Mesoscopic simulation of the dynamics of confined complex ...Mesoscopic simulation of the dynamics...
Transcript of Mesoscopic simulation of the dynamics of confined complex ...Mesoscopic simulation of the dynamics...
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Mesoscopic simulation of the dynamics ofconfined complex fluids
M. D. GrahamDept. of Chemical and Biological Engineering
Univ. of Wisconsin-Madison
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
• Microfluidics and DNA• Richard Jendrejack• Hongbo Ma• Juan Hernandez• Yeng-Long Chen•Eileen Dimalanta• Kyubong Jo•Chris Stoltz• Prof. Juan de Pablo• Prof. David C. Schwartz• NSF: UW-NSEC
• Cellular fluid mechanics• Sam Anekal• Patrick Underhill• Pratik Pranay• Juan Hernandez• NSF: CBET
Group/Collaborators/support
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Microfluidics and complex fluids
genomic DNA (Schwartz)
biological fluids
particle-based assays(Edelstein et al.)
droplet reactors (Ismagilov et al.)
dispersions (Anna)
Surface-based biochemicalassays (Jensen et al.)
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Confined polymers: fundamental issues
• Diffusion• Relaxation• Polymer conformations in flow• Effect of flow on adsorption• Mechanisms of center of mass migration and apparent slip in polymer solutions
Goal: simulations and theory to shed light on these issues
2Rg2h
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Optical mapping for genomics(Schwartz lab; OpGen Inc.)
DNA is stretched by flow and electrostatically adsorbed to a surfacewithin a microfluidic device.
The optical map can be usedas a scaffold to orient short, detailedsequences (contigs) and thus accel-erate construction of a full genome.
Restriction enzymes cleave the adsorbed DNA molecules at well-defined positions, providing landmarks along the sequence, an “optical map” of the genome.
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Adsorption depends on flow protocol
Reservoir of DNA solution
Syringe pump
Advancing meniscus
Dimalanta 2004
Negligible adsorption
Significant adsorption
Cationic capture surface
Cationic capture surface
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Coarse-grained models of polymers
atomistic (~ Å)
worm-like chain (~ 100 nm)(bending potential)
bead-spring chain (~µm)
Representation Resolution
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Force balance for a chain in solution
dri = v i + Mij ⋅ f j + kT∂∂rj
⋅M ji
⎛
⎝⎜⎞
⎠⎟j=1
Nb
∑⎡
⎣⎢⎢
⎤
⎦⎥⎥dt + 2 Bij ⋅dw j
j=1
Nb
∑
imposed flow
spring forceexcluded volume
wall forcethermal fluctuations
Mij =1ζ(Iδ ij +ζΩij )
mobility tensor
BijBkjj=1
Nb
∑ = kTMik
fluctuation-dissipationtheorem
′ v = Ωiji, j=1
Nb
∑ ⋅ fj
hydrodynamic interaction tensor
→ velocity field generated by motion of macromolecule→ Solution to a Stokes problem
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Coupling polymer dynamics to fluid dynamics
• Each “bead” is a regularized point force• Beads experience Stokes drag and Brownian force • Noise and drag are coupled!• Two-way coupling between fluid motion and polymer motion
• Approaches to this coupling:• Continuum solvent
• Green’s function methods (us, John Brady & co.)• Stochastic Navier-Stokes (Paul Atzberger)• Lattice Boltzmann (Tony Ladd)
•Particulate solvent• Dissipative particle dynamics (Bruce Caswell)• Stochastic rotation dynamics (Ron Larson)
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Coupling polymer dynamics to fluid dynamics:Green’s function methods
Computation time for a naïve algorithm in an unboundedgeometry!!: N3 -- not good!
But…
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Coupling polymer dynamics to fluid dynamics:Green’s function methods
Fast Stokes solver for (regularized) point forces in yourfavorite geometry => V = M . F
+Chebychev polynomial approximation to M1/2 (Fixman 1986)
+Derivative-free stochastic integration scheme (Fixman 1978, Hinch etal.1995)
Matrix-free linear algebra routines (GMRES etc.)+
=Brownian dynamics with hydrodynamic interactions in O(N) (or so)time in your favorite geometry.
