Single Particle Dynamics

Overview

This module implements equations for the dynamics of single aerosol particles, covering the fundamental physics that govern particle motion, transport, and interactions with gas molecules. These equations are essential for understanding aerosol behavior in the atmosphere, including sedimentation, diffusion, and electrical migration.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, 2nd Edition, Chapter 9: Dynamics of Single Aerosol Particles, pp. 396-433. John Wiley & Sons, Inc.

Aerosol.SingleParticleDynamics — Function
SingleParticleDynamics(; name=:SingleParticleDynamics)

Comprehensive component combining all single particle dynamics equations.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9: Dynamics of Single Aerosol Particles.

This component calculates key aerosol particle properties:

  • Mean free path of air (Eq. 9.6)
  • Knudsen number (Eq. 9.1)
  • Cunningham slip correction factor (Eq. 9.34)
  • Terminal settling velocity (Eq. 9.42)
  • Relaxation time (Eq. 9.38)
  • Brownian diffusion coefficient (Eq. 9.73)
  • Particle mobility (Eq. 9.78)
  • Mean thermal speed (Eq. 9.87)
  • Particle mean free path (Eq. 9.89)
  • Aerodynamic diameter (Eq. 9.110)
source
Aerosol.MeanFreePath — Function
MeanFreePath(; name=:MeanFreePath)

Component for calculating the mean free path of air molecules.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eq. 9.6.

The mean free path is calculated from:

\[\lambda = \frac{2\mu}{p\sqrt{8M/(\pi R T)}}\]

where μ is dynamic viscosity, p is pressure, M is molecular weight, R is the gas constant, and T is temperature.

source
Aerosol.SlipCorrection — Function
SlipCorrection(; name=:SlipCorrection)

Component for calculating the Cunningham slip correction factor.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eq. 9.34.

The slip correction factor accounts for noncontinuum effects when the particle size approaches the mean free path of gas molecules:

\[C_c = 1 + \frac{2\lambda}{D_p}\left[A_1 + A_2 \exp\left(-\frac{A_3 D_p}{2\lambda}\right)\right]\]

where the constants A₁ = 1.257, A₂ = 0.4, A₃ = 1.1 are from Allen & Raabe (1982).

source
Aerosol.SettlingVelocity — Function
SettlingVelocity(; name=:SettlingVelocity)

Component for calculating particle terminal settling velocity and relaxation time.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eqs. 9.38, 9.42.

The terminal settling velocity is given by:

\[v_t = \frac{D_p^2 \rho_p g C_c}{18\mu}\]

The relaxation time (time constant for velocity adjustment) is:

\[\tau = \frac{\rho_p D_p^2 C_c}{18\mu}\]

source
Aerosol.BrownianDiffusion — Function
BrownianDiffusion(; name=:BrownianDiffusion)

Component for calculating Brownian diffusion properties of aerosol particles.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eqs. 9.73, 9.78, 9.87, 9.89.

The Stokes-Einstein-Sutherland relation gives the diffusion coefficient:

\[D = \frac{k T C_c}{3 \pi \mu D_p}\]

The particle mobility is:

\[B = \frac{C_c}{3 \pi \mu D_p}\]

The mean thermal speed of the particle is:

\[\bar{c}_p = \sqrt{\frac{8 k T}{\pi m_p}}\]

The particle mean free path is derived from D = ½ c̄p λp (Eq. 9.88):

\[\lambda_p = \frac{2D}{\bar{c}_p}\]

source
Aerosol.ParticleMobility — Function
ParticleMobility(; name=:ParticleMobility)

Component for calculating particle mechanical mobility.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eq. 9.78.

The particle mobility relates external force to drift velocity:

\[B = \frac{C_c}{3 \pi \mu D_p}\]

With the Einstein relation connecting mobility to diffusion:

\[D = B k T\]

source
Aerosol.ElectricalMobility — Function
ElectricalMobility(; name=:ElectricalMobility)

Component for calculating electrical mobility and migration velocity.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eqs. 9.49, 9.50, 9.51.

