Energy landscapes as nested containers. Why errors are exponentially suppressed.
The container metaphor has a direct physical realization. A potential energy well is a container. The barrier height $\Delta E$ is the energy required to cross the boundary. A landscape of nested wells — large wells containing smaller wells containing even smaller wells — is a hierarchical energy landscape.
At the root ($k=0$), the barrier is $E_0$ — the largest. At depth $k$, barriers are $b^{-\alpha k}$ times smaller. Deep levels have tiny barriers (the system moves freely). Shallow levels have large barriers (the system is strongly confined).
Store a value as a point on a line — water column height, capacitor voltage, needle angle. Perturbations displace the point continuously. Error accumulates with time. Constant measurement and correction required. Energy cost scales with precision.
Store a value as the identity of a container at depth $D$. Perturbations below barrier $\Delta E_D$ cause jitter within the container but cannot change the logical state. Only above-threshold perturbations can cause errors — and these are exponentially rare. Energy cost is near-zero during idle.
For a system with hierarchical energy landscape $\Delta E_k = E_0 \cdot b^{-\alpha k}$, logical information encoded at depth $D$, and noise energy $\varepsilon$ (with $\varepsilon < \Delta E_D$):
$$P(\text{error}) \leq C \cdot \exp\!\left(-\left(\frac{\Delta E_D}{\varepsilon}\right)^{\beta}\right)$$where $C$ is an attempt frequency (typically $10^9$–$10^{13}$ Hz), and $\beta$ is:
The error rate is exponentially suppressed in $\Delta E_D / \varepsilon$. Double the ratio, square the error rate. Increase $E_0$ (larger barriers) or decrease $\varepsilon$ (lower temperature), and the protection improves dramatically.
A classical particle in a well of depth $\Delta E$, in contact with a heat bath at temperature $T$, has probability $\propto e^{-\Delta E / k_B T}$ of acquiring enough thermal energy to escape (Boltzmann factor). The escape rate multiplies this by an attempt frequency $\nu_0 \sim 10^{12}$ Hz:
$$\Gamma_{\text{escape}} \approx \nu_0 \cdot e^{-\Delta E / k_B T}$$Quantum tunneling provides an alternative path, with rate $\propto \exp(-\text{const} \cdot \sqrt{\Delta E})$ for parabolic barriers. In either case, the suppression is exponential in the barrier height.
Key difference from continuous: In a flat potential, the particle drifts freely — error grows as $\sqrt{t}$. In a well, it jiggles but stays — error is threshold-suppressed. The difference is a floor vs. a bowl.
With $E_0 = 100 \cdot k_B T$, $\alpha = 1$, $b = 2$, $C = 10^{12}$ Hz, $\beta = 1$:
| Depth $D$ | Barrier ($k_B T$) | Error Rate (s⁻¹) | Mean Time Between Errors |
|---|---|---|---|
| $0$ | $100$ | $\sim 10^{-32}$ | $\sim 10^{24}$ years |
| $1$ | $50$ | $\sim 10^{-10}$ | $\sim 170$ years |
| $2$ | $25$ | $\sim 10^{1}$ | $\sim 0.07$ seconds |
| $3$ | $12.5$ | $\sim 10^{6}$ | $\sim 0.27 \; \mu\text{s}$ |
Shallower encoding (smaller $D$) gives stronger protection. At $D = 2$, the barrier is only $25 \, k_B T$ — errors are frequent. At $D = 1$, errors are astronomically rare. The crossover is where $\Delta E_D \approx 20 \, k_B T$.