Chapter 6: Instanton Formalism for Rare Urban Transitions

6.1 Rare Events in Urban Systems

Most of the time, cities evolve gradually. But occasionally, dramatic transitions occur: a neighbourhood gentrifies seemingly overnight, an industrial district collapses, or a housing bubble bursts. These rare events are not well described by deterministic models. They require a stochastic framework that can compute the probability of unlikely but impactful transitions.

The instanton formalism, borrowed from quantum field theory and statistical mechanics, provides exactly this. An instanton is the most probable pathway for a rare transition between two metastable states—the “optimal fluctuation” that the system follows during the transition event.

6.2 The Bistable Urban Potential

Consider a scalar order parameter$\phi$ representing the state of a neighbourhood (e.g., average property value normalised between 0 and 1). The deterministic dynamics are governed by a double-well potential:

$$\frac{d\phi}{dt} = -\frac{\partial V}{\partial \phi} + \sqrt{2\varepsilon} \, \eta(t)$$

where $\eta(t)$ is Gaussian white noise with$\langle\eta(t)\eta(t')\rangle = \delta(t-t')$, and $\varepsilon \ll 1$ is the noise intensity. The potential has two minima:

$$V(\phi) = \frac{1}{4}(\phi^2 - 1)^2 = \frac{1}{4}\phi^4 - \frac{1}{2}\phi^2 + \frac{1}{4}$$

The minima at $\phi = \pm 1$ represent two stable urban states (e.g., “low-income residential” and “gentrified commercial”), separated by a barrier at $\phi = 0$ of height$\Delta V = 1/4$.

6.3 The Action Functional

In the Freidlin-Wentzell (large deviation) framework, the probability of a path$\phi(t)$ over the time interval$[0, T]$ is:

$$P[\phi] \propto \exp\left(-\frac{1}{2\varepsilon} S[\phi]\right)$$

where the action functional is:

$$S[\phi] = \int_0^T \mathcal{L}(\phi, \dot{\phi}) \, dt = \int_0^T \frac{1}{2}\left(\dot{\phi} + V'(\phi)\right)^2 dt$$

The Lagrangian is:

$$\mathcal{L}(\phi, \dot{\phi}) = \frac{1}{2}\left(\dot{\phi} + V'(\phi)\right)^2$$

Note: $\mathcal{L} = 0$ along the deterministic trajectory $\dot{\phi} = -V'(\phi)$, confirming that the most probable path in the absence of a transition is the deterministic one (zero action).

6.4 The Instanton: Saddle-Point of the Action

The instanton is the path that minimises the action$S[\phi]$ subject to the boundary conditions$\phi(0) = -1$ (initial state) and$\phi(T) = +1$ (final state). Setting$\delta S / \delta \phi = 0$ gives the Euler-Lagrange equation:

$$\ddot{\phi} = V''(\phi)\dot{\phi} + V'(\phi)V''(\phi) = \frac{d}{d\phi}\left[\frac{1}{2}(V'(\phi))^2\right] \cdot \frac{d\phi}{dt} / \frac{d\phi}{dt}$$

For the specific case of the double-well, the instanton satisfies the time-reversed deterministic equation:

$$\boxed{\dot{\phi}_{\text{inst}} = +V'(\phi) = \phi^3 - \phi}$$

This is the uphill equation—the system climbs the potential barrier. The instanton trajectory from $\phi = -1$ to$\phi = +1$ can be found analytically:

$$\phi_{\text{inst}}(t) = \tanh\left(\frac{t - t_0}{\sqrt{2}}\right)$$

The WKB Approximation

The action along the instanton gives the exponential suppression of the transition rate:

$$S_{\text{inst}} = \int_{-1}^{+1} |V'(\phi)| \, d\phi = \int_{-1}^{+1} |\phi^3 - \phi| \, d\phi = 2 \int_0^1 (\phi - \phi^3) \, d\phi = 2\left[\frac{\phi^2}{2} - \frac{\phi^4}{4}\right]_0^1 = \frac{1}{2}$$

The transition rate (Kramers rate) follows the WKB/Arrhenius form:

$$\boxed{\Gamma \sim \exp\left(-\frac{S_{\text{inst}}}{2\varepsilon}\right) = \exp\left(-\frac{1}{4\varepsilon}\right)}$$

For small noise ($\varepsilon \ll 1$), transitions are exponentially rare. The instanton is the optimal fluctuation that the system follows during the exponentially unlikely transition event.

6.5 Path Integrals and Large Deviation Theory

The transition probability can be expressed as a path integral:

$$P(\phi_f, T \,|\, \phi_i, 0) = \int_{\phi(0)=\phi_i}^{\phi(T)=\phi_f} \mathcal{D}\phi \, \exp\left(-\frac{S[\phi]}{2\varepsilon}\right)$$

In the limit $\varepsilon \to 0$, this integral is dominated by the saddle point (instanton), exactly as in the semiclassical limit of quantum mechanics. The connection to large deviation theory is:

$$\lim_{\varepsilon \to 0} (-2\varepsilon) \ln P(\phi_f, T \,|\, \phi_i, 0) = \inf_{\phi: \phi_i \to \phi_f} S[\phi] = S_{\text{inst}}$$

This is Varadhan’s lemma applied to the urban stochastic dynamics. The rate function$I = S_{\text{inst}} / 2$ controls the exponential decay of the transition probability.

Connection to JCP manuscript: The instanton formalism developed here connects directly to the path-integral approach to rare reactive events in chemical physics. The urban transition problem has the same mathematical structure as a chemical reaction crossing an activation barrier, with the noise intensity playing the role of temperature.

