Long-range correlations in driven systems David Mukamel Outline Will discuss two examples where long-range correlations show up and consider some consequences Example I: Effect of a local drive on the steady state of a system Example II: Linear drive in two dimensions: spontaneous symmetry breaking Example I :Local drive perturbation T. Sadhu, S. Majumdar, DM, Phys. Rev. E 84, 051136 (2011) Local perturbation in equilibrium Particles diffusing (with exclusion) on a grid occupation number ππ = 0,1 N particles V sites Prob. of finding a particle at site k π π π = π Add a local potential u at site 0 1 1 π π½π’ N particles V sites 1 The density changes only locally. 1 1 0 1 Effect of a local drive: a single driving bond Main Results In d ≥ 2 dimensions both the density corresponds to a potential of a dipole in d dimensions, decaying as π π ~1/π π−1 for large r. The current satisfies π(π)~1/π π . The same is true for local arrangements of driven bonds. The power law of the decay depends on the specific configuration. The two-point correlation function corresponds to a quadrupole In 2d dimensions, decaying as πΆ(π, π )~1/ π 2 + π 2 π for π = 1/2 The same is true at other densities to leading order in π (order π 2 ). Density profile (with exclusion) The density profile 1/π 2 π(π)~ 1/π along the y axis in any other direction Non-interacting particles • Time evolution of density: ππ‘ π π, π‘ = π» 2 π π, π‘ + ππ 0, π‘ [πΏπ,0 − πΏπ,π1 ] π» 2 = π π + 1, π + π π − 1, π + π π, π + 1 + π π, π − 1 − 4π(π, π) The steady state equation π» 2 π π = −ππ(0)[πΏπ,0 − πΏπ,π1 ] particle density → electrostatic potential of an electric dipole π» 2 π π = −ππ(0)[πΏπ,0 − πΏπ,π1 ] Green’s function solution π» 2 πΊ π, ππ = −πΏπ,ππ π π = π + ππ 0 [πΊ π,0 − πΊ(π,π1 )] Unlike electrostatic configuration here the strength of the dipole should be determined self consistently. Green’s function of the discrete Laplace equation p\q 0 0 0 1 2 1 − 4 2 −1 π 1 1 − 4 1 − π 1 − 4 2 2 −1 π 1 2 − 4 π 4 − 3π πΊ π, ππ 1 ≈− ln |π − π0 | 2π determining π(0) To find π(0) one uses the values πΊ 0, 0 = 0 , πΊ 0, π1 = π 0 = π 1− π 4 1 − 4 at large π density: current: ππ 0 π1 π 1 π π =π− + π( 2 ) 2 2π π π ππ 0 1 2 π1 π π 1 π π = [π1 − + π( 3 ) 2 2 2π π π π Multiple driven bonds π π = π + ππ π1 πΊ π, π1 − πΊ π, π1 + π1 + ππ π2 πΊ π, π2 − πΊ π, π2 + π1 Using the Green’s function one can solve for π π1 , π π2 … by solving the set of linear equations for π = π11 , π2 , β― +β― Two oppositely directed driven bonds – quadrupole field The steady state equation: π» 2 π π = −ππ(0)[2πΏπ,0 − πΏπ,π1 −πΏπ,−π1 ] ππ 0 π π =π− 2π 1 π1 π −2 2 2 π π 2 1 + π( 4 ) π π ≠ 2 dimensions π=1 π π₯ =π− πΊ π₯, π₯π = − π≥2 π 2 π 0 π ππ(π₯) π₯−π₯π 2 π π ∼ 1 π π−1 The model of local drive with exclusion Here the steady state measure is not known however one can determine the behavior of the density. ππ‘ π π, π‘ = π» 2 π π, π‘ + π π 0 1 − π π1 [πΏπ,0 − πΏπ,π1 ] π = 0,1 is the occupation variable π π 0 1 − π π1 π π =π− 2π π1 π 1 + π( 2 ) 2 π π π π 0 1 − π π1 π π =π− 2π π1 π 1 + π( 2 ) 2 π π The density profile is that of the dipole potential with a dipole strength which can only be computed numerically. Simulation results Simulation on a 200 × 200 lattice with π = 0.6 For the interacting case the strength of the dipole was measured separately . Two-point correlation function πΆ π, π =< π π π π >- π(r) π(π ) In d=1 dimension, in the hydrodynamic limit πΆπΏ π, π = 1 π