1 Fractional dynamics in underground contaminant transport: introduction and applications Andrea...

Post on 26-Dec-2015

219 views 2 download

Tags:

transcript

1

Fractional dynamics in underground

contaminant transport:introduction and

applications

Andrea Zoia

Current affiliation: CEA/SaclayDEN/DM2S/SFME/LSET

Past affiliation: Politecnico di Milano and MIT

MOMAS - November 4-5th 2008

2

Outline

CTRW: methods and applications

Conclusions

Modeling contaminant migrationin heterogeneous materials

3

Transport in porous media

Highly complex velocity spectrum

ANOMALOUS (non-Fickian) transport: <x2>~t

Relevance in contaminant migration Early arrival times (): leakage from repositories

Late runoff times (): environmental remediation

Porous media are in general heterogeneous

Multiple scales: grain size, water content, preferential flow streams, …

4

[Kirchner et al., Nature 2000] Chloride transport in catchments.

Unexpectedly long retention times

Cause: complex (fractal) streams

An example

5

Continuous Time Random Walk

t

xx0

Main assumption: particles follow stochastic trajectories in {x,t} Waiting times distributed as w(t) Jump lengths distributed as (x)

Berkowitz et al., Rev. Geophysics 2006.

6

CTRW transport equation P(x,t) = probability of finding a particle in x at time t =

= normalized contaminant particle concentration

P depends on w(t) and (x): flow & material properties

)()(1

)()(1),( 0

kuw

kP

u

uwukP

Probability/mass balance (Chapman-Kolmogorov equation) Fourier and Laplace transformed spaces: xk, tu, P(x,t)P(k,u)

Assume: (x) with finite std and mean

‘Typical’ scale for space displacements

7

CTRW transport equation

')',(2

)'(),(2

22

dttxPxx

ttMtxPt

Rewrite in direct {x,t} space (FPK):

Heterogeneous materials: broad flow spectrum multiple time scales w(t) ~ t , 0<<2, power-law decay

M(t-t’) ~ 1/(t-t’): dependence on the past history

Homogeneous materials: narrow flow spectrum single time scale w(t) ~ exp(-t/) M(t-t’) ~ (t-t’) : memoryless = ADE

)(1

)()(

uw

uuwuM

Memory kernel M(u): w(t) ?

8

Asymptotic behavior

),(2

),(2

221 txP

xxtxP

t t

Fractional Advection-Dispersion Equation (FADE) Fractional derivative in time ‘Fractional dynamics’

The asymptotic transport equation becomes:

Analytical contaminant concentration profile P(x,t)

)(1)()( 11 uoucucuwttw

Long time behavior: u0

9

Long jumps

(x)~|x| , 0<<2, power law decay

),(),( txPx

txPt

The asymptotic equation is

Fractional derivative in space

Physical meaning: large displacements Application: fracture networks?

10

Monte Carlo simulation

CTRW: stochastic framework for particle transport Natural environment for Monte Carlo method

Simulate “random walkers” sampling from w(t) and (x) Rules of particle dynamics

Describe both normal and anomalous transport

Advantage: Understanding microscopic dynamics link with macroscopic

equations

11

Developments

Advection and radioactive decay

Macroscopic interfaces

Asymptotic equations

Breakthrough curves

CTRW Monte Carlo

12

1. Asymptotic equations Fractional ADE allow for analytical solutions However, FADE require approximations Questions:

How relevant are approximations? What about pre-asymptotic regime (close to the

source)?

FADE good approximation of CTRW Asymptotic regime rapidly attained

Quantitative assessment via Monte Carlo

Exact CTRW .

Asymptotic FADE _

P(x,t)

x

13

If 1 (time) or 2 (space): FADE bad approximation

1. Asymptotic equations

Exact

CTRW .Asymptotic FADE _

P(x,t)

x

FADE* _

P(x,t)

x

New transport equations including higher-order corrections: FADE* Monte Carlo validation of FADE*

14

2. Advection

How to model advection within CTRW? x x+vt (Galilei invariance) <(x)>= (bias: preferential jump direction)

Water flow: main source of hazard in contaminant migration

Fickian diffusion: equivalent approaches (v = /<t>) Center of mass: (t) ~ t

Spread: (t) ~ t

0 1 2 3 4 5 6 7 8 9 100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

(t)~t

x

P(x,t)

2(t)~t

15

2. Advection

Even simple physical mechanisms must be reconsidered in presence of anomalous diffusion

Anomalous diffusion (FADE): intrinsically distinct approaches

x x+vt

<(x)>=

P(x,t)

xx

P(x,t)

Contaminant migration

2(t)~t

(t)~t

2(t)~t2

(t)~t

16

2. Radioactive decay

Coupling advection-dispersion with radioactive decay

),(),(2

2

txPx

vx

DtxPt

Normal

diffusion:Advection-dispersion

),(1

),(),(2

2

txPtxPx

vx

DtxPt

… & decay

),(),(2

21 txP

xv

xDtxP

t t

Anomalou

s diffusion:Advection-dispersion

),(1

),(),(2

2/1/ txPtxP

xv

xDeetxP

tt

tt

… & decay

/),(),( tetxPtxP

17

3. Walking across an interface Multiple traversed materials, different physical properties

{,,}1 {,,}2

Set of properties {1}

Set of properties {2}

Two-layered medium

Stepwise changes

Interface

What happens to particles when crossing the interface?

