nsf: ef-0830117 constant-number monte carlo simulations of nanoparticles agglomeration yoram cohen,...
TRANSCRIPT
NSF: EF-0830117
Constant-Number Monte Carlo Simulations of Nanoparticles Agglomeration
Yoram Cohen, Haoyang Haven Liu, Sirikarn Surawanvijit, Robert Rallo and Gerassimos Orkoulas
Center for Environmental Implications of nanotechnology
and
Department of Chemical and Biomolecular Engineering
University of California, Los Angeles
http://www.cein.ucla.edu/
This materials is based on work supported by the National Science Foundation and Environmental Protection Agency under Cooperative Agreement # NSF-EF0830117. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation or the Environmental Protection Agency.
NSF: EF-0830117
OUTLINE
Motivation Toward predictive models of NP
agglomerationo Basic approacho Monte Carlo numerical simulationso Comparison of predictions
with experimental datao Dependence of NP agglomeration
on basic system parameterso Future work
NSF: EF-0830117
• eNMs may be released to the environment throughout their life-cycle
• Preliminary in vitro (with various cell lines) and in-vivo studies with simple organisms (e.g., zebrafish) suggest that certain eNMs may be toxic at certain exposure concentration levels
• The transport and fate of eNMs in the environment is governed by their agglomeration state
• The toxicity of eNMs may be impacted by their primary size and their agglomeration state
• The removal of eNMs from aqueous streams can be facilitated by controlling their aggregation state
Motivation
Nanoparticle Toxicity
ExposureFate & Transport
Particle Size Distribution
Particle-Cell Interactions
Nanoparticle Aggregation
NSF: EF-0830117
Environmental Multimedia Fate & Transport of eNMs
The transport and fate of nanoparticles is governed by their agglomeration state
Atmosphere
Water Body
SedimentSoil
eNMs input
Aerosolization
Sedimentation
Dry/wet Deposition
Resuspension
FloodingAdsorption
Resuspension
Aggregation
DisaggregationAdsorption
Desorption
Dispersion Convection
Dry/wet Deposition
Runoff
NSF: EF-0830117
Environmental Intermedia Transport of Particles
Dry Deposition
Wind Soil Resuspension
Wet Scavenging
Aerosolization
NSF: EF-0830117
Atmospheric Deposition of Particles onto Water Surfaces
• The dry deposition velocity of particles varies with particle size
Dep
ositi
on V
eloc
ity (
cm/s
)
Particle Diameter, (µm)
1 nm
Diffusion
Impaction
1 10
10-2
10-3
10-1
1
10
102
0.1
0.01
Williams, R.M., A model for the dry deposition of particles to natural water surfaces. Atmospheric Environment (1967), 1982. 16(8): p. 1933-1938
The dry deposition velocity of particles varies with particle size
NSF: EF-0830117
Rain Scavenging of Nanoparticles
• Efficiency of NP removal from the atmosphere via wet deposition depends on particle size Cohen, Y. and P. A. Ryan, "Multimedia Transport of Particle Bound Organics: Benzo(a)Pyrene Test Case,” Chemosphere, 15, 31-47 (1986).
Cohen, Y. and P. A. Ryan, "Multimedia Transport of Particle Bound Organics: Benzo(a)Pyrene Test Case,” Chemosphere, 15, 31-47 (1986).
