ikuna0@engr.uky.edu
Department of Mechanical Engineering
University of Kentucky, Lexington, KY 40506
Advanced Computational Fluid Dynamics
C F D
• Overview of different models
• Classical heat conduction model
• Hyperbolic heat conduction model
• Two-step models (parabolic and hyperbolic two-step models)
• Dual phase lag heat conduction model
• Mathematical formulation
• Numerical analysis
• Stability criterion
• Numerical results
• 1D problem (short-pulse laser heating of gold film)
• 3D problem (short-pulse laser heating of gold film at different locations)
• Grid function convergence tests
Advanced Computational Fluid Dynamics
C F D
• Heat flux directly proportional to temperature gradient (Fourier's law)
q ( r
, t )
k
T ( r
, t )
• Incorporation into first law of thermodynamics yields parabolic heat conduction equation
T
t
2
T
Anomalies associated with Fourier law
• Heat conduction diffusion phenomenon in which temperature disturbances propagate at infinite velocities. Assumes instantaneous thermodynamic equilibrium
• Heat flow starts (vanishes) simultaneous with appearance (disappearance) of temperature gradient,violating causality principle
Fourier's law fails to predict correct temperature distribution
• Transient heat flow for extremely short periods of time (applications involving laser pulses of nanosecond and femtosecond duration)
• High heat fluxes
• Temperatures near absolute zero (heat conduction at cryogenic temperatures)
Advanced Computational Fluid Dynamics
C F D
• Modified heat flux that accommodates finite propagation speed of observed thermal waves proposed by Vernotte and Cattaneo (1958)
q ( r
, t
q
)
k
T ( r
, t )
q (
r , t )
t
q ( r
, t )
k
T ( r
, t )
• Combined with equation of energy conservation gives hyperbolic heat conduction equation (HHCE)
t
2
T
2
1 q
T t
c
2 2
T c
1
2
q c is the speed of thermal wave propagation
• HHCE suffers from theoretical problem of compatibility with second law of thermodynamics
• predicts physically impossible solutions with negative local heat content
• Neglects energy exchange between electrons and the lattice, applicability to short pulselasers becomes questionable
• No clear experimental evidence of hyperbolic heat conduction even though wave behavior has been observed
• Earliest experiments detecting thermal waves performed by Peshkov (1944) using superfluid liquid helium at temperature near absolute zero
• He referred to this phenomenon as “second sound” , because of similarity between observed thermal and ordinary acoustic waves
Advanced Computational Fluid Dynamics
C F D
Stage I Stage II
Heated electrons
Energy quanta;
Phonons no temperature rise at time t temperature rise at time t + τ
Photon energy at time t
Electron-gas heating by photons Metal lattice heating by phonon-electron interactions
* D. Y. Tzou, Macro-microscale Heat Transfer, the lagging behavior
Advanced Computational Fluid Dynamics
C F D
Anisimov et al. (1974) proposed two-step model to describe the electron temperature and lattice temperature during the short-pulse laser heating of metals
C e
C l
T e
t
T l
t
G
( T e
q
G(T e
T l
)
T l
) (Heating of electrons)
(Heating of metal-lattice) eliminating electron-gas temperature ( T e
)
G
4
( n e
V s k
B
)
2 k
1
C
E
2
2
T l
t
2
1 e
T l
t
C
2
E e
(
t
2
T l
)
2
T l eliminating metal-lattice temperature ( T l
)
1
C
E
2
Where,
2
T e
t
2
1
e
T e
t
e
C
E
2
(
t
2
T e
)
2
T e
G = Phonon-electron coupling factor
V s
= speed of sound n e
= number density of free electrons per unit volume k
B
= Boltzmann constant k = Thermal conductivity
C e
= Heat capacity of electrons
C l
= Heat capacity of metal lattice
e
C e k
C l
C
E
kG
C e
C l
Advanced Computational Fluid Dynamics
C F D
•
Modified heat flux vector represented by Tzou (1995)
T
q ( r
, t
q
)
k
T ( r
, t
)
T
~ delay behavior in establishing the temperature gradient
•
q
~ delay behavior in heat-flow departure
Taylor expansion gives
q ( r
, t )
q
q ( r
,
t t )
k
T ( r
, t )
T
[
T ( r
, t )]
t
• coupled with energy equation gives dual phase lag (DPL) equation
q
t
2
T
2
1
T
t
T
(
t
2
T
(
C
C e l
C e
)
•
Comparing coefficients of DPL model with those of two-step model we can represent microscopic properties by
C e k
C l
T
C l
G
q
G C l
)
2
T
G
4
( n e
V s k k
B
)
2 n e
T
G
0
q
2
T
t
2
q
1
T
T
t
0
T
q
2
T
t
2
1
T t
(
2
T )
t
2
T
T
(
2
T )
t
2
T
classical diffusion equation classical wave equation
Advanced Computational Fluid Dynamics
C F
D
Volumetric heating in the sample
S ( z , t )
S
0 e
z
I ( t ) where,
S
0
0 .