6.6 Python: Computing the Instanton Trajectory

We compute the instanton for the double-well potential, verify the action integral, and demonstrate the Kramers escape rate dependence on noise intensity.

Instanton: Bistable Urban Transition Pathway

Python

script.py99 lines

import numpy as np

# ── Instanton for V(phi) = (phi^2 - 1)^2 / 4 ──

print("=" * 62)
print("  Instanton Formalism: Bistable Urban Transition")
print("=" * 62)
print()

# Potential and its derivative
def V(phi):
    return 0.25 * (phi**2 - 1)**2

def Vprime(phi):
    return phi**3 - phi

# ── 1. Analytical instanton trajectory ──
t = np.linspace(-5, 5, 200)
t0 = 0  # center of transition
phi_inst = np.tanh((t - t0) / np.sqrt(2))

print("  Instanton trajectory phi(t) = tanh(t/sqrt(2)):")
print(f"  {'t':>8} {'phi':>10} {'V(phi)':>10} {'V_prime':>10}")
print("-" * 42)
for i in range(0, len(t), 20):
    print(f"  {t[i]:8.2f} {phi_inst[i]:10.4f} {V(phi_inst[i]):10.4f} {Vprime(phi_inst[i]):10.4f}")
print()

# ── 2. Verify action integral ──
# S = integral |V'(phi)|^2 dt along instanton
dphi_dt = Vprime(phi_inst)  # instanton satisfies dphi/dt = +V'(phi)
dt_val = t[1] - t[0]
S_numerical = np.trapezoid(dphi_dt**2, t)
S_analytical = 0.5  # from integral of |phi^3 - phi| dphi from -1 to 1

print(f"  Action integral:")
print(f"    S_numerical  = {S_numerical:.6f}")
print(f"    S_analytical = {S_analytical:.6f}")
print(f"    Agreement:     {abs(S_numerical - S_analytical)/S_analytical * 100:.3f}%")
print()

# ── 3. Kramers escape rate vs noise intensity ──
print("  Kramers escape rate: Gamma ~ exp(-S_inst / (2*epsilon))")
print(f"  {'epsilon':>10} {'S/(2*eps)':>12} {'log10(Gamma)':>14} {'Mean wait (yr)':>16}")
print("-" * 56)
epsilons = [0.5, 0.2, 0.1, 0.05, 0.02, 0.01, 0.005]
for eps in epsilons:
    exponent = S_analytical / (2 * eps)
    log_gamma = -exponent / np.log(10)  # log10
    mean_wait = np.exp(exponent)  # in natural time units
    wait_str = f"{mean_wait:.2e}" if mean_wait < 1e15 else ">10^15"
    print(f"  {eps:10.3f} {exponent:12.2f} {log_gamma:14.2f} {wait_str:>16}")
print()
print("  Small noise => exponentially long wait times")
print("  This explains why gentrification events are rare but sudden")
print()

# ── 4. Numerical instanton via shooting method ──
# Solve dphi/dt = V'(phi) from phi=-1+delta to phi=+1-delta
print("  Numerical instanton (RK4 shooting):")
phi_start = -0.999
dt_rk = 0.01
phi_traj = [phi_start]
t_traj = [0.0]

phi_n = phi_start
for _ in range(2000):
    k1 = dt_rk * Vprime(phi_n)
    k2 = dt_rk * Vprime(phi_n + k1/2)
    k3 = dt_rk * Vprime(phi_n + k2/2)
    k4 = dt_rk * Vprime(phi_n + k3)
    phi_n = phi_n + (k1 + 2*k2 + 2*k3 + k4) / 6
    phi_traj.append(phi_n)
    t_traj.append(t_traj[-1] + dt_rk)
    if phi_n > 0.999:
        break

phi_traj = np.array(phi_traj)
t_traj = np.array(t_traj)

print(f"  Transition time: {t_traj[-1]:.2f} (natural units)")
print(f"  Start: phi = {phi_traj[0]:.4f}, End: phi = {phi_traj[-1]:.4f}")
print()

# Action along numerical trajectory
Vp = np.array([Vprime(p) for p in phi_traj])
S_num2 = np.trapezoid(Vp**2, t_traj)
print(f"  Action along numerical instanton: S = {S_num2:.6f}")
print(f"  Exact: S = 0.500000")

# Profile near the barrier
print()
print("  Instanton near barrier top (phi ~ 0):")
mask = np.abs(phi_traj) < 0.3
t_near = t_traj[mask]
phi_near = phi_traj[mask]
for i in range(0, len(t_near), max(1, len(t_near)//8)):
    bar_left = ' ' * int(20 + phi_near[i] * 18)
    print(f"  t={t_near[i]:6.2f}  phi={phi_near[i]:+.4f}  |{bar_left}*")

Click Run to execute the Python code

Code will be executed with Python 3 on the server

6.7 Summary & Key Takeaways

• Rare urban transitions (gentrification, collapse) are modelled as noise-driven escapes from metastable states
• The action functional $S[\phi] = \int \frac{1}{2}(\dot\phi + V')^2 dt$ measures path improbability
• The instanton (most probable transition path) satisfies the time-reversed deterministic equation
• Transition rate: $\Gamma \sim e^{-S_{\text{inst}}/(2\varepsilon)}$ (Kramers/WKB)
• Connects to path integrals, large deviation theory, and the JCP formalism for reactive trajectories

← Fisher-KPP Waves Next: Stochastic CA →