?

x

18

3. Walking across an interface

“Physics-based” Monte Carlo sampling rules

Linking Monte Carlo parameters with equations coefficients

Case study: normal and anomalous diffusion (no advection)

Analytical boundary conditions at the interface

),(),()(),()(2

)(),(

0),()0,(),(

uxPuxMxuxMxx

xuxJ

uxJx

xPuxuP

19

3. Walking across an interface

Fickian diffusion

layer1 layer2

P(x,t)

x Interface

Anomalous

Interface

P(x,t)

xExperimental

results

Key feature: local particle velocity

layer1 layer2

20

4. Breakthrough curves Transport in finite regions

A

Injection

x0

0 1 2 3 4 5 6 7 8 9 10

x 105

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Breakthrough curve (t)

t

Outflow

The properties of (t) depend on the eigenvalues/eigenfunctions of the transport operator in the region [A,B]

Physical relevance: delay between leakage and contamination

Experimentally accessible

B

21

4. Breakthrough curves Time-fractional dynamics: transport operator = Laplacian

)()(2

2

xxx

2

2

x

Well-known formalism

Space-fractional: transport operator = Fractional Laplacian

)()( xxx

|| x

Open problem…

Numerical and analytical characterization of eigenvalues/eigenfunctions

x

x

(t)

22

Conclusions

Current and future work:

link between model and experiments (BEETI: DPC, CEA/Saclay)

Transport of dense contaminant plumes: interacting particles.

Nonlinear CTRW?

Strongly heterogeneous and/or unsaturated media:

comparison with other models: MIM, MRTM…

Sorption/desorption within CTRW: different time scales?

Contaminant migration within CTRW model

23

Fractional derivatives

ttt )1(

)1(,0

Definition in direct (t) space:

Definition in Laplace transformed (u) space:

Example: fractional derivative of a power

24

Generalized lattice Master Equation

''

),'()',(),(),'(),(ss

tsCssrtsCssrtsCt

' 0' 0

')','()','(')',()','(),(s

t

s

t

dttspttssdttspttsstspt

sus

usuus

),(1

),(),(

Master Equation

Normalized particle concentration

Transition ratesMass conservation at each lattice site

s

Ensemble average on possible rates realizations:

),( ts Stochastic description of traversed medium

Assumptions:

lattice continuum )()(),( stwts

25

Chapman-Kolmogorov Equation

P(x,t) = normalized concentration (pdf “being” in x at time t)

Source terms

(t) = probability of not having moved

p(x,t) = pdf “just arriving” in x at time t

Contributions from the past history

26

Higher-order corrections to FDE

)(1)()( 11 uoucucuwttw

)(1)()( 222

1kokckckxx

FDE: u0

FDE: k0

),(),(),( 12

21 txP

tqtxP

xtxP

t tt

),(),(),(2

2

txPx

qtxPx

txPt

Fourier and Laplace transforms, including second order contributions

Transport equations in direct space, including second order contributions

FDE

FDE

27

Standard vs. linear CTRW

t

x

(x): how far

w(t): how long

Linear CTRW

28

3. Walking across an interface

“Physics-based” Monte Carlo sampling rules

Sample a random jump:

t~w(t) and x~(x)

Start in a given layer

The walker lands in the same layer

The walker crosses the interface

“Reuse” the remaining portion of the jump in the other layer

29

Re-sampling at the interface

xx’,t’ v’=x’/t’

t=x/v’

x’ = -1(Rx), t’ = W-1(Rt) Rx = (x), Rt = W(t)

1 2

x = -1(Rx) - -1(Rx), t = W-1(Rt) - W-

1(Rt)

,w ,w

x,t

v’

v’’

30

3. Walking across an interface

Analytical boundary conditions at the interface

),(),()(),()(2

)(),(

0),()0,(),(

uxPuxMxuxMxx

xuxJ

uxJx

xPuxuP

JJ Mass conservation:

Concentration ratio at the interface: PuMPuM )()( P(x,u)

31

Local particle velocity

Normal diffusion: M(u)=1/

Equal velocities: ()+=()-

Anomalous diffusion: M(u)=u1-/

Equal velocities: ()+=()- and +=-

(x/t)-

(x/t)+

Different concentrations at the

interface

Equal concentrations at

the interface

(x/t)-

(x/t)+Monte Carlo simulation:

Local velocity: v=x/t

PuMPuM )()(

Boundary conditions

32

5. Fractured porous media

Experimental NMR measures [Kimmich, 2002] Fractal streams (preferential water flow)

Anomalous transport

Develop a physical model Geometry of paths

df

Schramm-Loewner Evolution

33

5. Fractured porous media

Compare our model to analogous CTRW approach [Berkowitz et al., 1998] Identical spread <x2>~t( depending on df)

Discrepancies in the breakthrough curves

Anomalous diffusion is not universal There exist many possible realizations and descriptions

0 5 10 15 20 25 30 35 400

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Both behaviors observed in different physical contexts

(t)

t

CTRW

Our model