94 J
t p
R
I ( t )
I
0 e
a t t p
Gold Film
Laser
Femtosecond laser heating is modeled by energy absorption rate
S ( z , t )
0 .
94 J
t p
R
exp
z
1 .
992 t t p
2 t p
(1D)
S ( r
, t )
0 .
94 J
t p
R
exp
x
2 r
0
2 y
2
z
1 .
992 t t p
2 t p
(3D)
S (
r t
q
2
T
t
2
1
T
t
T
(
2
T )
t
2
T
1 k
S
q
S
t
L = 100 nm
• Laser fluence J = 13.4 Jm 2
• Reflectivity R=0.93
• Thermalization time t p
=96 fs
• Depth of laser penetration = 15.3nm
• Radius of laser beam r
0
= 100nm
• S
0 is the intensity of laser absorption
•
•
•
•
• I(t) is the intensity of laser pulse
= 1.2
10 4 m 2 s 1
= 8.5
ps
q
T
= 90 ps k = 315 Wm 1 K 1 z
Advanced Computational Fluid Dynamics
C F D
q ( r
, t )
q
q ( r
, t )
t
k
T ( r
, t )
T
[
T ( r
, t )]
t q
1
q
q
t
1 k
T
x
T
t
T
x q
2
q
q
t
2 k
T
y
T
t
T
y
q
3
q
q
t
3 k
T
z
T
t
q
S
C
T
t
500nm
( r
, t )
T
z
q
2
T
t
2
1
T
t
T
(
2
T )
t
2
T
1 k
S
q
S
t x z
100nm y
500nm
500nm
500nm
Initial Conditions
T ( r
, 0 )
T o
T t
( r , 0 )
0
Boundary Condition
T
r
0 , r
x , y , z
Advanced Computational Fluid Dynamics
C F D
Explicit finite-difference scheme employed to solve the DPL equation
Centered differencing approximates second-order derivatives in space
Centered differencing is employed for time derivative in the source term
2
T
x
2
1
x
2
[ T i n
1 , j , l
2 T i , n j , l
T i n
1 , j , l
]
S
t
1
2
t
[ S i n
,
1 j
S i n
,
1 j
]
Mixed derivative is approximated using centered difference in space and backward difference in time
Stability criterion for 3-D DPL model obtained using von Neumann eigenmode analysis (Tzou)
3
T
t
x
2
1
[ T
t
x
2
T i n
1
1 , j , l i n
1 ,
j , l
2 T i , n j
1
, l
2 T i , n j , l
T
i n
1
1 , j
T
, l i n
1 ,
] j , l
t ( 2
t
x
2
(
t
4
T
2
q
)
)
t ( 2
t
y
2
(
t
4
T
2
q
)
)
t ( 2
t
z
2
(
t
4
T
2
q
)
)
1
Forward differencing approximates first-order derivative in time
T
t
1
t
[ T i , n j
1
, l
T i , n j , l
]
t
b
a b
2
4 ac
2 a
2
(
y
2 z
2 x
2 z
2 x
2 y
2
)
Centered differencing approximates secondorder derivative in time b
x
2 y
2 z
2
4
y
2 z
2
T
4
x
2 z
2
T
4
x
2 y
2
T
2
T
t
2
1
(
t )
2
[ T i , n
1 j , l
2 T i , n j , l
T i , n j
1
, l
] c
2
x
2 y
2 z
2 q
Advanced Computational Fluid Dynamics
C F D
Laser
Gold Film
X
L = 100 nm
Fig. 1. Normalized Temperature change at front surface of a gold film of thickness 100nm: = 1.2
10 4 m 2 s 1 , k = 315 Wm 1 K 1 ,
T
= 90 ps, q
= 8.5
ps .