Jendrejack et al., 2000, 2003, Hernandez-Ortiz et al. 2007, 2008
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
GGEM: fast method for point particles inStokes flow
• Near-field details– easy to regularize– Exponentially decaying
(short range)– Easy to calculate, i.e.
analytical expression– O(Nparticles)
• Long-range interactions– Correct boundary
conditions at walls,– Solution through O(Nmesh)
CFD method– Generalizes to
Stokesian dynamics,immersed boundarymethod
Jendrejack et al ‘03, Hernandez et al ‘07, Anekal et al. ‘08
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
GGEM
Force density: ρ(x) = f(xv ) δ (x − xv ) − g(x − xv )[ ] + g(x − xv ) v
N
∑“local” “global”
• Conventional Ewald sum approachto electrostatics:
• Local: exponentially decaying fcns.
• Global: smooth solution in Fourierspace
• This idea is not restricted to periodicdomains.
ρ = ρl + ρg ⇒ v = v l + vg
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
GGEM: local
Analytical FREE-SPACE solns.(i.e. no Boundary Conditions)
Neighbor list with size given by the smearing function
v l (x) = Gl (x − xv ) ⋅ f(xv )v
N
∑O(N) Green’s function-like calculation:
Gl (x) =18πη
δ +xxr2
⎡⎣⎢
⎤⎦⎥erfc(αr)
r−18πη
δ −xxr2
⎡⎣⎢
⎤⎦⎥2απ 1/2
e(−α2r2 )
⎥⎦⎤
⎢⎣⎡ −= − 22)(
2/3
3
25)(
22
rerg r απα αSmearing function:
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
GGEM: global
Find the solution on a mesh with αΔx << 1
Sparse/iterative Methods: O(Mesh points)
−∇pg +η∇2vg + ρg = 0∇ ⋅vg = 0
Solution of Stokes Equations:
with no-slip boundary Conditions: vg = −v l
After the O(Mesh) calculation a O(N) interpolation is performed:
GGEM = O(N)
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Point force regularizationSmearing function g is aregularized delta function:Use it to replace the point force,with
GlR (x) = 1
8πηδ +
xxr2
⎡⎣⎢
⎤⎦⎥erf(ξr)r
−erf(ξr)r
⎡⎣⎢
⎤⎦⎥+
18πη
δ −xxr2
⎡⎣⎢
⎤⎦⎥2επ 1/2
e(−ξ2r2 ) −
2απ 1/2
e(−α2r2 )⎡
⎣⎢⎤⎦⎥
Regularized:
Beadhydrodynamic
radius: a
α−1
ξ−1
ξ−1 =3aπ
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Jendrejack et al., JCP (2000,2002)
L = 21 µm (λ-phage)Ns = 10 springsNk,s = 19.8 Kuhns/spring
v = 0.0019 µm3/Kuhna = 0.077 µm (h* = 0.15)bk = 0.106 µm (Nk = 198Kuhns)
equilibrium stretch (1.5 µm)relaxation time (4.1 s)diffusivity (0.0115 µm2/s)
Smith and Chu, 1998Smith, Perkins and Chu, 1996
match
model parameters experimental data
• Model parameters determined by direct comparison to equilibrium experimental data for 21 µm DNA.
• Quantitative agreement with experiment for both transient and steady flows over a wide range of Weissenberg numbers.
• Molecular weight dependence of properties is “built in” because HI and EV are included in model
• Expect good performance in microfluidic simulations
Parametrization with free solution data
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Expt
Channel Heights = 0.8 microns – 10 micronsλ-phange DNA: chain Lengths Ns = 10 – 120 (21 microns – 234 microns)
Square channels
Slit channels
• Static confinement transition at Rg,bulk/H ≈0.4
• Power-law scaling in highly confined regime as predicted by good solvent blob models(de Gennes, Douad, Brochard)
• Hydrodynamic screening in highly confinedchannels -> Rouse dynamics
• Good agreement with experimentalmeasurements of Doyle group for λ-DNA
Equilibrium results for confined chains
Stretch Diffusivity
Jendrejack et al. JCP, 2003, Chen et al. 2004
2/3
1/4
-2/3
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
• Polymer migrates from wall: hydrodynamic depletion layer forms• Confinement (wall effect) is crucial• Molecules move faster than average fluid velocity
Dilute solution: pressure-driven flow
0We =
We = 1.3, ˙ γ = 3.98s−1
Center of mass distribution42 µm DNA, H = constant = 9.3 Sbulk
Distance along channel vs. time
10.5 µm21 µm42 µm84 µm
We = 10, ˙ γ = 30.8s−1
We = 100, ˙ γ = 308s−1
Jendrejack et al 2003 ,2004
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
DNA migration in oscillatory flow
40 µm
Before Oscillation 8 min Later
Experiment (T2 DNA, We=50, 0.25 Hz, square channel,Jo & Schwartz 2005)
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Experimental resultsSimulation results
Before Oscillation 40 seconds 6 minutes
Chen et al 2005
Migration in oscillatory flow:experiment vs. simulation
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
velocityfluid.maxvelocityDNAavg.=fR )2//(max Hv=γ
Sugarman & Prud’homme (1988), Chen et al 2005
25 µm
Detection points at 25 cm and 200cm
detector
λ-DNA in microcapillary flow
Simulations: square channelExperiments: cylindrical channel
Comparison to capillary flow experiments
No adjustable parameters
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Hydrodynamic migration mechanism
The flow induced by each endof a horizontal relaxing chainpushes the other end awayfrom the wall. (Jendrejack et al.2004)y
A horizontally aligned stressletabove a wall induces a wall-normal flow (cf. Leal 1980, Smart& Leighton 1991, Hudson 2003)
dumbbell
equivalent stresslet
induced velocity
Point force above a wallred: vy>0, blue: vy<0.
