coupled rotors and classical spin chains: predictions from ... · lecture notes in phys. vol 921:...
TRANSCRIPT
Coupled rotors and classical spin chains: predictions fromfluctuating hydrodynamics and numerical tests
Abhishek DharInternational centre for theoretical sciences
TIFR, Bangalorewww.icts.res.in
Suman G. Das (NCBS, Bangalore)Kedar Damle (TIFR, Mumbai)
Manas Kulkarni (ICTS, Bangalore)Herbert Spohn (TU, Munich)
David Huse (Princeton)
Numerical aspects of nonequilibrium dynamics, IHP, ParisApril 25 - 27, 2017
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 1 / 27
Outline
Signatures of anomalous heat conduction in one dimensional momentum systems.
Introduction to Fluctuating Hydrodynamic Theory and results for anharmonic chains.
Other systems: Rotor models and Discrete NonLinear Schrodeinger equation(Gross-Pitaevski).
Heisenberg spin-chains: hydrodynamic theory, numerical results.
References1 Fermi-Pasta-Ulam chains: S. Das, AD, K. Saito, C. Mendl, H. Spohn, PRE 90, 012124 (2014).
2 Hydrodynamics of the rotor model: S. Das, AD, arXiv:1411.5247 (2014) / Spohn:arxiv:1411.3907 (2014)
3 Hydrodynamics and Correlations in the Toda lattice and other integrable models: Aritra Kunduand AD, PRE (2016). POSTER
4 Discrete nonlinear Schrodinger equation: Kulkarni, Huse, Spohn, PRA (2015), Mendl, Spohn,J. Stat. Mech. (2015).
5 Heisenberg spins: Damle, Spohn, AD, Kulkarni, Huse (Ongoing work).
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 2 / 27
Breakdown of Fourier’s law — Anomalous heat transport
Fourier’s law is probably not valid in low-dimensional momentum-conserving systems.
Lecture notes in Phys. vol 921: Thermal transport in low dimensions (2016).
Divergent conductivity: κ ∼ Lα.
Nonlinear (and possibly singular) temperature profiles, EVEN for small temperaturedifferences.
Anomalous spreading of heat pulses — Levy walk instead of random walk.
Anomlaous behaviour of equilibrium space-time correlations of conserved quantities.
Predictions of fluctuating hydrodynamics — Levy heat peak and KPZ sound peaks.
Slow temporal decay of total energy-current correlations and conclusions from Green-Kuboformula.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 3 / 27
Anomalous transport - spreading of energy pulses
Look at propagation of a energy pulse
The energy profile followsthe Levy-stable distribution.Power-law decay at large x .Finite speed of propagationof front.〈x2〉 ∼ t1+α
(Super-diffusive).
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 4 / 27
Understanding anomalous transport
Nonlinear fluctating hydrodynamics - a general framework.
Narayan, Ramaswamy (2002), H. vanBeijeren (2012)
Very detailed predictions: H. Spohn and C. Mendl (2013,2014)
Look at decay of energy fluctuations in a system in thermal equilibrium. Thus one can look atspatio-temporal correlation functions such as
C(x , t) = 〈δε(x , t) δε(0, 0)〉,
where δε(x , t) is fluctuation in local energy density.Anomalous transport would imply super-diffusive spreading of such correlation functions.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 5 / 27
Systems studied so far in the framework of NlFHT
Fermi-Pasta-Ulam chains: Zhao, Das-AD-Saito-Mendl-Spohn
Hard particle gases: Mendl-Spohn
Rotor chains: Das-AD, Spohn, Mendl-Spohn
DNLS: Kulkarni-Huse-Spohn, Mendl-Spohn
Stochastic models: Stoltz-Spohn, Lepri et al, Cividini-Kundu-Miron-Mukamel
Coupled exclusion processes: Popkov-Schmidt-Schütz
What about spin-chains ?
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 6 / 27
Basics of fluctuating hydrodynamics
Fermi-Pasta-Ulam Hamiltonian:
H =N∑
x=1
p2x
2+ V (qx+1 − qx) , V (r) = k2
r2
2+ k3
r3
3+ k4
r4
4.
Identify the conserved fields. For the FPU chain they areExtension: rx = qx+1 − qxMomentum: pxEnergy: ex
Using equations of motion one can directly arrive at the following conservation laws (Eulerequations):
∂r∂t
=∂p∂x,
∂p∂t
= −∂P∂x
,∂e∂t
= −∂pP∂x
,
where Px = 〈−V ′(rx )〉 is the pressure.
Consider constant T ,P and zero momentum ensemble.Let (u1, u2, u3) be fluctuations of conserved fields about equilibrium values:rx = 〈rx〉+ u1(x), px = u2(x), ex = 〈ex〉+ u3(x).
