granular thermoelectrics · spillmann, teschner, “cosserat nets”, 2009 • twist buckling...
TRANSCRIPT
Andreas Glatz1, Igor Aronson1, George Crabtree1, Hanqi Guo2, Gregory Kimmel1, Alexei Koshelev1, Todd Munson2, Carolyn Phillips2, Ivan Sadovsky1, Jason Sarich2
1Materials Science Division, Argonne National Laboratory 2Mathematics and Computer Science Division, Argonne National Laboratory
http://www.oscon-scidac.org/
http://oscon-scidac.org
Outline
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
Critical Current by Design
Experimental validation
Non-additivity of defects
Prediction of critical currents
Intelligent optimization of Superconductors
Approach
Examples
Extracting, Tracking, and Visualizing of Vortices
Tracking methods
Tools
Next steps
2
SDAV
FASTMath & SUPER
see Poster A1 (18)
http://oscon-scidac.org
Model and Pinning
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
Time dependent GL equations:
In dimensionless units:
3
Inclusions and defects are modeled by critical temperature Tc(r) [ε(r)=Tc(r)/T-1] modulation
http://oscon-scidac.org
Examples: Order parameter ψ(r) follows Tc-
pattern
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
4
Ch
ecke
rbo
ard
Po
lycry
sta
llin
e
Re
cta
ng
ula
r
|ψ|2
Tc
http://oscon-scidac.org
Experimental motivation
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
5
Angular dependence of the critical current in commercial REBCO coated conductors
TEM image of heavy-
ion damage tracks
Irradiation
YBCO tape made by
http://oscon-scidac.org
Observation and scientific questions
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
6
• Disappearance of central peak
after irradiation Pinning
effects are non-additive.
Explanation?
• Overall increase of the critical
current. Can we increase it
more?
• Can we get a more
homogeneous critical current?
• What is the optimal irradiation
track concentration?
Here we use large-scale Ginzburg-Landau simulations to address these
questions
Simulation-assisted design of mixed-pinning landscapes with
predictable “critical-current-by-design”
http://oscon-scidac.org
Sample realization
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
7
Sample is realized as a cuboid, discretized using a regular mesh of 108 grid points with mesh size of x0/2
(quasi-)periodic boundary conditions Inclusions and irradiation tracks are modeled by a
different low-Tc component Anisotropy in c-direction is implemented by an
anisotropy factor g=5 For each field and pinning configuration an IV curve
is calculated from which the critical current is obtained
http://oscon-scidac.org
Comparison
Experiment &
Simulation
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
8
http://oscon-scidac.org
Validation of the simulation
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
9
http://oscon-scidac.org
Depinning in fixed field B (θ = 0°/45°)=0.08
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
10
http://oscon-scidac.org
Discovery of novel mechanism: Non-additivity of
defects
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
11
B (θ = 0°) FL
Columnar defects work like shortcuts for vortices
http://oscon-scidac.org
Predication of the critical current in irradiated samples
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
12
BΦ
[Hc2]
BΦ
[Hc2]
Jc
[Jd
p]
θ [°] Angle of field, θ [°]
Sa
me
cu
rre
nt, J
> J
c
nanorods + columns B || c
Sadovskyy et. al., submitted to Nature Materials (2015)
http://oscon-scidac.org
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
13
Critical Current by Design
Experimental validation
Non-additivity of defects
Prediction of critical currents
Intelligent Optimization of Superconductors
Approach
Example
Extracting, Tracking, and Visualizing of Vortices
Tracking methods
Tools
Next steps
SDAV
FASTMath & SUPER
see Poster A2 (19)
http://oscon-scidac.org
Optimization Challenge & Approach
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
14
Desired output
Optimal inclusion configurations for maximal critical current
Robustness of these configurations
Dependence on magnetic field and temperature
Challenge
We need to calculate the critical current for many configurations with small fluctuations (requires disorder averaging)
A typical pinning landscape has about 10 free parameters
Approach
Fully automated derivative-free optimizer: Define possible parameter ranges and the optimizer samples promising pinning configurations
Can handle arbitrary combination of different defect types
http://oscon-scidac.org
Example: 2 parameter optimization of random
spherical inclusions
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
15
System size: 100x100x50 x3 [256x256x128 mesh points] Simulation time=105 per current
Occupied volume fraction f npVp
Overlap: f f - f2/2
with diameter a and density np
http://oscon-scidac.org
Optimal particle density at fixed magnetic field
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
16
IVs for different numbers of particles (occupied volume fractions)
0.000 0.005 0.010 0.015 0.020 0.025 0.0300
5
10
15
20 300 (0.02) 4800 (0.27)
500 (0.033) 5600 (0.305)
800 (0.052) 6400 (0.337)
1200 (0.077)
1600 (0.1) flux flow
2400 (0.148) 0.02 ff J
3200 (0.191)
4000 (0.232)
a=4x, g=5,
Bz=0.1H
c2
E [
x1
04]
J
0.0 0.1 0.2 0.30.00
0.01
0.02
0.03
0.04
0.053 Jdp
0.076 Jdp
B=0.016
B=0.05
B=0.1 a=4
J c
Volume fraction
0.1 Jdp
http://oscon-scidac.org
Optimal parameters
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
17
Critical current vs particle density for different particle sizes
Dependences of optimal parameters on magnetic field
0.0 0.1 0.2 0.30.000
0.005
0.010
0.015
0.020
Jc
a = 3
a = 2
a = 6
B = 0.1
f
a = 4
0.00 0.05 0.10 0.15 0.200.00
0.01
0.02
0.03
0.04
B
J c,m
ax
0.00
0.02
0.04
0.06
0.08
0.10
J c,m
ax/J
dp
0
1
2
3
4
5
aopt/x
0.0
0.1
0.2
0.3
0.4
0.5
fopt
(b)
(a)
Contour plots of critical current
• Optimal particle diameter
decreases with field
