🔭 astrophysics

Investigating Nonlinear Quenching Effects on Polar Field Buildup in the Sun Using Physics-Informed Neural Networks

This study employs Physics-Informed Neural Networks to solve the surface flux transport equation, revealing how the nonlinear interplay between tilt and latitude quenching regulates solar polar field buildup and cycle amplitude while demonstrating the framework's superior accuracy and efficiency over traditional models.

Original authors: Jithu J. Athalathil, Mohammed H. Talafha, Bhargav Vaidya

Published 2026-03-17

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Jithu J. Athalathil, Mohammed H. Talafha, Bhargav Vaidya

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The Sun's Magnetic Heartbeat

Imagine the Sun as a giant, churning ball of hot gas with a massive magnetic heartbeat. This heartbeat follows a rhythm: every 11 years, the Sun goes through a "solar cycle" where it gets very active (lots of sunspots and storms) and then quiets down.

Scientists have long known that the strength of the next solar cycle depends on the Sun's polar magnetic fields (the magnetic fields at the North and South poles). Think of these polar fields as the "seed" or the "fuel tank" for the next cycle. If the fuel tank is full, the next cycle will be a monster. If it's empty, the next cycle will be weak.

But here's the mystery: Why does the fuel tank sometimes fill up a lot and other times only a little?

The Problem: The "Leaky" Tank and the "Smart" Sun

The Sun has a natural mechanism to move magnetic fuel from the equator to the poles. However, the Sun isn't a dumb machine; it has a way of regulating itself to prevent the cycle from getting too wild. It uses two "brakes" (nonlinear feedback mechanisms) to stop the polar fields from growing too huge:

The Tilt Brake (Tilt Quenching): Imagine sunspots as pairs of magnets (one North, one South) that pop up on the surface. Usually, they are tilted like a slanted fence. The stronger the cycle, the more the Sun "squints" these magnets, making them stand straighter. When they are straighter, the North and South poles cancel each other out more easily before they can reach the poles.
The Latitude Brake (Latitude Quenching): When the Sun is having a super-strong cycle, it tends to spawn these magnetic pairs further away from the equator (closer to the poles). Because they start so far north/south, they have a harder time crossing the equator to get to the opposite pole, which is necessary to build up the polar field.

The New Tool: The "Physics-Brain" (PINNs)

Traditionally, scientists used standard computer simulations (like a grid-based map) to model this. But the Sun is messy and non-linear, and these old maps sometimes get "jittery" or inaccurate when things get complicated.

In this paper, the authors used a new, fancy tool called Physics-Informed Neural Networks (PINNs).

The Analogy: Imagine you are trying to teach a robot to drive a car.
- Old Way: You show the robot a million photos of roads and tell it, "If you see a stop sign, stop." It learns by guessing patterns.
- PINN Way: You give the robot the photos, but you also program the laws of physics directly into its brain. You tell it, "You must obey gravity, friction, and momentum." Even if the robot has never seen a specific road before, it knows the physics of how a car moves.

The authors used PINNs to solve the equations of the Sun's magnetic field. Because the "laws of physics" are built into the AI's brain, it doesn't need as much data, and it doesn't make silly mistakes that grid-based models sometimes do.

What They Discovered

Using this "Physics-Brain," the authors ran thousands of simulations to see how the two brakes (Tilt and Latitude) interact. Here is what they found:

1. The "Traffic Jam" of Magnetic Fields
They discovered that the effectiveness of these brakes depends on how fast the Sun's surface flows (like a river) versus how much the magnetic fields spread out (like ink in water).

Fast River, Slow Spread (Advection): If the surface flow is fast, the Latitude Brake is the boss. It stops the fuel from reaching the poles effectively.
Slow River, Fast Spread (Diffusion): If the magnetic fields spread out easily, the Tilt Brake takes over. It makes the magnets cancel each other out before they can do any good.