The electrical mobility relates particle velocity to electric field strength:

\[B_e = \frac{q C_c}{3 \pi \mu D_p}\]

The migration velocity in an electric field E is:

\[v_e = B_e E\]

source
Aerosol.StokesNumber — Function
StokesNumber(; name=:StokesNumber)

Component for calculating Stokes number and stop distance.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eqs. 9.98, 9.101.

The stop distance is the distance a particle travels before coming to rest:

\[s_p = \tau U\]

The Stokes number characterizes particle inertia relative to flow:

\[St = \frac{\tau u_0}{L} = \frac{D_p^2 \rho_p C_c u_0}{18 \mu L}\]

When St ≫ 1, particles cannot follow fluid streamlines (inertial impaction). When St ≪ 1, particles follow fluid motion closely.

source
Aerosol.AerodynamicDiameter — Function
AerodynamicDiameter(; name=:AerodynamicDiameter)

Component for calculating equivalent particle diameters.

Reference: Seinfeld, J. H. and Pandis, S. N. (2006) Atmospheric Chemistry and Physics, 2nd Edition, Chapter 9, Eqs. 9.102, 9.106, 9.108, 9.110, 9.114.

The volume equivalent diameter is the diameter of a sphere with the same volume:

\[D_{ve} = \left(\frac{6 V_p}{\pi}\right)^{1/3}\]

The aerodynamic diameter is defined for a sphere of unit density (1000 kg/m³) with the same settling velocity:

\[D_{ca} = D_p \sqrt{\frac{\rho_p}{\rho_p^\circ}} \sqrt{\frac{C_c(D_p)}{C_c(D_{ca})}}\]

In the continuum regime (large particles), this simplifies to:

\[D_{ca} \approx D_p \sqrt{\frac{\rho_p}{\rho_p^\circ}}\]

The electrical mobility equivalent diameter relates to mobility measurement:

\[D_{em} = D_{ve} \chi \frac{C_c(D_{em})}{C_c(D_{ve})}\]

source

Implementation

The implementation provides several components that can be used individually or combined through the comprehensive SingleParticleDynamics component.

Key Physical Quantities

The implementation calculates the following key quantities:

QuantitySymbolEquationDescription
Mean free pathλEq. 9.6Average distance traveled by air molecules between collisions
Knudsen numberKnEq. 9.1Ratio of mean free path to particle diameter
Slip correctionCcEq. 9.34Correction for non-continuum effects
Relaxation timeτEq. 9.38Time constant for velocity adjustment
Terminal velocityvtEq. 9.42Steady-state settling velocity
Diffusion coefficientDEq. 9.73Brownian diffusivity
Particle mobilityBEq. 9.78Mechanical mobility

State Variables

using Aerosol, ModelingToolkit, DataFrames, DynamicQuantities

sys = SingleParticleDynamics()
vars = unknowns(sys)
DataFrame(
    :Name => [string(v) for v in vars],
    :Units => [string(ModelingToolkit.get_unit(v)) for v in vars],
    :Description => [ModelingToolkit.getdescription(v) for v in vars]
)
11×3 DataFrame
RowNameUnitsDescription
StringStringString
1λ(t)1.0 mMean free path of air
2Kn(t)1.0 Knudsen number (dimensionless)
3C_c(t)1.0 Cunningham slip correction factor (dimensionless)
4m_p(t)1.0 kgParticle mass
5τ(t)1.0 sRelaxation time
6v_t(t)1.0 m s⁻¹Terminal settling velocity
7D_B(t)1.0 m² s⁻¹Brownian diffusion coefficient
8B(t)1.0 kg⁻¹ sParticle mobility
9c_p(t)1.0 m s⁻¹Mean thermal speed of particle
10λ_p(t)1.0 mParticle mean free path
11D_ca(t)1.0 mAerodynamic diameter (continuum limit)