Efficiency of NP removal from the atmosphere via wet deposition depends on particle size
NSF: EF-0830117
Gravitational Sedimentation of Nanoparticles in Aqueous Media
NSF: EF-0830117
eNM Size Distribution in Aqueous Systems
DLS is the standard approach to quantifying the size distribution of nanoparticles The reliability of DLS measurements is dependent on the NP
concentration and suspension stability Suspension stability is impacted by NP agglomeration
(aggregation)/disaggregation which directly affect particle gravitational sedimentation
Detector
90°~40μm
NSF: EF-0830117
1000 ppm suspension
Nanoparticle powder
Detector
DLS20ppm
suspension
Sonicate for 30 minutes in T-
controlled bath
Sonicate for 5 minutes
Time delays between consecutive steps ~5 s
Dilute
IS adjustedpH adjusted
aqueous solution
NPs: TiO2 (21 nm, IEP=6.5, 21% A/79%R) CeO2 (15 nm, IEP=7.8)
NSF: EF-0830117
• van der Waals Attraction
• EDL Repulsion– 2 cases, quantified
by inverse Debye Length
𝜅𝑟>5
𝜅𝑟<5 Φ𝐸𝐷𝐿 ,𝑖𝑗=4𝜋𝜀𝜀𝑜𝑟 𝑖𝑟 𝑗𝑌 𝑖𝑌 𝑗𝜓𝑜2 (𝑘𝑇𝑒 )
2 exp (−κH )𝐻+𝑟 𝑖+𝑟 𝑗
Φ𝐸𝐷𝐿 ,𝑖𝑗=4𝜋𝜀𝜀𝑜𝜓𝑜2 𝑟 𝑖𝑟 𝑗
𝑟 𝑖+𝑟 𝑗
ln(1+exp (−𝜅𝐻 ))
Φ𝑣𝑑𝑊 ,𝑖𝑗=−𝐴𝐻
6[
2𝑟 𝑖𝑟 𝑗
𝑅2− (𝑟 𝑖+𝑟 𝑗 )2+
2𝑟 𝑖𝑟 𝑗
𝑅2− (𝑟 𝑖−𝑟 𝑗 )2+𝑙𝑛
𝑅2− (𝑟 𝑖+𝑟 𝑗 )2
𝑅2− (𝑟 𝑖+𝑟 𝑗 )2 ]
eNP eNP
Particle-Particle Interactions (Classical DLVO)
NSF: EF-0830117
DLVO Theory (slide shows forumlas for types of interactionType of Interactions Expression
EDL<5
EDL>5
vdW2 2
, 2 2 2 2 2 2
2 2 ( )ln
6 ( ) ( ) ( )i j i j i jH
vdW iji j i j i j
rr rr R r rA
R r r R r r R r r
r
2, 4 ln 1 expi j
EDL ij o oi j
rr
r r
r
2
,
exp4EDL ij o i j i j
i j
kTrr YY
e H r r
NSF: EF-0830117
Particle-Particle Interactions
• Classical DLVO only accounts for vdW and EDL• Classical DLVO assumes hard sphere
– O.K. for environmental application as most frequently used eNMs are spherical
– Non-spherical particles exist• Nano-rod, nano-wire, etc.
• DLVO does not account for:– Steric, hydration, magnetic, etc.
• Modified DLVO can be utilized to account for additional interaction energies and particle shape (e.g., sphericity)
NSF: EF-0830117
Size distribution of NPs in Aqueous Systems
• Basis: Smoluchowski Coagulation Theory
– is the agglomeration frequency function: – is the collision frequency: – is the inverse sticking coefficient:– is the total interaction energy between and
• Estimated using classical DLVO theory
– Time step to next agglomeration event:
2
ij
tC N K
NSF: EF-0830117
Nanoparticle Brownian Motion & Settling
• Stokes’ Settling velocity
• Diffusion length22x D t
6Bk T
Dr
22
9p f
sed
g rv
r <x>
𝑑𝑠𝑒𝑑
𝑑𝑠𝑒𝑑=𝑣𝑠𝑒𝑑 ∙ Δ 𝑡
NSF: EF-0830117
Monte Carlo Simulation of eNM Agglomeration
.
NSF: EF-0830117
Constant-Number MC Simulations of Particles in a Box
Box is expanded to maintain the particle concentration upon aggregation events and replenishment of particles to maintain a constant number
NSF: EF-0830117
Simulations of Nanoparticles Agglomeration
Dynamic Monte Carlo Simulation Solver
Primary NP Information(e.g., primary size, surface chemistry)
Solution Chemistry/Media Parameters
(e.g., ionic strength, pH, temperature, dielectric
constant)
Output:- Particle size distribution (PSD)
- NP concentration
Measured or Calculated Model Parameters
(e.g., dp, zeta potential, IS Hamaker constant)
Aggregation Model:- DLVO
- Sedimentation- Particles in a “box”
Computational (Constant-Number Monte Carlo) model of NP agglomeration making use of the DLVO theory accounting for NP sedimentation
Computational Cluster: 10 Nodes with a total of 20 Intel Quad-Core Xeon processors (2.2 – 3.0 GHz) with 176 GB RAM
NSF: EF-0830117
Importance of Including Sedimentation in Model Simulations
Average of 10 simulations of 5000 particles
CeO2TiO2
ζCeO2 = -24.5 mV ζTiO2 = -29 mV AH, = 42 zJAH, = 21 zJ
pH = 8, IS= 0.065 mM
NSF: EF-0830117
Convergence of Simulations
<d>
10 (
nm)
Number of Simulation Particles
<d>
n (n
m)
Number of Simulations, n
Expand box to maintain mass concentration
Determine diffusion and settling distances during the previously determined time step
for all NPs
Final NP Dispersion
Distribute NPs in a box
End Start
t<tfinal
Calculate agglomeration
frequency for all NP pairs
Select a pair of NPs for agglomeration
based on their agglomeration
frequency
Calculate size and position of
agglomerated pair
Calculate the time step between the
agglomeration events
Replenish particles based on PSD
sampling to maintain constant NP numbers
Replace particles based on periodic
boundary condition
Replace particle based on PSD of settled particles
NPs diffuse or settle out
of box?
