For each solution record

$$\Delta_R = \frac{R_{\mathrm{num}}-R_{\mathrm{tw}}}{R_{\mathrm{num}}}, \qquad \Delta_B = \frac{B_{\mathrm{num}}-B_{\mathrm{tw}}}{B_{\mathrm{num}}},$$

where the definition of $R_{\mathrm{num}}$—for example the radius of maximum wall energy density—must be fixed across the sequence. Also record the boundary residual, virial residual, and changes under radial-grid and outer-boundary refinement. Thin-wall control is demonstrated when $\eta_R$, $\eta_U$, $|\Delta_R|$, and $|\Delta_B|$ all decrease in the near-degenerate sequence while the numerical residuals remain smaller than the claimed truncation error.

Here is that comparison for $\lambda=v=1$ in $d=4$. The leading estimate obeys $\ell/R_{\mathrm{tw}}=\epsilon/2$, but the table reports $\eta_R=\ell/R_{\mathrm{num}}$ using the numerical radius. With the potentials aligned at the false vacuum, the tilt deformation reaches $\epsilon$ across $-v\leq\phi\leq v$ while the degenerate barrier height is $1/4$, so $\eta_U=4\epsilon$. In the table, $R_{\mathrm{num}}$ is the unique outward zero crossing, $\phi_b(R_{\mathrm{num}})=0$. The dependency-free reproduction script uses damped Newton relaxation on second-order radial finite differences, solves at $h=0.02$, $0.01$, and $0.005$, and Richardson-extrapolates the radius and action. Its finest-grid maximum discrete equation residual is below $2.5\times10^{-11}$; the relative radial virial residual is below $1.4\times10^{-5}$; the observed action-convergence order is $2.000$; and the extrapolation correction is below $6.7\times10^{-6}$. Increasing the outer tail from $20$ to $24$ leaves both the radius and action unchanged at the displayed precision. Adaptive collocation and shooting provide useful independent formulations of the same boundary problem Devoto et al. 2022, Appendix A.6, pp. 96–98.

From the repository root, reproduce the table and its self-checks with node scripts/benchmark-vacuum-decay-thin-wall.mjs.

$\epsilon$	$\eta_R$	$\eta_U$	$R_{\mathrm{tw}}$	$R_{\mathrm{num}}$	$\Delta_R$	$B_{\mathrm{tw}}$	$B_{\mathrm{num}}$	$\Delta_B$
$0.30$	$0.152$	$1.200$	$9.428$	$9.287$	$-1.5198\%$	$3899.103$	$3728.296$	$-4.5814\%$
$0.20$	$0.101$	$0.800$	$14.142$	$14.050$	$-0.6581\%$	$13159.473$	$12904.162$	$-1.9785\%$
$0.10$	$0.050$	$0.400$	$28.284$	$28.239$	$-0.1620\%$	$105275.780$	$104766.214$	$-0.4864\%$

The monotone decrease of $\eta_R$, $\eta_U$, $\lvert\Delta_R\rvert$, and $\lvert\Delta_B\rvert$ supports the approach to the thin-wall regime. The displayed $\eta_U=1.200$, $0.800$, and $0.400$ also show that these are convergence benchmarks, not deeply controlled thin-wall points. At $\epsilon=0.10$, a relative exponent error below one percent is still an absolute error $B_{\mathrm{tw}}-B_{\mathrm{num}}\simeq509.6$. Because the rate is exponential, that approximation is inadequate for any claim requiring order-one multiplicative accuracy. The required accuracy must be set on $\ln(\Gamma/V)$, not only on $B$ as a percentage.

Where corrections come from

Write $\rho=r-R$ in the wall region and let $w_0(\rho)$ be the planar wall action density. Expanding the radial measure gives

$$\begin{aligned} \Omega_{d-1}\int \mathrm d\rho\,(R+\rho)^{d-1}w_0(\rho) = \Omega_{d-1}R^{d-1} \bigg[ \sigma_0 +\frac{d-1}{R}\mu_1\\ \qquad +\frac{(d-1)(d-2)}{2R^2}\mu_2 +\cdots \bigg], \end{aligned}$$

with wall moments

$$\mu_n=\int\mathrm d\rho\,\rho^n w_0(\rho).$$

Shifting the convention for $R$ changes $\mu_1$. A surface-of-tension convention sets $\mu_1=0$; for a reflection-symmetric planar wall centered at $\rho=0$, it vanishes automatically. Geometric corrections then begin with a term of relative order $(\ell/R)^2$. A deformed potential or an asymmetric wall can independently change the tension and profile at order $\eta_U$. Consequently, there is no universal claim that the first correction is always linear or always quadratic in one chosen thin-wall parameter.