g( πΏ πΏ , π ) πΏ T. Bodineau, B. Derrida, J.L. Lebowitz, JSP, 140 648 (2010). In higher dimensions local currents do not vanish for large L and the correlation function does not vanish in this limit. T. Sadhu, S. Majumdar, DM, in progress Symmetry of the correlation function: πΆ π, π =< π π π π >- π(r) π(π ) πΆπ r, s = C−π (−π, −π ) inversion particle-hole πΆπ,π r, s = C−π,1−π (r, s) at π = 1/2 πΆπ π, π = πΆ−π (π, π ) ππ π πΆ π, π = 0 Δπ + Δπ πΆ π, π = π π, π πΆ(π, π ) corresponds to an electrostatic potential in 2π induced by π Consequences of the symmetry: The net charge =0 At π = 0.5 πΆπ π, π is even in π Thus the charge cannot support a dipole and the leading contribution in multipole expansion is that of a quadrupole (in 2d dimensions). πΆ(π, π )~1/ π 2 + π 2 π For π ≠ 0.5 one can expand π in powers of π One finds: The leading contribution to π is of order π 2 implying no dipolar contribution, with the correlation decaying as πΆ(π, π )~1/ π 2 + π 2 π πΆ π, π = ππΌ1 π, π + π 2 πΌ2 π, π + β― πΌ2π−1 −π, −π = −πΌ2π−1 π, π πΌ2π −π, −π = πΌ2π (π, π ) Since πΌ1 = 0 (no dipole) and the net charge is zero the leading contribution is quadrupolar Δπ + Δπ πΆ π, π = π1 π, π + π2 π, π + π3 (π, π ) π1 π, π = π π π π πΏπ ,0 − πΏπ +π1 ,0 1 − πΏπ,0 − πΏπ+π1 ,0 + π π π π πΏπ,0 − πΏπ+π1 ,0 1 − πΏπ ,0 − πΏπ +π1 ,0 1 2 2 2 π2 π, π = − π π πΏπ,−π1 πΏπ ,0 + π π + ππ − π π πΏπ ,π+ππ π π 1 2 2 2 − π π πΏπ ,−π1 πΏπ,0 + π π + ππ − π π πΏπ ,π−ππ π π π3 π, π = πΆ π + ππ , π − 2πΆ π, π + πΆ π, π − ππ πΏπ ,π+ππ + πΏπ ,π π + πΆ π, π + ππ − 2πΆ π, π + πΆ π − ππ , π πΏπ ,π−ππ + πΏπ ,π π π = π 0 1 − π −π1 + π(−π1 ) 1 − π(0) Summary Local drive in π > 1 dimensions results in: Density profile corresponds to a dipole in d dimensions π π ~ 1/π π−1 Two-point correlation function corresponds to a quadrupole in 2d dimensions πΆ(π, π )~1/ π 2 + π 2 π At density π = 0.5 to all orders in π At other densities to leading (π 2 ) order Example II: a two dimensional model with a driven line The effect of a drive on a fluctuating interface T. Sadhu, Z. Shapira, DM PRL 109, 130601 (2012) Motivated by an experimental study of the effect of shear on colloidal liquid-gas interface. D. Derks, D. G. A. L. Aarts, D. Bonn, H. N. W. Lekkerkerker, A. Imhof, PRL 97, 038301 (2006). T.H.R. Smith, O. Vasilyev, D.B. Abraham, A. Maciolek, M. Schmidt, PRL 101, 067203 (2008). What is the effect of a driving line on an interface? - + In equilibrium- under local attractive potential + - Local potential localizes the interface at any temperature π < ππ Transfer matrix: 1d quantum particle in a local attractive potential, the wave-function is localized. no localizing potential: < Δβ2 > ~ πΏ with localizing potential: < π½ < Δβ2 > ~ππππ π‘ Schematic magnetization profile my 1.0 0.5 20 10 10 20 y 0.5 1.0 The magnetization profile is antisymmetric with respect to the zero line with π π¦ = 0 = 0 Consider now a driving line +-→-+ with rate min(1, π −π½Δπ»+π½πΈ ) -+→+- with rate min(1, π −π½Δπ»−π½πΈ ) π» = −π½ ππ ππ <ππ> Ising model with Kawasaki dynamics which is biased on the middle row Main results The interface width is finite (localized) A spontaneous symmetry breaking takes place by which the magnetization of the driven line is non-zero and the magnetization profile is not symmetric. The fluctuation of the interface are not symmetric around the driven line. These results can be demonstrated analytically in certain limit. Results of numerical