Data and Code to reproduce the results in the manuscript:
Reconstructions of Jupiter's magnetic field using physics-informed neural networks (PINN)
Note
This is the Vector Potential Version
- Juno vector magnetic data
- Orbit 1-33 (Prime mission)
- Orbit 1-50 (Prime + Extended mission)
- Sampling rate: 30 sec
- Within 4.00
$R_J$ , orbit 2 not used - Format: PJ, Year, Decimal-Day, x, y, z, Bx, By, Bz
-
Spherical Harmonic Models
of Juno magnetic data-
JRM33_I30
: degree 30 model of (Connerney et al., 2022) -
Bloxham_I32
: degree 32 model of (Bloxham et al., 2022) - Format: IND, SHC, g | h, n, m
-
- Collocation cloud
- Random points within the region
$[1.00 R_J, 4.00 R_J]$ - Random points within the region
$[0.80 R_J, 1.00 R_J]$ - Random points on the surface
$r = 1.00 R_J$ - Format: x, y, z
Generated based on the fact: Normalized vector of Gaussian variables is uniformly distributed on the sphere
- Random points within the region
- Training diagnostics from training model
PINN50e
: Jupyter Notebook - Plot comparison of RMS errors of
PINN Models
andSH models
on each orbit of Juno Observation Orbits: Jupyter Notebook - Plot showing
PINN models
predicted physical misfit (current density$|\mathbf{J}|$ ) on Multiple$R_J$ : Jupyter Notebook - Plot comparison of
$B_{r}$ predicted byPINN models
andSH models
on Multiple$R_J$ : Jupyter Notebook - Plot comparison of Lowes spectrums of
PINN models
andSH models
: Jupyter Notebook - Table showing the RMS errors of
Spherical Harmonic Models
with increasing degree$n$ at different subset of Juno Observation Orbits: Jupyter Notebook - Table showing the RMS errors of
PINN Models
at different subset of Juno Observation Orbits: Jupyter Notebook - PINN training Above 1.00
$R_J$ (NN Jupiter Magnetic Model): Python Code - PINN training Below 1.00
$R_J$ (Downward Continuation): Python Code - Predict magnetic vector fields at Juno Observation Locations using
PINN models
and write to ascii file: Python Code - Predict magnetic vector fields at Multiple
$R_J$ usingPINN models
and write to ascii file:-
Above 1.00
$R_J$ : Python Code -
Below 1.00
$R_J$ : Python Code
-
Above 1.00
- Predict electric currents at Multiple
$R_J$ usingPINN models
and write to ascii file:-
Above 1.00
$R_J$ : Python Code -
Below 1.00
$R_J$ : Python Code
-
Above 1.00
- Predict magnetic vector fields at Interface
$r=1.00 R_J$ usingPINN models
and write to ascii file:-
Above 1.00
$R_J$ : Python Code -
Below 1.00
$R_J$ : Python Code
-
Above 1.00
- Predict magnetic vector fields at Juno Observation Locations using
Spherical Harmonic Models
and write to ascii file: Python Code - Predict magnetic vector fields at Multiple
$R_J$ usingSpherical Harmonic Models
and write to ascii file: Python Code
- Models (Last 10 models of each training)
-
Above 1.00
$R_J$ :-
PINN33e
: model trained using Orbit 1-33 -
PINN50e
: model trained using Orbit 1-50
-
-
Below 1.00
$R_J$ :-
PINN33i
: downwards extrapolation model ofPINN33e
in$[0.80 R_J, 1.00 R_J]$ -
PINN50i
: downwards extrapolation model ofPINN50e
in$[0.80 R_J, 1.00 R_J]$
-
-
Above 1.00
-
PINN models
andSpherical Harmonic Models
predicted magnetic vector fields at Juno Observation Locations:- Format: PJ, Year, Decimal-Day, x, y, z, Bx, By, Bz
-
PINN models
andSpherical Harmonic Models
predicted gridded Jupiter magnetic vector fields at Multiple$R_J$ :- Format: Lon, Lat,
$B_{\theta}$ ,$B_{\lambda}$ ,$B_{r}$ ,$|\mathbf{B}|$ ;
- Format: Lon, Lat,
-
PINN models
predicted gridded data of electric currents ($\mathbf{J}$ ) at Multiple$R_J$ :- Format: Lon, Lat,
$J_{\theta}$ ,$J_{\lambda}$ ,$J_{r}$ ,$|\mathbf{J}|$ ;
- Format: Lon, Lat,
-
PINN models
training process data:- Data and physics Loss terms, Dynamic weights, Learning Rate;
- Connerney, J. E. P., Timmins, S., Oliversen, R. J., Espley, J. R., Joergensen, J. L., Kotsiaros, S., et al. (2022). A new model of Jupiter's magnetic field at the completion of Juno's Prime Mission. Journal of Geophysical Research: Planets, 127, e2021JE007055. https://doi.org/10.1029/2021JE007055
- Bloxham, J., Moore, K. M., Kulowski, L., Cao, H., Yadav, R. K., Stevenson, D.J., et al. (2022). Differential rotation in Jupiter's interior revealed by simultaneous inversion for the magnetic field and zonal flux velocity. Journal of Geophysical Research: Planets, 127, e2021JE007138. https://doi.org/10.1029/2021JE007138
- Edwards, T. M., Bunce, E. J., & Cowley, S. W. H. (2001). A note on the vector potential of Connerney et al.'s model of the equatorial current sheet in Jupiter's magnetosphere. Planetary and Space Science, 49, 1115-1123. https://doi.org/10.1016/S0032-0633(00)00164-1
- Wieczorek, M. A., & Meschede, M. (2018). SHTools: Tools for working with spherical harmonics. Geochemistry, Geophysics, Geosystems, 19, 2574–2592. https://doi.org/10.1029/2018GC007529