24.02.2014 Views

Partial Differential Equations - Modelling and ... - ResearchGate

Partial Differential Equations - Modelling and ... - ResearchGate

Partial Differential Equations - Modelling and ... - ResearchGate

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

218 J. Hao et al.<br />

<strong>and</strong> solve for V n+5/6<br />

G<br />

<strong>and</strong> ω n+5/6 from<br />

⎧<br />

⎪⎨ 〈e i , u n+5/6 − V n+5/6 −−−−−→<br />

n+5/6<br />

G<br />

− ω G n+4/6 ⊥<br />

x 〉 n+4/6 B<br />

=0, for i =1, 2,<br />

h<br />

⎪⎩ 〈 −−−−−→<br />

G n+4/6 ⊥<br />

x , u n+5/6 − V n+5/6 −−−−−→<br />

n+5/6<br />

G<br />

− ω G n+4/6 ⊥<br />

x 〉 n+4/6 B<br />

=0.<br />

h<br />

(30)<br />

Step 7. Finally, take V n+1,0<br />

G<br />

= V n+5/6<br />

G<br />

<strong>and</strong> G n+1,0 = G n+4/6 . Then predict<br />

the final position <strong>and</strong> translation velocity as follows:<br />

For k =1,...,N,<br />

Call Adhesive Dynamics Algorithm,<br />

̂V n+1,k<br />

G<br />

= V n+1,k−1<br />

G<br />

+ F r (G n+1,k−1 )△t/2N, (31)<br />

Ĝ n+1,k = G n+1,k−1 +(̂V n+1,k<br />

G<br />

V n+1,k<br />

G<br />

= V n+1,k−1<br />

G<br />

G n+1,k = G n+1,k−1 +(V n+1,k<br />

G<br />

+ V n+1,k−1<br />

G<br />

)△t/4N, (32)<br />

+(F r (Ĝn+1,k )+F r (G n+1,k−1 ))△t/4N,<br />

(33)<br />

+ V n+1,k−1<br />

G<br />

)△t/4N, (34)<br />

enddo;<br />

<strong>and</strong> let V n+1<br />

G<br />

= Vn+1,N G<br />

, G n+1 = G n+1,N ; <strong>and</strong> set u n+1 = u n+5/6 ,<br />

ω n+1 = ω n+5/6 .<br />

In Algorithm 1, we have t n+s =(n + s)△t, W n+1<br />

g = W 0,h g 0,h(t n+1 ), Λ n+s<br />

0,h =<br />

Λ 0,h (t n+s ), B n+s<br />

h<br />

is the region occupied by the particle centered at G n+s ,<br />

<strong>and</strong> F r is the combination of a short range repulsion force which prevents the<br />

particle/particle <strong>and</strong> particle/wall penetration (see, e.g., [GPHJ99, GPH + 01])<br />

<strong>and</strong> the force obtained from the adhesive dynamics algorithm for the cell<br />

adhesion. Finally, α <strong>and</strong> β verify α + β =1;wehavechosenα =1<strong>and</strong>β =0<br />

in the numerical simulations discussed later.<br />

The degenerated quasi-Stokes problem (21) is solved by a preconditioned<br />

conjugate gradient method introduced in [GPP98], in which discrete elliptic<br />

problems from the preconditioning are solved by a matrix-free fast solver from<br />

FISHPAK by Adams et al. in [ASS80]. The advection problem (22) for the<br />

velocity field is solved by a wave-like equation method as in [DG97]. The<br />

problem (24) is a classical discrete elliptic problem which can be solved by<br />

the same matrix-free fast solver. To enforce the rigid body motion inside<br />

the region occupied by the particles, we have applied the conjugate gradient<br />

method discussed in [PG02, PG05].<br />

4 Numerical Results <strong>and</strong> Discussion<br />

We consider the detachment of 20 cells in shear flow as the test problem for<br />

cell adhesion model at the initial stage of the adhesion. The computational<br />

domain is Ω =(0, 23)×(0, 10) (unit: 10 µm). Cells have the shape of an ellipse,

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!