simulations Example of configurations in the two mesoscopic states for a 100X101 with fixed boundary at T=0.85Tc Schematic magnetization profiles my my 20 1.0 1.0 0.5 0.5 10 10 20 y 20 10 10 0.5 0.5 1.0 1.0 20 my 1.0 0.5 unlike the equilibrium antisymmetric profile 20 10 10 0.5 1.0 20 y y Averaged magnetization profile in the two states L=100 T=0.85Tc Time series of Magnetization of driven lane for a 100X101 lattice at T= 0.6Tc. Switching time on a square LX(L+1) lattice with Fixed boundary at T=0.6Tc. Analytical approach In general one cannot calculate the steady state measure of this system. However in a certain limit, the steady state distribution (the large deviations function) of the magnetization of the driven line can be calculated. Typically one is interested in calculating −πΉ π π₯, π¦ the large deviation function of a magnetization profile π(π π₯, π¦ )~π −πΏ2 πΉ(π π₯,π¦ ) We show that in some limit a restricted large deviation function, that of the driven line magnetization, π0 , can be computed π π0 = π −πΏπ(π0) The following limit is considered Slow exchange rate between the driven line and the rest of the system Large driving field πΈ β« π½ Low temperature In this limit the probability distribution of m0 is π π0 = π −πΏπ(π0) where the potential (large deviations function) π π0 can be computed. The large deviations function π π π0 = π −πΏπ(π0) Slow exchange between the line and the rest of the system ππ€(Δπ») +− β ππ€(−Δπ») − + 1 π < π( 3 ) πΏ π€ Δπ» = min(1, π −π½Δπ» ) In between exchange processes the systems is composed of 3 sub-systems evolving independently Fast drive πΈ β« π½ the coupling π½ within the lane can be ignored. As a result the spins on the driven lane become uncorrelated and they are randomly distributed (TASEP) The driven lane applies a boundary field π½π0 on the two other parts Due to the slow exchange rate with the bulk, the two bulk sub-systems reach the equilibrium distribution of an Ising model with a boundary field π½π0 Low temperature limit In this limit the steady state of the bulk sub systems can be expanded in T and the exchange rate with the driven line can be computed. ππ βΆ ππ + ππ βΆ ππ − 2 πΏ 2 πΏ π(π0 ) with rate π(ππ ) with rate π(ππ ) π(π0 ) ππ performs a random walk with a rate which depends on ππ 2 π−1 π(0) β― β― π( ) 2π πΏ π ππ = = ≡ π −π(ππ) 2 2π πΏ π( ) β― β― π( ) πΏ πΏ ππ πΏ 2 −1 π ππ = − π=0 2π ln π + πΏ ππ πΏ 2 π=1 2π π( ) πΏ Calculate p ππ at low temperature - - + + + + + + + + + + + + - - + + + + + contribution to p ππ : + + 1 8 + + + + + 1 − ππ 1 + ππ 2 π −2π½π½ π −2π½π½1 π½1 is the exchange rate between the driven line and the adjacent lines The magnetization of the driven lane πβ changes in steps of 2/πΏ Expression for rate of increase, p(ππ ) π π0 1 = 1 + m0 8 2 1 + [2 1 + π0 1 − π0 8 1 − m0 π −2π½ 2 π½+π½1 π −2π½π½1 + π 2π½π½1 π0 + 1 + π0 2 (1 − π0 )π 2π½π½1 π0 + 1 − π0 3 π 2π½π½1 π0 ]π −6π½π½ + π(π −8π½π½ ) π ππ = π(−ππ ) π π π =− 0 π ln π π ππ + Type ln πequation π ππ 0 here. π This form of the large deviation function demonstrates the spontaneous symmetry breaking. It also yield the exponential flipping time at finite πΏ. (π = 0.6ππ , π½1 = π½) π π0 = π −πΏπ(π0) < ππ > 1 − π(π −4π½π½ ) Summary Simple examples of the effect of long range correlations in driven models have been presented. A limit of slow exchange rate is discussed which enables the evaluation of some large deviation functions far from equilibrium.