A systematic calculation expands

$$\phi(r) = \phi_0(\rho)+\phi_1(\rho)+\cdots, \qquad R=R_{\mathrm{tw}}+\delta R+\cdots,$$

and solves the wall fluctuation equation order by order. Its translational zero mode $\phi_0'(\rho)$ imposes a solvability condition; that condition fixes the radius correction. The same renormalization prescription and perturbative order must be used for $\epsilon$, $\sigma$, the corrected profile, and the determinant prefactor.

Breakdown diagnostics

Use the full bounce rather than the leading formula when any of the following occurs:

• $\ell/R$ is not small, so curvature varies appreciably across the wall;
• the tilt changes the barrier shape or wall tension by an order-one amount;
• the bounce center is not exponentially close to the true-vacuum basin;
• two wall paths or bounce branches have comparable actions;
• the radial residual or virial identity fails at the claimed error level;
• the leading action is sensitive to the chosen definition of wall position;
• higher-derivative operators become important at gradients of order $1/\ell$;
• temperature or spacetime curvature competes with $R^{-1}$.

The last two failures are not repaired by adding more terms to the canonical flat-space wall expansion.

Shared calculation. The bounce control map places thin wall as an optional approximation after the boundary-value and spectrum checks, not as an independent definition of the rate.

Shared comparison. The instanton–bounce boundary and mode comparison shows why the thin-wall bubble still uses return-to-false-vacuum bounce boundary conditions.

Common pitfalls

Using the biased potential in the degenerate tension integral without a prescription. Once the vacua have unequal energies, the planar integral includes an infinite bulk contribution. Define a degenerate reference potential or subtract the appropriate bulk pieces before assigning a wall tension.

Equating a small energy splitting with a small total error. Thin-wall control also requires a stable wall path, $\ell/R\ll1$, controlled gradients, and convergence of the full bounce and fluctuation calculation.

Comparing only exponents at one parameter point. A single agreement can be accidental. Vary the tilt and numerical resolution, and test the predicted approach to the degenerate limit.

Exercises

1. Derive $R_{\mathrm{tw}}$ and $B_{\mathrm{tw}}$ in general $d$, and verify their mass dimensions.

Solution

Differentiating

$$B(R)=\Omega_{d-1}\sigma_0R^{d-1} -\frac{\Omega_{d-1}}{d}\epsilon R^d$$

gives $R=(d-1)\sigma_0/\epsilon$. Substitution yields

$$B_{\mathrm{tw}} = \frac{\Omega_{d-1}}{d}(d-1)^{d-1} \frac{\sigma_0^d}{\epsilon^{d-1}}.$$

Because $[\sigma_0]=M^{d-1}$ and $[\epsilon]=M^d$, the radius has dimension $M^{-1}$ and the action is dimensionless.

2. For the tilted quartic, verify $\ell/R_{\mathrm{tw}}=\epsilon/(2\lambda v^4)$ in $d=4$.

Solution

Using

$$\ell=\frac{\sqrt2}{\sqrt\lambda\,v}, \qquad R_{\mathrm{tw}} = \frac{3}{\epsilon} \left(\frac{2\sqrt2}{3}\sqrt\lambda\,v^3\right) = \frac{2\sqrt2\sqrt\lambda\,v^3}{\epsilon},$$

their ratio is

$$\frac{\ell}{R_{\mathrm{tw}}} = \frac{\epsilon}{2\lambda v^4}.$$

3. Explain why the first geometric correction vanishes for a reflection-symmetric wall centered at $\rho=0$.

Solution

For a reflection-symmetric wall, $w_0(\rho)=w_0(-\rho)$. The moment

$$\mu_1=\int\mathrm d\rho\,\rho\,w_0(\rho)$$

has an odd integrand and vanishes. The next measure correction is proportional to $\mu_2/R^2$, hence is relatively of order $(\ell/R)^2$. Potential-deformation corrections need not share that order.

References

• Coleman, S. (1977). “The Fate of the False Vacuum. I. Semiclassical Theory.” Physical Review D 15, 2929–2936. doi:10.1103/PhysRevD.15.2929.
• Devoto, F., Devoto, S., Di Luzio, L., and Ridolfi, G. (2022). “False Vacuum Decay: An Introductory Review.” Journal of Physics G: Nuclear and Particle Physics 49, 103001. doi:10.1088/1361-6471/ac7f24. Open PDF.