Parameters

params = parameters(sys)
DataFrame(
    :Name => [string(p) for p in params],
    :Units => [string(ModelingToolkit.get_unit(p)) for p in params],
    :Description => [ModelingToolkit.getdescription(p) for p in params]
)
14×3 DataFrame
RowNameUnitsDescription
StringStringString
1μ1.0 m⁻¹ kg s⁻¹Dynamic viscosity of air
2M_air1.0 kg mol⁻¹Molecular weight of air
3T1.0 KTemperature
4π_val1.0 Pi (dimensionless)
5R1.0 m² kg s⁻² K⁻¹ mol⁻¹Molar gas constant
6P1.0 m⁻¹ kg s⁻²Pressure
7D_p1.0 mParticle diameter
8A11.0 Slip correction constant A1 (dimensionless)
9A21.0 Slip correction constant A2 (dimensionless)
10A31.0 Slip correction constant A3 (dimensionless)
11ρ_p1.0 m⁻³ kgParticle density
12g1.0 m s⁻²Gravitational acceleration
13k_B1.0 m² kg s⁻² K⁻¹Boltzmann constant
14ρ_p01.0 m⁻³ kgReference density (1 g/cm³)

Equations

equations(sys)
11-element Vector{Symbolics.Equation}:
 λ(t) ~ (2μ) / (P*sqrt((8M_air) / (R*T*π_val)))
 Kn(t) ~ (2λ(t)) / D_p
 C_c(t) ~ 1 + (A1 + A2*exp((-A3) / Kn(t)))*Kn(t)
 m_p(t) ~ (1//6)*(D_p^3)*π_val*ρ_p
 τ(t) ~ (C_c(t)*(D_p^2)*ρ_p) / (18μ)
 v_t(t) ~ g*τ(t)
 D_B(t) ~ (C_c(t)*T*k_B) / (3D_p*μ*π_val)
 B(t) ~ C_c(t) / (3D_p*μ*π_val)
 c_p(t) ~ sqrt((8T*k_B) / (m_p(t)*π_val))
 λ_p(t) ~ (2D_B(t)) / c_p(t)
 D_ca(t) ~ D_p*sqrt(ρ_p / ρ_p0)

Analysis

Slip Correction Factor vs Particle Size

The Cunningham slip correction factor (Eq. 9.34) accounts for non-continuum effects when particle size approaches the mean free path of gas molecules. For particles much larger than the mean free path, Cc approaches 1 (continuum regime). For smaller particles, Cc increases significantly.

using Plots, NonlinearSolve

sys = SingleParticleDynamics()
csys = mtkcompile(sys)

# Calculate Cc for different particle sizes
D_p_range = 10 .^ range(-9, -4, length = 50)  # 1 nm to 100 μm
C_c_values = Float64[]
Kn_values = Float64[]

for D_p in D_p_range
    prob = NonlinearProblem(csys, Dict(), Dict(csys.D_p => D_p))
    sol = solve(prob)
    push!(C_c_values, sol[csys.C_c])
    push!(Kn_values, sol[csys.Kn])
end

p1 = plot(D_p_range * 1e6, C_c_values,
    xscale = :log10, yscale = :log10,
    xlabel = "Particle Diameter (μm)",
    ylabel = "Slip Correction Factor Cc",
    title = "Cunningham Slip Correction (Eq. 9.34)",
    legend = false, linewidth = 2)
hline!([1.0], linestyle = :dash, color = :gray, label = "Continuum limit")
Example block output

Terminal Settling Velocity vs Particle Size

The terminal settling velocity (Eq. 9.42) shows the balance between gravitational force and drag. Larger particles settle faster due to their greater mass.

v_t_values = Float64[]

for D_p in D_p_range
    prob = NonlinearProblem(csys, Dict(), Dict(csys.D_p => D_p))
    sol = solve(prob)
    push!(v_t_values, sol[csys.v_t])
end

p2 = plot(D_p_range * 1e6, v_t_values * 100 * 3600,  # Convert to cm/h
    xscale = :log10, yscale = :log10,
    xlabel = "Particle Diameter (μm)",
    ylabel = "Terminal Velocity (cm/h)",
    title = "Settling Velocity (Eq. 9.42)",
    legend = false, linewidth = 2)
Example block output

Brownian Diffusion Coefficient vs Particle Size

The Stokes-Einstein-Sutherland relation (Eq. 9.73) shows that smaller particles have larger diffusion coefficients, allowing them to spread more rapidly through random Brownian motion.