• Optimal volume fraction = 18-
20% 0.10 0.15 0.20 0.25 0.302.0
2.2
2.4
2.6
2.8
3.0
3.2
3.4
3.6
3.8
4.0
4.2
4.4
f
a
0.0136
0.0140
0.0144
0.0148
0.0152
0.0156
0.0160
0.0164
0.0168
0.0172
0.0176
0.0180
0.0184
0.0188
0.0192
0.0196
0.0200
0.0204
0.0206
0.0208
B=0.1
0.10 0.15 0.20 0.253.0
3.2
3.4
3.6
3.8
4.0
4.2
4.4
4.6
4.8
5.0
f
a
0.0250
0.0252
0.0254
0.0256
0.0258
0.0260
0.0262
0.0264
0.0266
0.0268
0.0270
0.0272
0.0274
0.0276
0.0278
0.0280
0.0282
0.0284
0.0285
0.0286
B=0.05
http://oscon-scidac.org
Sampling in parameter space
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
18
http://oscon-scidac.org
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
19
Critical Current by Design
Experimental validation
Non-additivity of defects
Prediction of critical currents
Intelligent Optimization of Superconductors
Approach
Examples
Extracting, Tracking, and Visualizing of Vortices
Extraction & tracking methods
Tools
Next steps
SDAV
FASTMath & SUPER
see Poster B1 (19)
http://oscon-scidac.org
Vortices in superconductors
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
20
Vortices are defined as topological defects of the order parameter field ψ, which are a locus of points satisfy n is usually +/-1, indicating the chirality
Vortices are 1D curves in 3D space
The vortices are fundamentally different from vortices in fluid flow
Amplitude |ψ|
Phase θ
http://oscon-scidac.org
TDGL Data Visualization
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
21
Little research has been done in visualizing complex-field data
– Volume rendering/isosurface blur fine features
– Vortex extraction for single frames in regular grid data is proposed by Phillips et al. recently
Extracting, tracking, and visualizing vortices are the keys to understand the dissipative material behaviors and the impact of adding material inclusions.
http://oscon-scidac.org
Deliverables of a detection algorithm
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
22
A vortex extraction algorithm for both structured and unstructured mesh TDGL data
A vortex tracking algorithm, which is as accurate as the data discretization
Application of various visualization techniques
http://oscon-scidac.org
Overview of the Vortex Extraction Algorithm
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
23
• Vortex extraction locates vortex line at single time frames
By definition, singularities can be localized by checking phase jumps over mesh faces.
As there are always equal numbers of “ins” and “outs” for each cell, the punctured faces are further connected into vortex lines based on the mesh connectivity.
http://oscon-scidac.org
Overview of the Vortex Tracking Algorithm
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
24
• Vortex tracking algorithm relates vortex lines over adjacent frames, unless there are topological changes.
The movement of a vortex line is detected by checking each space-time edge to see whether it is intersected at an intermediate time between two adjacent frames.
http://oscon-scidac.org
Algorithm Pipeline
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
25
Load data
Extract punctured spatial faces
Extract intersected space-time edges
Graph-based vortex extraction and tracking
http://oscon-scidac.org
Algorithm Details –
Punctured Face/Intersected Edge Detection
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
26
Phase jump over faces
Phase jump over space-time edges
• Gauge transformation
*) Always use local transformations
to compute phase jumps
http://oscon-scidac.org
Event Detection and Visualization
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
27
Events are defined as topological changes of vortex graphs: merging, splitting, birth, death, recombination/crossing, etc.
Event visualization is based on storyline-like visual representations [Tanahashi and Ma 2012]
http://oscon-scidac.org
Results – 3D Structured Grid Data
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
28
http://oscon-scidac.org
Results – 3D Unstructured Data
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
29 Guo et. al., Proc. IEEE SciVis (2015)
http://oscon-scidac.org
Software Development and ParaView Plugins
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
30
A standalone visualization tool, as well as a ParaView plugin are developed for loading, analyzing, and visualizing TDGL simulation data.
The unstructured mesh data structures are based on libMesh, which is the finite element library used by the simulation. The framework can be integrated with the simulation for in-situ analysis in the future.
http://oscon-scidac.org
Next steps: Twist, Writhe, and Stabilizing Helical
Vortices
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
31
When electrostatic and magnetic fields aligned, phase field twists around the vortex.
Shielding supercurrents spirals around vortex rather than form planar loops.
Twisting phase field can be numerically extracted from GL simulations. Represented as ribbon.
No Twist
Twisting phase field Twisting phase field
http://oscon-scidac.org
Excess Twist leads to Writhe
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
32
Spillmann, Teschner, “Cosserat Nets”, 2009
• Twist Buckling Instability Writhe of Filament/Rod/Vortex
• Perpendicular Lorentz Forces on writhing vortex helical coils unstable. “Blows up” (Superconducting state lost): New criterion for upper limit of the critical current
• However, if vortex inside a cylindrical inclusion, cylinder can “traps” helical vortex state. (Superconducting state retained)
Twisted Rod Buckles and Writhes
Stable Helical Twisted Vortex
in Cylindrical Inclusion
http://oscon-scidac.org
Acknowledgements
Optimizing SuperConductor Transport Properties Through Large-Scale Simulation
Igor Aronson Alexei
Koshelev
Dmitry
Karpeev
Carolyn
Phillips
Ivan
Sadovskyy
33
George
Crabtree
Todd
Munson
Gregory
Kimmel
The Team
Jason
Sarich Stefan Wild Hanqi Guo