Statistical Physics Approach to Understanding the Multiscale Dynamics of Earthquake Fault Systems Theory Statistical Physics Approach to Understanding the Multiscale Dynamics of Earthquake Fault Systems Systems composed of large number of simple, interacting elements Uninterested in small-scale (random) behaviour Use methods of statistics (averages!) Huge range of scale Phenomenology of dynamics Overview • • • • • Motivation Scaling laws Fractals Correlation length Phase transitions – boiling & bubbles – fractures & microcracks • Metastability, spinodal line Limitations to Observational Approach • Lack of data (shear stress, normal stress, fault geometry) • Range of scales: Fault length: ~300km Fault slip ~ m Fault width ~ cm Scaling Laws log(y) = log(c) - b log(x) y = c x-b b>0 x0 Why are scaling laws interesting? Consider interval (x0, x1) minimum of y is y1 = c (x1)-b maximum of y is y2 = c (x0)-b ratio y2/y1 = (x0/x1)-b now consider interval ( x0, x1) minimum of y is y1 = c ( x1)-b maximum of y is y2 = c ( x0)-b ratio y2/y1 = (x0/x1)-b → ratio independent of scale λ! x1 λ x0 λ x1 compare with: y = ex/b, b>0 on (x0, x1), y1 = ex0/b, y2 = ex1/b y2/y1 = e(x1-x0)/b on ( x0, x1), y1 = e x0/b, y2 = e x1/b y2/y1 = e (x1-x0)/b → power-law relation ≈ scale-free process Earthquake scaling laws Gutenberg-Richter Law • Log Ngr(>m) = -b m + a – m = magnitude, measured on logarithmic scale – Ngr(>m) = number of earthquakes of magnitude greater than m occurring in specified interval of time & area – Valid locally & globally, even over small time intervals (e.g. 1 year) Omori law: dNas/dt = 1/t0 (1+t/t1)-p Nas = number of aftershocks with m>specified value t = time after main shock Benioff strain: N = number of EQs up to time t ei = energy release of ith EQ i.e. intermediate EQ activity increases before big EQ Fractals Fractal = self-similar = scale-free e.g. Mandelbrot set Fractals are ubiquitous in nature (topography, clouds, plants, …) Why? c.f. self-organized criticality, multifractals, etc. Correlation Length Correlations measure structure On average, how different is f(x) for two points a distance L apart? Let correlation length = scale where correlation is maximal L L Lc L Correlation length ~ largest structure size Correlation length → ∞ ~ all scales present = scale-free Phase transition model… Let’s look at earthquakes as phase transitions! 1st order phase transition 2nd order phase transition involves latent heat NO latent heat involved solid/liquid/gas QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture. supercritical fluid QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture. Phase diagram of a pure substance: coexistence of liquid and vapor phases! Isothermal decrease in pressure Liquid boils at constant P Vapor equilibrium curve Formation of metastable, superheated liquid Spinodal curve: limit of stability. No superheating beyond!!! Reduction in P, leads to isothermal expansion Explosive nucleation and boiling (instability) at constant P,T s’more about stability… why a spinodal line? Qu ickT i me ™ an d a T IF F (Un co mp re ss ed ) d ec om p res so r a re ne ed ed to se e thi s pic tu re. Consequence of… Van der Waals equation (of state) (real gas) Ideal Gas Law QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture. Qu ickTi me ™ an d a TIFF (Un co mp re ss ed ) d ec omp res so r a re ne ed ed to se e thi s pic tu re. isotherms QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture. pressure pressure correction term for intermolecular force, attraction between particles correction for the real volume of the gas molecules, volume enclosed within a mole of particles isotherms volume volume Incompressible Metastable region: fluid2-phase (liquid): at small V and at intermediate low V compressible fluid (gas): atcoexistence large V and low P: P: isotherms isotherms and low Pshow with horizontal large increase in P for small decrease in V show small decrease in Pisotherms for large decrease in V The spinodal line is interesting! Limit of stability! It acts like a line of critical points for nucleating bubbles Now let’s look at brittle fracture of a solid as a phase change… DamageDamage occurs along constant strain path or until stress is reduced to yield occurs at constant stress stress (IH)… similar to constant volume boiling (DH) pressure Undergoes phase change at B Let’s look at a plot of Stress vs. Strain… Elasticsolid solidstrained rapidly rapidly loaded with with a Elastic Deforms elastically until failure at B constant stress ( < yield stress) constant strain ( < yield strain) When damage occurs along a constant strain path… QuickTim e™ and a TIFF (Un co m press ed ) de com pres sor are n eed ed to se e this p ictu re . We call it stress relaxation! Applicable to understanding the aftershock sequence that follows an earthquake Rapid stress! Rapid stress! earthquake Rapid stress! Rapid stress! If rapid stress is greater than yield stress: microcracks form, relaxing stress to yield stress Time delay of aftershock relative to main shock = time delay of damage Why? Because it takes time to nucleate microcracks when damage occurs in form of microcracks. Damage is accelerated strain, leading to a deviation from linear elasticity. How do we quantify derivation from linear elasticity? a damage variable!! as failure occurs as increases : brittle solid weakens due to nucleation and coalescence of Microcracks. Spinodal Line Increasing correlation length Metastable region nucleation coalescence phase change Metastability – an analogy Consider a ball rolling around a ‘potential well’ Gravity forces the ball to move downhill If there is friction, the ball will eventually stop in one of the depressions (A, B, C) What happens if we now perturb the balls? (~ thermal fluctuations) B is globally stable, but A & C are only metastable A C B If we now gradually make A & C shallower, the chance of a ball staying there becomes smaller Eventually, the stable points A & C disappear – this is the limit of stability, the spinodal Tomorrow, we will consider a potential that changes in time the end Qu ickT i me ™ an d a T IF F (Un co mp re ss ed ) d ec om p res so r a re ne ed ed to se e thi s pic tu re.