Near-field
Far-field
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Kinetic theory for a single chain near awall: dumbbell model
number density(probability)
Migration tensor (3rd order)
Polymer stressKirkwood diffusivity(conformation-dependent)
Flux expression in low Re, point-dipole limit
Thick depletion layers in uniformshear flow
Migration and diffusionbalance at steady state:
Depletion layer thickness
can be much larger than Rgat high Weissenberg number
Ma et al. (2005, 2006)
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Spatial development of depletion layer
Depletion layer thickness:
Uniform shear flow in semi-infinite domain
These balance in entrance region We=10,FENE-P model
Weak singularity (smoothed by diffusion)
Migration is fastest near the wall => molecules “pile up” to form a moving, spreading shock-like concentration profile
Ld
Lx
Similarity solution
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Spatial development of depletion layerPlane Poiseuille flow, full numerical solution
We = 10
B/Ld ≈ 20
FENE-P Dumbbell
• From theory:
• Entrance length forFENE-P model: Lx ~ We3
• Fully developed profileonly when δh ≈ Ld:
“pileup”residence time
>>diffusion time across Ld
y
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Concentration and geometry effects
Hernandez-Ortiz et al. (2008)
Equilibrium Wi = 20, φ = 10-4, HI Wi = 20, φ = 0.12, HI
Complex geometry >104 beads FEM for ud, Fixman methods for
random term
Migration away from wallsweakens with concentration Dilute: Significant depletion
from cavity Concentrated: Very weak
depletion from cavity Consistent with experiments
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Adsorption depends on flow protocol
Reservoir of DNA solution
Syringe pump
Migration suppressesadsorption
“Fountain flow”at meniscus promotesadsorption
Cationic capture surface
Cationic capture surface
Consider reference frame moving with meniscus:
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Blood flow in the microcirculation Red blood cells flow in center of arterioles,
white blood cells “marginate” What do drug delivery particles do? Many diseases/injurues disrupt normal blood
flow (heart disease, diabetes, hemorrhage) Minute amounts of “drag reducing” polymer
added to the bloodstream can improve tissueoxygenation
Dynamics of blood flow in small vessels iscomplex and poorly understood
Distribution of red and white blood cells in microcirculation
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Polymer additives decrease cell-free layer thickness in invitro experiments
RBCs in buffer: fewcells near bottomwall
buffer+50 ppm PEO:many more cells nearbottom wall
Proposed particles for drug delivery: (a) liposomes, (b)wormlike micelles, (c-d) nonspherical polymeric particles
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
GGEM/IBM GGEM can be used as the basis for a variant of the immersed boundary method, valid
for Stokes flow. The local interactions are not put on the mesh but are included in thelocal (analytical) solution.
This can also be viewed as a low budget boundary integral method. Is there a convergence proof for this variant of IBM?
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
-8 -6 -4 -2 0 2 4 6 80
0.5
1
1.5
2
Δx/a
Δy/a
SimulationLac et. al.(2007)
Pair collision of elasticcapsules in shear
Δy
Δx
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Red blood cells in capillaries - preliminary!
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Flowing Complex Fluids Research Group Department of Chemical and Biological EngineeringUniversity of Wisconsin-Madison
Summary• Particle level simulation and theory for flexible polymers (DNA) in microchannels, including hydrodynamic effects of confinement
• Fast simulation methods allow exploration of concentration, geometry, flow
• DNA dynamics in microchannel flow- Thick depletion layers in flow: dominated by wall-modified hydrodynamic interactions- nontrivial concentration dependence in complex flows -- not understood
• Flux expression for solution of dumbbells near a wall- depletion layer thickness normal stress- slow evolution of depletion layer
• Similar framework holds for other systems, such as blood.