JunoMag_PINN_VP3

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

---

Input Data

• 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

---

Python Code and Jupyter Notebook

• Training diagnostics from training model PINN50e: Jupyter Notebook
• Plot comparison of RMS errors of PINN Models and SH 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 by PINN models and SH models on Multiple $R_J$: Jupyter Notebook
• Plot comparison of Lowes spectrums of PINN models and SH 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 $R_J$ using PINN models and write to ascii file:
  • Above 1.00 $R_J$: Python Code
  • Below 1.00 $R_J$: Python Code
• Predict electric currents at Multiple $R_J$ using PINN models and write to ascii file:
  • Above 1.00 $R_J$: Python Code
  • Below 1.00 $R_J$: Python Code
• Predict magnetic vector fields at Interface $r=1.00 R_J$ using PINN models and write to ascii file:
  • Above 1.00 $R_J$: Python Code
  • Below 1.00 $R_J$: Python Code
• 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 $R_J$ using Spherical Harmonic Models and write to ascii file: Python Code

---

Output NN Models

• 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 of PINN33e in $[0.80 R_J, 1.00 R_J]$
    • PINN50i: downwards extrapolation model of PINN50e in $[0.80 R_J, 1.00 R_J]$

---

Output data

• PINN models and Spherical Harmonic Models predicted magnetic vector fields at Juno Observation Locations:
  • Format: PJ, Year, Decimal-Day, x, y, z, Bx, By, Bz
• PINN models and Spherical Harmonic Models predicted gridded Jupiter magnetic vector fields at Multiple $R_J$:
  • Format: Lon, Lat, $B_{\theta}$, $B_{\lambda}$, $B_{r}$, $|\mathbf{B}|$;
• PINN models predicted gridded data of electric currents ($\mathbf{J}$) at Multiple $R_J$:
  • Format: Lon, Lat, $J_{\theta}$, $J_{\lambda}$, $J_{r}$, $|\mathbf{J}|$;
• PINN models training process data:
  • Data and physics Loss terms, Dynamic weights, Learning Rate;

---

References

• 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