1.2.2. Kardar-Parisi-Zhang (KPZ) equation (7) - v0 can be absorbed. - Scaling exponents: 1d: α = 1/2, β = 1/3, z = 3/2 (shown below) 2d: α ≈ 0.39, β ≈ 0.24, z ≈ 1.6 Note: naive calculation done below Eq.(4) gives wrong results here. Coefficients can also be scale-dependent. 3 1.2.3. Quenched KPZ equation If noise comes from spatial heterogeneity, one should use the quenched KPZ eq. (8) - F cannot be absorbed (because noise depends on h) - Pinning-depinning transition F > Fc: interface grows, KPZ scaling (η(x,h) is equivalent to η(x,t)) F < Fc: interface pinned F ≈ Fc: critical scaling α = β ≈ 0.63 “quenched KPZ class” [homework (advanced)] with exponents for directed percolation class. In fact, Read [4] and understand the relation to directed percolation. 1.2.4. Conserved growth Consider particle deposition. If (time scale of particle motion) << (time scale of deposition) then the particle number is effectively conserved. Mullins-Herring (MH) equation (9) (MH class) [homework] Derive these exponents. 4 In practice, nonlinearity often plays role, then we obtain: (10) Confirmed in numerics and some experiments. Molecular beam epitaxy (MBE) class. 2. Basics on KPZ equation In the following, we set v0=0. 2.1. Relation to noisy Burgers equation (11) This is the noisy Burgers equation (if λ=1). Galilei symmetry: invariant under For KPZ, invariant under (statistical tilt symmetry) Galilei symmetry remains valid under scale transformation (4) λ is scale-invariant, (12) [homework] Show Eq.(12) from Eqs.(4), (7), and scale invariance of λ. 5 2.2. Stationary KPZ interface (1d) Fokker-Planck equation for KPZ eq. (13) For 1d (only), stationary solution (5) for the EW eq., (Brownian motion) is also the stationary solution for the KPZ eq. [homework] Show this claim. ∴ α = 1/2. Using Eq.(12), z = 3/2, β = 1/3. 2.3. EW KPZ crossover Starting from flat initial condition h(x,0)=0, is small at short times, so KPZ eq. behaves like EW eq. 2.4. Well-definedness of KPZ equation In fact, KPZ equation is ill-defined. ( is incompatible with white noise) Using the Cole-Hopf transformation (14) 6 KPZ eq. (7) can be formally rewritten as: (stochastic heat equation, SHE) (15) with (16) Eq.(15) is a linear equation with multiplicative noise. Ito or Stratonovich product? Transformation from Eqs.(14) to (15) is valid for Stratonovich product, but then with By contrast, SHE with Ito product has a well-defined solution Z(x,t). Then, h(x,t) obtained by Eq.(14) is called the “solution” of KPZ eq. [homework (advanced)] Consider colored noise with Then argue how fast h(x,t) grows in KPZ eq. 7 3. Exact results on 1d KPZ class [5,6] 3.1. Polynuclear growth (PNG) model Rules: (1) Random nucleation (at constant nucleation rate ρ) (2) Lateral expansion (at constant speed v) We set ρ=1 and v=1. 3.2. PNG circular interface [6] 1st nucleation at (x,t)=(0,0), subsequent nucleations occur only on top of the 1st plateau. Space-time plot h(0,t) = # of lines to pass from (0,0) to (0,t) = Max # of points passed by directed polymer with fixed end points (0,0) & (0,t) (= its ground-state energy under random potential U(x,t) = -Σδ(x-xn)δ(t-tn) ) = Length of longest increasing subsequences of random permutations with Poisson-distributed length (Ulam’s problem) = … (combinatorics, Young tableau, etc.) (17) largest-eigenvalue distribution of GUE random matrices (GUE Tracy-Widom distribution) 8 [homework] Draw a new space-time plot and convince yourself of each step of Eq.(17). h(x,t) behaves analogously, but ’s area is Semicircle mean profile. 9
© Copyright 2026 Paperzz