Expand the curents about their equilibrium value (to second order in nonlinearity) and writehydrodynamic equations for these fluctuations.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 7 / 27
Fluctuating hydrodynamics basics
Let u = (u1, u2, u3). Equations have the form:
∂u∂t
= − ∂
∂x[Au + uHu] +
[D∂2u∂x2 + B
∂ξ
∂x
].
1D noisy Navier-Stokes equation
A,H known explicitly in terms of microscopic model.D, B unknown but satisfy fluctuation dissipation.
Neglecting nonlinear terms, one can construct normal mode variables (φ+, φ0, φ−), as linearcombinations of the original fields φ = Ru. These satisfy equations of the form
∂φ+∂t
= −c∂φ+∂x
+ Ds∂2φ+∂x2 +
∂η+∂x
∂φ0
∂t= Dh
∂2φ0
∂x2 +∂η0
∂x∂φ−∂t
= c∂φ−∂x
+ Ds∂2φ−∂x2 +
∂η−∂x
NOTE: two propagating sound modes ( φ±) and one diffusive heat mode ( φ0).(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 8 / 27
Predictions of fluctuating hydrodynamics
Including the nonlinear terms:
∂φ+
∂t=
∂
∂x[−cφ++G+φ2] + Ds
∂2φ+
∂x2+∂η+
∂x∂φ0
∂t=
∂
∂x[G0φ2] + Dh
∂2φ0
∂x2+∂η0
∂x∂φ−
∂t=
∂
∂x[cφ−+G−φ2] + Ds
∂2φ−
∂x2+∂η−
∂x
Given V (r),T ,P, the form of the G-matrices is completely determined.
Generic case: To leading order, the oppositely moving sound modes are decoupled from theheat mode and satisfy noisy Burgers equations. For the heat mode, the leading nonlinearcorrection is from the two sound modes.
Solving the nonlinear hydrodynamic equations within mode-coupling approximation, one canmake predictions for the equilibrium space-time correlation functionsC(x , t) = 〈φα(x , t)φβ(0, 0)〉.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 9 / 27
Predictions of fluctuating hydrodynamics
Predictions for equilibrium space-time correlation functions C(x , t) = 〈φα(x , t)φβ(0, 0)〉.
Sound− mode : Cs(x , t) = 〈φ±(x , t)φ±(0, 0)〉 =1
(λs t)2/3fKPZ
[(x ± ct)(λs t)2/3
]
Heat− mode : Ch(x , t) = 〈φ0(x , t)φ0(0, 0)〉 =1
(λet)3/5fLW
[x
(λet)3/5
]c, the sound speed and λ are given by the theory.fKPZ - universal scaling function that appears in the solution of the Kardar-Parisi-Zhangequation.fLW – Levy-stable distribution with a cut-off at |x | = ct .
Cross correlations negligible at long times.
Also find 〈J(0)J(t)〉 ∼ 1/t2/3.
Correlations from direct simulations of FPU chains and comparisions with theory.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 10 / 27
Equilibrium space-time correlation functions
Numerically compute heat mode and sound mode correlations in the α− β-Fermi-Pasta-Ulamchain with periodic boundary conditions.
Average over ∼ 107 thermal initial conditions. Dynamics is Hamiltonian.
Parameters — k2 = 1, k3 = 2, k4 = 1, T = 5.0, P = 1.0, N = 16384.
Speed of sound c = 1.803.
0
0.0025
0.005
0.0075
0.01
C(x,t)
−5000 0 5000
x
t = 800t = 2400t = 3200
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 11 / 27
Equilibrium simulations of FPU
Sound mode scaling: λtheory = 0.396, λsim = 0.46.
0
0.2
0.4
0.6
(λst)
2/3C
−−(x,t)
−4 −2 0 2 4
(x + ct)/(λst)2/3
t = 800t = 3200t = 4000KPZ
Heat mode scaling: λtheory = 5.89, λsim = 5.86.
0
0.1
0.2
0.3
(λet)
3/5C
00(x,t)
−5 0 5
x/(λet)3/5
t = 800t = 2400t = 3200Levy
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 12 / 27
Equilibrium simulations of FPU
Other results:
Other parameter regimes: KPZ and Levy scaling are always very good. Values of scalingparameters sometimes far from theory. Fit to KPZ scaling function not always good.
Second universality class: even potential and zero pressure.
Sound modes diffusive, heat mode Levy with different exponent [ f (k , t) = exp(−|k |3/2t) ].
Other possible special points — See “Fibonacci family of dynamical universality classes”, V.Popkova, A. Schadschneidera, J. Schmidta, and G. M. Schütz [PNAS 112, 12645 (2015)].