Advanced Computational Fluid Dynamics
C F D
200nm laser beam
Work piece-Gold
500nm
500nm
250nm
250nm
500nm
500nm
Fig. 2. 3-D schematic of laser heating of gold film at different locations
Advanced Computational Fluid Dynamics
C F D
DPL
DPL DPL
Parabolic
Parabolic
Hyperbolic
DPL
DPL
Parabolic
Parabolic Hyperbolic
Fig. 3. Temperature distribution at top surface of gold film predicted by different models
Advanced Computational Fluid Dynamics
C F D
At t = 0.3
ps
At t = 0.9
ps
DPL
DPL DPL
Parabolic Hyperbolic
DPL
DPL
Parabolic
Parabolic
Hyperbolic
Hyperbolic
Fig. 3a. Temperature distribution at top surface of gold film predicted by different models
Advanced Computational Fluid Dynamics
C F D
At t = 1.56
ps
At t = 2.23
ps
DPL
DPL
Parabolic
Parabolic
Hyperbolic
DPL Parabolic
Parabolic
Hyperbolic
Fig. 3b. Temperature distribution at top surface of gold film predicted by different models
Advanced Computational Fluid Dynamics
C F D
312
310
308
306
304
302
300
51×51×11
101×101×21
201×201×41
0 50 100 150
Radial distance (nm )
200
Fig. 4. Temperature plots of front surface of gold film at t = 0.3
ps in radial direction using different grids: 51 51 11, 101 101 21 and 201 201 41
250
Advanced Computational Fluid Dynamics
C F D
• DPL model agrees closely with experimental results in one dimension compared to the classical and the hyperbolic models
• Energy absorption rate used to model femtosecond laser heating modified to accommodate for three-dimensional laser heating
• Simulation of 3-D laser heating at various locations of thin film carried out using pulsating laser beam (~ 0.3 ps pulse duration) to compare different models
• Stability criterion for selecting a numerical time step obtained
• using von Neumann eigenmode analysis: x = y = z = 5nm
t = 3.27 fs
• Different grids (51 51 11, 101 101 21 and 201 201 41) were used to check convergence in numerical solution
• Classical and hyperbolic models over predict temperature distribution during ultra-fast laser heating, whereas DPL model gives temperature distribution comparable to experimental data
Advanced Computational Fluid Dynamics
C F D
• Compared to experimental data large difference in diffusion model due to negligence of both micro structural interaction in space and fast transient effect in time .
• Hyperbolic model redeems difference between experimental and diffusion, but still overestimates transient temperature because it neglects micro structural interaction in space.
• DPL model incorporates delay time caused by phonon-electron interaction in micro scale
– Time delay due to fast transient effect of thermal inertia absorbed in phase lag of heat flux
– Time delay due to finite time required for phonon-electron interaction to take place absorbed in phase lag of temperature gradient transient temperature closer to experimental observation.
Advanced Computational Fluid Dynamics
C F D