Beyond the Dilute Instanton Gas: Resurgence with Exact Saddles in the Double Well

1. Problem Statement

The paper addresses fundamental limitations in the standard Dilute Instanton Gas (DIG) approximation used in the path-integral approach to quantum mechanics, specifically for the symmetric double-well potential $V(x) = \frac{1}{8}(x^2 - 1)^2$.

• Limitations of DIG:
  • Infinite Temperature Limit: The DIG relies on taking the Euclidean time period $T \to \infty$. This discards all finite-$T$ corrections, meaning it can only extract the ground-state energy splitting. It fails to systematically compute energy splittings for excited states or their dependence on the level number $N$.
  • Lack of Systematic Corrections: The DIG constructs multi-instanton configurations by "sewing" together well-separated instanton-anti-instanton pairs (tanh kinks). This ignores interactions and provides only leading-order contributions, lacking a framework for subleading corrections.
  • Ad Hoc Prescriptions: The standard method (Bogomolny-Zinn-Justin or BZJ) handles quasi-zero modes (associated with instanton separation) via an ad hoc analytic continuation of the coupling $g \to -g$. While this produces the correct cancellation of Borel ambiguities (resurgence), it is not derived from a principled decomposition of the path integral.

The authors aim to bypass the DIG approximation entirely by utilizing exact finite-$T$ saddle points to derive a rigorous, systematic resurgent structure for all energy levels.

2. Methodology

The authors replace the approximate instanton gas with exact solutions to the Euclidean equations of motion at finite temperature $T$.

• Exact Saddles: The classical solutions are parameterized by the conserved energy $\varepsilon$ and expressed using Weierstrass elliptic functions ($\wp$). The boundary conditions impose a quantization condition on $\varepsilon$ involving the half-periods of the associated elliptic curve:
  $$T = 2k \omega_N(\varepsilon) + 2k' \omega_P(\varepsilon)$$
  where $k$ is the instanton number and $k'$ counts windings around the imaginary period. Real saddles correspond to $k'=0$.
• Picard–Lefschetz Decomposition: The partition function $Z$ is decomposed into a sum over steepest-descent thimbles ($J_{k,k'}$) associated with these exact saddles. The authors prove that for the real partition function, only real saddles ($k'=0$) contribute directly to the decomposition; complex saddles govern the Stokes structure indirectly but do not enter the final sum.
• Fluctuation Analysis:
  • The one-loop determinant around these saddles is computed exactly using Lamé operators (arising from the fluctuation potential $V''(x^*)$).
  • The periods of the elliptic curve satisfy Picard–Fuchs equations, allowing higher-order derivatives of the action to be expressed as linear combinations of the action and its first derivative.
• Quasi-Zero Mode Integration: The integration over the quasi-zero modes (instanton separations) is performed as a finite-dimensional contour integral over the effective action. This replaces the ad hoc BZJ prescription with a geometric interpretation of the thimble decomposition.

3. Key Contributions and Results

A. Systematic Computation of Energy Splittings

Unlike the DIG, which yields a uniform splitting for all levels, this approach captures the full dependence on the quantum number $N$.

• By analyzing the twisted partition function (with anti-periodic boundary conditions), the authors derive the non-perturbative splitting $\Delta_N$ for all levels.
• Result: The leading non-perturbative splitting is found to be:
  $$\Delta_N = -\frac{\hbar}{\sqrt{2\pi}} \frac{(8/\hbar)^{\kappa}}{\Gamma(\kappa + 1/2)}, \quad \text{where } \kappa = N + 1/2$$
  This result matches the Exact WKB calculation perfectly. The DIG fails here because it discards $e^{-T}$ corrections in the action, leading to an incorrect factor of $(4/\hbar)^N$ instead of $(8/\hbar)^N$.

B. Resolution of Resurgence and Ambiguities

The paper provides a rigorous geometric derivation of how ambiguities cancel, replacing the BZJ prescription.

• Thimble Geometry: For the $n=2$ (one instanton-anti-instanton) sector, the integration contour over the separation mode $\alpha$ is decomposed into three parts: a real segment, a vertical segment connecting to a complex saddle, and an arm extending to infinity.
• Cancellation Mechanism:
  • The real segment generates an asymptotic series (Borel singularity) corresponding to perturbative vacuum bubbles.
  • The vertical segment and arm generate imaginary contributions.
  • The sum of these thimbles yields a result where the imaginary part (ambiguity) from the lateral Borel resummation of the perturbative series is exactly canceled by the imaginary part arising from the instanton sector.
• Multi-Instanton Structure: For $n=4$, the authors identify a "2:1 channel structure" (two separation modes vs. one breathing mode). They verify a three-way ambiguity cancellation relation:
  $$\frac{1}{2} \text{Im}[\Delta_4 Z_0] + \frac{1}{2} \text{Im}[\Delta_2 Z_2] + \text{Im}[Z_4] = 0$$
  This confirms the consistency of the trans-series structure across different instanton numbers.

C. Consistency Checks via Logarithmic Terms

The authors perform a stringent consistency check involving logarithmic terms ($\ln u$, where $u=e^{-T}$) appearing in the expansion of the twisted partition function.

• These logarithmic terms arise from three distinct sources:
  1. The one-loop functional determinant (Lamé spectrum).
  2. The tree-level exact action (finite-$T$ corrections to the instanton profile).
  3. Two-loop perturbative vacuum bubbles.
• The paper demonstrates that these terms, derived from completely different loop orders and mathematical structures, cancel exactly when matched against the spectral decomposition. This validates the exact finite-$T$ computation.

4. Significance and Implications

• Beyond Approximation: The work demonstrates that the "dilute gas" is an unnecessary approximation. By keeping $T$ finite and using exact elliptic solutions, one can systematically compute non-perturbative effects for all excited states, not just the ground state.
• Geometric Resurgence: It establishes a clear geometric link between Picard–Lefschetz theory (thimble decomposition) and resurgent trans-series. The cancellation of ambiguities is no longer an ad hoc trick but a direct consequence of the topology of the integration contours in the complex plane.
• Mathematical Framework: The synthesis of Weierstrass elliptic functions (for saddles), Lamé operators (for fluctuations), and Picard–Fuchs equations (for periods) creates a coherent mathematical toolkit that complements Exact WKB methods.
• Future Applications: The authors argue that this framework is essential for Quantum Chromodynamics (QCD). In QCD, the dilute instanton gas is unreliable due to renormalon effects and the lack of stable multi-instanton solutions at infinite volume. The paper suggests that compactifying QCD on $\mathbb{R}^3 \times S^1$ (finite volume) would allow for exact multi-instanton saddles, enabling a rigorous resurgent analysis of the QCD path integral similar to the double-well model.

Conclusion

This paper successfully upgrades the path-integral approach to quantum mechanics by replacing the dilute instanton gas with exact finite-$T$ saddles. It provides a rigorous, geometric derivation of resurgence, correctly reproduces the energy spectrum for all levels (matching Exact WKB), and resolves the ambiguity cancellation mechanism without ad hoc prescriptions. This sets a new standard for analyzing non-perturbative physics in quantum theories.