CIG Workshop 2005 Boulder, Colorado K. Stemmer, H. Harder and U. Hansen stemmer@uni-muenster.de

Preview:

DESCRIPTION

Münster University, Germany Department of Geophysics. A finite volume solution method for thermal convection in a spherical shell with strong temperature- und pressure-dependent viscosity. CIG Workshop 2005 Boulder, Colorado K. Stemmer, H. Harder and U. Hansen stemmer@uni-muenster.de. - PowerPoint PPT Presentation

Citation preview

A finite volume solution method for thermal convection in a spherical shell with strong temperature- und pressure-dependent viscosity

CIG Workshop 2005Boulder, Colorado

K. Stemmer, H. Harder and U. Hansen

stemmer@uni-muenster.de

Münster University, GermanyDepartment of Geophysics

Outline

Motivation: Importance of mantle rheology

Basic principles of thermal convection with variable viscosity Mathematical model Numerical model

Simulation results: Thermal convection in a spherical shell Temperature-dependent viscosity Temperature- and pressure-dependent viscosity

Conclusions

Münster University

MotivationImportance of mantle rheology

Laboratory experiments of mantle material: viscosity is temperature-, pressure- and stress-dependent

Many models have constraints: Cartesian isoviscous / depth-dependent viscosity

High numerical and computational effort for lateral variable viscosity mode coupling sophisticated numerical methods

Münster University

Thermal convectionmathematical model

Rayleigh-Bénard convection

0=⋅∇ ur

continuity equation

equation of motion

02 =−∇−∇+∂∂

RaRa

TTutT Qr

heat transport equation

( )[ ] 0)( =∇−+∇+∇⋅∇ peRaTuu r

T rrrη

Rayleigh numberref

TdgRa

κηαρ 3Δ

=ref

Q k

QdgRa

ηκαρ 5

=

Arrhenius equation ( ) ( )))(ln())(ln(exp, 0 refprefT TrRTTpT −−Δ+−Δ= ηηη

.

pηΔTηΔ

viscosity contrast with pressure

viscosity contrast with temperature

Münster University

Thermal convection with lateral variable viscositynumerical model

Implemented methods: Discretization with Finite Volumes (FV) Collocated grid Equations in Cartesian formulation Primitive variables Spherical shell topologically divided in 6 cube surfaces Massive parallel, domain decomposition (MPI)

Time stepping: implicit Crank-Nicolson method Solver: conjugate gradients (SSOR) Pressure correction: SIMPLER and PWI

Münster University

control volume

Thermal convection with lateral variable viscositynumerical model

grid generation lateral grid

Münster University

Advantages of this spatial discretization: Efficient parallelization No singularities at the poles Approximately perpendicular grid lines Implicit solver (finite volumes)

discretization of the viscous term

Problem: required: derivatives of velocities in x-,y- und z-direction

available: curved gridlines (not in x-,y- und z-direction)

( )[ ]Tuu )(rr

∇+∇⋅∇ η

Solution:transformation of the viscous termapplying Gauß / Stokes theorem and lokal CV coordinate systems simplification of integrals

( ) ( )uurr

×∇×∇+∇⋅∇ ηη2

CV: control volume

Thermal convection with lateral variable viscositynumerical model

Münster University

SduIVS

rr∫ ∂=

∇= η21

( )∫ ∂=×∇×=

VSuSdIrr

η2

Gauss integral theorem

known Laplacian solution

applying Stokes theoremchange to local orthonormal basis

to simplify notation: ( ) ( )i

i

i S

S

S hcgbfa

c

b

a

urrrr

++=⎟⎟⎟

⎜⎜⎜

⎛=×∇ :

iSlocal orthonormal basis of the CV surface( )hgfrrr,,

∫ ∫ ∂=⋅=⋅∇

V VSSdudVurrr

Thermal convection with lateral variable viscositydiscretisation of the viscous term

( )[ ] ( ) ( )dVudVudVuuV VV

T ∫ ∫∫ ×∇×∇+∇⋅∇=∇+∇⋅∇rrrr

ηηη 2)(

stress tensor 1I 2I

viscous term:

Münster University

( ) i

SS

iiS SVS

dS

c

b

a

n

n

n

uSdI

ii

ii

⎟⎟⎟

⎜⎜⎜

⎛×

⎟⎟⎟

⎜⎜⎜

⎛=×∇×= ∑∫∫

=∂=

6

3

2

1

2 ηηrr

solution of integral :