Yes
No
Diffuse
Settle
Neither
Number of Simulation Particles
Average of 10 simulations
Sm
ean
part
icle
siz
e (%
)
Sm
ean
part
icle
siz
e, n
m
NSF: EF-0830117
Comparison of Experimental and Simulation Results
NSF: EF-0830117
eNP(a) Type
z [mV] (pH) IS [mM] dp[nm] dexp [nm] dsim [nm] % abs. error(b)
Jiang, et al. TiO2 38 (3.3) 1 15 80 96 20.5TiO2 36 (3.8) 1 15 85 102 19.8TiO2 34 (4.45) 1 15 87 108 23.7TiO2 28.5 (5.3) 1 15 233 252 8.1TiO2 -30 (7.8) 1 15 218 251 15.5TiO2 -38 (8.2) 1 15 162 121 25.2TiO2 -43 (8.7) 1 15 92 90 1.9TiO2 -47.5 (9.65) 1 15 93 85 8.7TiO2 -45 (10.4) 1 15 98 78 20.2TiO2 36 (4.6) 0.01 15 90 77 14.6TiO2 42 (4.6) 1 15 90 107 18.8TiO2 40 (4.6) 5 15 160 178 11.3TiO2 36 (4.6) 10 15 500 392 21.6
French, et al. TiO2 35 (4.5) 4.5 5 90 109 20.8TiO2 35 (4.5) 8.5 5 500 632 13.5TiO2 35 (4.5) 12.5 5 700 628 10.3
Ji, et al. TiO2 30.2 (6.1) 1 21 200 202 1.0Present Study TiO2 41 (3) 0.37 21 163 162 0.6
TiO2 -30 (8) 0.027 21 173 175 1.2TiO2 -35 (10) 0.12 21 172 171 0.6CeO2 32 (3) 0.37 15 271 269 0.7CeO2 -23.5 (8) 0.027 15 266 264 0.8CeO2 -30 (-30) 0.12 15 240 243 1.3
(b) % abs. error (a) eNP – Engineered Nanoparticle
Summary of Experimental & Simulation Conditions
NSF: EF-0830117
Particles Size Distributions (t=24 h)
NSF: EF-0830117
Dependence of TiO2 Agglomeration on pH
Simulations:
NSF: EF-0830117
Dependence of Agglomerate Size on Ionic Strength
NSF: EF-0830117
Dependence of NP Agglomeration on the Hamaker Constant
NSF: EF-0830117
Dependence of Agglomerate Size on Primary NP Size
NP primary size ↑ PSD tail of small aggregates ↑ Average NP aggregate size (in suspension) ↓
For present primary size range:
NSF: EF-0830117
Summary and Future workMonte Carlo (MC) simulations of NP agglomeration based on the Smoluchowski equation and classical DLVO theory demonstrated reasonable quantitative predictions of NP agglomeration (average size and size distribution) over a range of solution conditions (pH= 3-10, IS= 0.03-12.5 mM for TiO2 and CeO2 NPs)
The present approach can be extended to include various modifications/extensions of the DLVO theory
With extension and additional validation of the current modeling approach it will be feasible to develop a practical parameterized model of NP agglomeration
• New experimental DLS data are being generated over a wide range of conditions specifically for extended model extension and validation
• A machine learning approach is being developed to guide the task of data generation and parameterized model development
NSF: EF-0830117
Questions?