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 13 / 27
Rotor model
Hamiltonian of the Rotor model —
H =N∑
l=1
p2l
2m−
N−1∑l=1
V0 cos(ql+1 − ql )
Non-equilibrium simulations show that this model satisfies Fourier’s law and does not showanomalous transport, even though it is momentum conserving.Giardina, Livi, Politi, Vassali (2000), Gendelman, Savin (2000).
This can be understood in the framework of the hydrodynamics theory.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 14 / 27
Hydrodynamic theory for the Rotor model
Spohn: arxiv:1411.3907S. Das and AD, arXiv:1411.5247 (2014)
1 (a) The coordinate variables are now angles, defined modulo 2π. Hence stretch notconserved.(b) Since V (r) is bounded, the pressure is identically zero.
2 Recall the conservation laws —∂r∂t
= −∂p∂x,
∂p∂t
= −∂P∂x
,∂e∂t
= −∂pP∂x
+ [Dissipation + Noise] ,
Since P is zero, in the final description, only the dissipative and fluctuation terms survive.
Thus the hydrodynamic equations for ~u = (r , p, e) are
∂t uα = −∂x[−∂x Dαβuβ + Bαβξβ
].
[Chaikin and Lubensky!]
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 15 / 27
Rotor model
• Numerical observation: Cross elements Dα 6=β vanish.
V (r) = − cos(r),T = 1.• Momentum and Energy correlations decay diffusively — completely different from FPU chain.
-2 0 2
x/t1/2
0
0.1
0.2
0.3
0.4
0.5
t1/2
Cp
p(x
,t)
/ C
pp(0
,0)
t=2000t=4000t=6000t=8000D=0.304
-4 -2 0 2 4
x/t1/2
0
0.1
0.2
0.3
0.4
t1/2
Cee
(x,t
) /
Cee
(0,0
)
t=2000t=4000t=6000t=8000D=0.627
(a) (b)
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 16 / 27
Rotor model at low temperatures
At low temperatures the angles ql make small fluctuations around some fixed value (Brokensymmetry phase!) — replace − cos(r) by −1 + r2/2− r4/24 + r6/720.
Hence the restriction ql ∈ (0, 2π) is NOT relevant.
In this case one finds two diffiusive sound modes (Zero pressure and even potential) and aLevy heat mode.
The hydrodynamic equations for both the high temperature “disordered phase” and the“broken symmetry phase” have been discussed earlier (Chaikin and Lubensky, Principles ofcondensed matter physics).
However it was not realized that the heat mode is actually non-diffusive and with Levycharacteristics.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 17 / 27
XXZ classical Heisenberg spin chains
Heisenberg spins S = (Sx ,Sy ,Sz) on a ring with nearest neighbor interactions. Hamiltonian givenby
H =N∑`=1
−J[Sx`Sx`+1 + Sy
`Sy`+1
]− JzSz
`Sz`+1 .
Consider J > Jz — easy-plane magnetization.
Equations of motion − >
Sα` = −eαβγSβ`∂H∂Sγ`
. OR [S` = S` × Beff (`)] −−−− Symplectic dynamics
Explicitly
Sx` = −J
[Sy`+1 + Sy
`−1
]Sz` + Jz
[Sz`+1 + Sz
`−1]
Sy`
Sy` = −Jz
[Sz`+1 + Sz
`−1]
Sx` + J
[Sx`+1 + Sx
`−1]
Sz`
Sz` = −J
[Sx`+1 + Sx
`−1]
Sy` + J
[Sy`+1 + Sy
`−1
]Sx` .
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 18 / 27
Regular Hamiltonian form
Define new variables {s`, θ`}
Sz` = s`, Sx
` = (1− s2`)
1/2 cos θ`, Sy` = (1− s2
`)1/2 sin θ`.
This defines a canonical transformation leading to equations of motion
θ` =∂H∂s`
, s` = −∂H∂θ`
,
with
H = −JN∑`=1
(1− s2`)
1/2(1− s2`+1)
1/2 cos(θ`+1 − θ`)− Jzs`s`+1 =N∑`=1
ε`.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 19 / 27
Conservation laws
Two Exact conservation laws: Energy and z-magnetization-> continuity equations:
s` = −(js`+1 − js` ) , ε` = −(jε`+1 − jε` ) ,
with currents
js` = −J(1− s2`)
1/2(1− s2`+1)
1/2 sin(r`) ,
jε`+1 = −J2s`+1(1− s2`)
1/2(1− s2`+2)
1/2 sin (r`+1 − r`)
+ JJzs`(1− s2`+1)
1/2(1− s2`+2)
1/2 sin r`+1
− JJzs`+2(1− s2`)
1/2(1− s2`+1)
1/2 sin r` .
r` = θ`+1 − θ`.