D_B_values = Float64[]

for D_p in D_p_range
    prob = NonlinearProblem(csys, Dict(), Dict(csys.D_p => D_p))
    sol = solve(prob)
    push!(D_B_values, sol[csys.D_B])
end

p3 = plot(D_p_range * 1e6, D_B_values * 1e4,  # Convert to cm²/s
    xscale = :log10, yscale = :log10,
    xlabel = "Particle Diameter (μm)",
    ylabel = "Diffusion Coefficient (cm²/s)",
    title = "Brownian Diffusivity (Eq. 9.73)",
    legend = false, linewidth = 2)
Example block output

Comparison of Settling and Diffusion Timescales

The relative importance of gravitational settling versus Brownian diffusion depends on particle size. The characteristic timescale ratio τ_ds = 4D/vt² (Eq. 9.85) indicates which process dominates:

tau_ds_values = 4 .* D_B_values ./ (v_t_values .^ 2)

p4 = plot(D_p_range * 1e6, tau_ds_values,
    xscale = :log10, yscale = :log10,
    xlabel = "Particle Diameter (μm)",
    ylabel = "Characteristic Time τ_ds (s)",
    title = "Diffusion-Settling Timescale (Eq. 9.85)",
    legend = false, linewidth = 2)
hline!([1.0], linestyle = :dash, color = :gray)
Example block output

Summary Plot

plot(p1, p2, p3, p4, layout = (2, 2), size = (800, 600))
Example block output

Validation Against Table 9.3: Slip Correction Factors

The following compares computed slip correction factors against values from Table 9.3 in Seinfeld & Pandis (2006):

# Reference values from Table 9.3
table_data = [
    (0.001, 216.0),
    (0.01, 22.2),
    (0.1, 2.85),
    (1.0, 1.164),
    (10.0, 1.016),
    (100.0, 1.0016)
]

D_p_table = [d[1] * 1e-6 for d in table_data]  # Convert μm to m
Cc_table = [d[2] for d in table_data]

# Computed values
Cc_computed = Float64[]
for D_p in D_p_table
    prob = NonlinearProblem(csys, Dict(), Dict(csys.D_p => D_p))
    sol = solve(prob)
    push!(Cc_computed, sol[csys.C_c])
end

DataFrame(
    Symbol("Dp (μm)") => [d[1] for d in table_data],
    Symbol("Cc (Table 9.3)") => Cc_table,
    Symbol("Cc (Computed)") => round.(Cc_computed, digits = 2),
    Symbol("Error (%)") => round.(abs.(Cc_computed .- Cc_table) ./ Cc_table .* 100, digits = 1)
)
6×4 DataFrame
RowDp (μm)Cc (Table 9.3)Cc (Computed)Error (%)
Float64Float64Float64Float64
10.001216.0216.390.2
20.0122.222.160.2
30.12.852.860.4
41.01.1641.160.0
510.01.0161.020.0
6100.01.00161.00.0

Flow Regime Classification

The Knudsen number determines the applicable flow regime:

RegimeKn RangeDescription
ContinuumKn → 0Stokes' law applies (Cc ≈ 1)
TransitionKn ~ 1Slip correction required
Free molecularKn → ∞Kinetic theory applies
plot(D_p_range * 1e6, Kn_values,
    xscale = :log10, yscale = :log10,
    xlabel = "Particle Diameter (μm)",
    ylabel = "Knudsen Number",
    title = "Flow Regime Classification",
    legend = false, linewidth = 2)
hline!([0.1, 10], linestyle = :dash, color = :gray, label = "Regime boundaries")
annotate!([(0.001, 0.05, text("Continuum", 8)),
    (0.1, 5, text("Transition", 8)),
    (10, 500, text("Free Molecular", 8))])
Example block output