( )TSinnn 321 ,,normal vector

many terms are vanishing due to the use of local coordinates

( ) ( ) SSS SN

NN N dShafcdShafc ∫∫ +−+−++rrrr

ηη

( ) ( ) BBB BT

TT T dSgafbdSgafb ∫∫ −+++−+rrrr

ηη

( ) ( ) WWW WE

EE E dShbgcdShbgc ∫∫ −+++−=rrrr

ηη

remains the calculation of the curl of velocities on the CV surfaces

Thermal convection with lateral variable viscositydiscretisation of the viscous term

2I

Münster University

integration along selected paths

Thermal convection with lateral variable viscositydiscretisation of the viscous term

linear approximation of line integrals

Calculation of the curl of velocities on the CV surfaces:

applying Stokes theorem

( ) dzwdyvdxuwdzvdyudxrdu ++≈++=⋅∫ ∫rr

i

ii

Sc

b

a

Sj gf

j hf

j hg

Sh

g

f

cS

bS

aS

rdu

rdu

rdu

rdu

rdu

rdu

⎟⎟⎟

⎜⎜⎜

⎛=

⎟⎟⎟⎟⎟

⎜⎜⎜⎜⎜

=⎟⎟⎟⎟

⎜⎜⎜⎜

∑ ∫∑ ∫∑ ∫

∫∫∫

=

=

=

4

1 ,

4

1 ,

4

1 ,

rr

rr

rr

rr

rr

rr

a,b,c

∫∫ ⋅=×∇LS

rduSdurrrr

( )

( )

( )32,31,333,3

23,21,222,2

13,12,111,1

1

1

1

RvTuTDT

w

RwTuTDT

v

RwTvTDT

u

PPP

PPP

PPP

−−−=

−−−=

−−−=

entries adjacent R

term driving D

weightcentral T

i

i

ji ,

coupling of velocity components central weight is a vector

Münster University

Thermal convection with lateral variable viscositypressure weighted interpolation (PWI)

Solution: mathematical principle [Rhie and Chow, 1983] small regularizing terms are added that excludes spurious modes perturb the continuity equation with pressure terms regulating pressure terms do not influence the accuracy of the discretisation

( ) 2

1

11 42

1 +

+− Δ++=n

P

pressurenP

nP

nP p

cw

auuu

pressure of tiondiscretiza central :p

operator diffusion of weightcentral :cw

Δ

pressure is defined to an intermediate time levelpressure correction: fluxes are pertubated with pressure terms

Problem: insufficient coupling )()1( 211 tgcu jj

j+−= 21)1(2

jjj cp +−=

checkerboard oscillations

Münster University

Thermal convection in a spherical shelltemperature-dependent viscosity, basal heating 5102/1 =Ra

residualtemperaturedt = +/- 0.1

010=ΔT

η 310=ΔT

η 610=ΔT

η

temperatureisosurfacesand slices

Münster University

5102/1 =Ra010=Δ

Tη 310=Δ

Tη 610=Δ

T=0.25T=0.60 T=0.83

Thermal convection in a spherical shelltemperature-dependent viscosity, basal heating

Münster University

Three regimes: 1) mobile lid2) transitional (sluggish)3) stagnant lid

velocities minimum with depth of lateral velocities

Thermal convection in a spherical shelltemperature-dependent viscosity, basal heating

Münster University

100,10,10 54

0 =Δ=Δ= pTRa ηη

„highviscosityzone“

Thermal convection in a spherical shelltemperature- and pressure-dependent viscosity

Temperature dependence and pressuredependence of viscosity compete each other!

Münster University

0=QRa 4105⋅=QRa

isosurfaces:

100,10,10 54

0 =Δ=Δ= pTRa ηη

slices:

red = high viscousblue = low viscous

„high viscosity zones“

Thermal convection in a spherical shelltemperature- and pressure-dependent viscosity

)ln(η

1.0/−+=Tδ

Münster University

Conclusions

Mantle Convection Importance of spherical shell geometry Importance of mantle rheology ... ?

…thanks for your attention!

High numerical and computational effort for lateral variable viscosity

BUT: temperature-dependent viscosity has a strong effect on…

…convection pattern

…heat flow

…temporal evolution

Münster University, GermanyDepartment of Geophysics

stemmer@uni-muenster.de

Recommended