Equilibrium distribution − > e−β(H−µ∑
` s`).Equilibrium currents IDENTICALLY zero:〈js〉 = 〈jε〉 = 0.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 20 / 27
Hydrodynamics at high temperatures
Hence we expect diffusive hydrodynamic equations.
Numerical tests: β = 1.0, µ = 0.3 , Css(x , t) = 〈sx (t)s0(0)〉eq .
s − s correlations
-300 -200 -100 0 100 200 300x
0
0.0005
0.001
0.0015
0.002
Css
(x,t)
t=512t=768t=1024
Diffusive scaling
-12 -8 -4 0 4 8 12x/t1/2
0
0.01
0.02
0.03
0.04
0.05
t1/2 C
ss(x
,t)
t=512t=768t=1024
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 21 / 27
Low temperature: effective symmetry breaking
At low temperatures, one expects the xy -plane symmetry to be broken.Small fluctuations around a chosen θ∗ and r` = θ`+1 − θ` = ∇θ is a new “conserved” variable.
Clearly r` = −∇j r with j r = θ = ∂H/∂s.
Equilibrium measure is now − > ∼ e−β(H−µ∑
` s`−ν∑
` r`).
Equilibrium currents are now NON-ZERO. Magic Identity (Mendl,Spohn, 2016) gives
〈j r 〉eq = 〈−∂H/∂s〉eq = −µ, 〈js〉eq = 〈−∂H/∂r〉eq = −ν , 〈jε〉eq = µν .
— This is useful for computing linear and nonlinear coefficients in FHT equations.
Fluctuating hydrodynamics then predicts two KPZ sound modes and one Levy heatmode. [SPECIAL CASE: µ = 0, two diffusive sound modes, one Levy heat mode]
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 22 / 27
Low temperature correlations
Numerical tests: β = 8.0, µ = 0.3, N = 4096
s − s correlations
-2000 -1000 0 1000 2000x
0
0.00015
0.0003
0.00045
0.0006
Css
(x,t)
t=768t=1024t=1280t=1536t=1792t=2048
KPZ-scaling
-2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5t-2/3(x-ct)
0
0.01
0.02
0.03
0.04
0.05
t2/3 C
ss(x
,t)
t=768t=1024t=1280t=1536t=1792t=2048
Speed of sound: c = 0.819....
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 23 / 27
Low temperature: heat mode
ε− ε correlations
-2000 -1000 0 1000 2000x
0
1e-05
2e-05
Cεε
(x,t)
Levy-5/3 scaling
-15 -10 -5 0 5 10 15x/t3/5
0
0.00025
0.0005
0.00075
0.001
t3/5 C
εε(x
,t)
Still need to go to normal modes to see scaling clearly — i.e look at linear combinations of thebasic fields.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 24 / 27
Very low-temperature: integrable dynamics
s − s correlations
-1000 -500 0 500 1000x
0
0.0002
0.0004
0.0006
0.0008
Css
(x,t)
t=256t=512t=768
Ballistic scaling
-0.4 -0.2 0 0.2 0.4(x-ct)/t
0
0.05
0.1
0.15
0.2
t-1C
ss(x
,t)
t=256t=512t=768
Ballistic scaling implies integrability at low temperatures — but probably not harmonic ?
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 25 / 27
Numerical methods
We implemented an effective finite-time version of the dynamics, which preserves theconservation laws exactly [Damle (unpublished notes)]. Possible to study large system sizesand large times.
Dynamics is a variant of symplectic algorithms, normally written for (X ,P) systems,implemented here for spin dynamics.
Similar to odd-even dynamics which conserves ONLY energy exactly — alternately odd andeven sites updated in parallel.
Can check that in the limit of small update times, dynamics is equivalent to the exactPoisson-bracket dynamics.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 26 / 27
Conclusions
Equilibrium space-time correlations of conserved variables in one-dimensional interactingsystems.Very detailed theoretical predictions [Spohn, JSP (2014)] allow direct comparision withmicroscopic simulations.
Fermi-Pasta-Ulam chains, Rotor chain and integrable models.
A new class is investigated here — classical spin chains.
Simulations for XXZ chain verify the scaling predictions quite well. [Preliminary results- workin progress]High temperatures - Diffusive scalingLow temperatures - Anomalous scaling
— Levy scaling for heat mode
— KPZ scaling for sound-mode
Very low temperatures - Ballistic scaling
Open questions:
Derivation of hydrodynamic equations (Diffusion matrix ?).
Understanding strong finite size effects seen in simulations.
(Abhishek Dhar, ICTS-TIFR) IHP, April 25-27, 2017 27 / 27