2. The Perfect Balance (The Even-Odd Rule)
The most exciting finding is how these two brakes work together to create the Gnevyshev–Ohl rule. This is an observation that solar cycles often alternate between strong and weak (Strong, Weak, Strong, Weak).

The Analogy: Imagine a seesaw. If the Sun has a "Strong" cycle, the brakes kick in hard. The Tilt gets squashed, and the spots move too far north. This means the "fuel tank" for the next cycle doesn't fill up much. So, the next cycle is Weak.
Because the next cycle is weak, the brakes don't kick in as hard. The magnets stay tilted and near the equator, allowing the fuel tank to fill up completely. So, the cycle after that is Strong.
This creates a natural rhythm of "Strong-Weak-Strong-Weak."

3. No Need for a "Decay" Button
Older models often had to add a fake "decay" term (a button that just made the magnetic field disappear over time) to make the math work. The authors found that with PINNs, they didn't need this fake button. The training process of the AI naturally stabilized the system, just like a real physical system would.

Why This Matters

This paper is a big deal for two reasons:

Better Weather Forecasting: Just like we want to predict rain on Earth, we want to predict "Space Weather" (solar storms) that can knock out satellites and power grids. By understanding how the Sun regulates its own strength, we can predict if the next 11-year cycle will be a gentle breeze or a hurricane.
A New Way to Do Science: The authors proved that AI (specifically PINNs) is a superior tool for studying the Sun. It's more accurate, handles the messy "non-linear" parts better, and gives us a clearer picture of how the Sun's magnetic engine really works.

In short: The Sun is a self-regulating engine. It uses two different "brakes" to keep its cycles from going crazy. By using a smart AI that understands physics, we finally have a clear map of how these brakes work together to create the rhythmic heartbeat of our star.

1. Problem Statement

The solar dynamo, responsible for the Sun's 11-year magnetic cycle, relies on the regeneration of the poloidal magnetic field. This process is heavily modulated by nonlinear feedback mechanisms, specifically:

Tilt Quenching (TQ): Stronger solar cycles produce active regions (ARs) with smaller tilt angles relative to the equator (deviating from Joy's Law), leading to increased flux cancellation within hemispheres.
Latitude Quenching (LQ): Stronger cycles see ARs emerging at higher latitudes, reducing the efficiency of cross-equatorial flux transport.

These mechanisms regulate the buildup of the Sun's polar field, which acts as the seed for the next cycle. Traditional grid-based numerical models face limitations in capturing complex nonlinear interactions and often suffer from numerical instability or resolution sensitivity. Furthermore, the relative dominance of TQ versus LQ across different transport regimes (advection vs. diffusion) requires precise quantification to explain observed cycle-to-cycle variations, such as the Gnevyshev–Ohl rule (alternating strong and weak cycles).

2. Methodology

The authors employ Physics-Informed Neural Networks (PINNs) to solve the Surface Flux Transport (SFT) equation, offering a mesh-free alternative to traditional finite-difference schemes.

Governing Equation: The study solves the 1D SFT equation for the radial magnetic field $B(\lambda, t)$ :
$\frac{\partial B}{\partial t} = \frac{1}{R} \frac{\partial}{\partial \lambda} \left( u B \cos \lambda \right) + \frac{\eta}{R^2} \frac{\partial}{\partial \lambda} \left( \cos \lambda \frac{\partial}{\partial \lambda} \left( \frac{B}{\cos \lambda} \right) \right) - \frac{B}{\tau} + S(\lambda, t)$
Where $u$ is meridional flow, $\eta$ is diffusivity, $\tau$ is decay time, and $S$ is the source term representing bipolar magnetic regions (BMRs).
Nonlinear Source Modeling:
- TQ Implementation: The tilt angle $\delta$ is suppressed by a factor dependent on cycle amplitude ( $A_n$ ): $\delta = \delta_0 \sin(1 - b_{joy} \frac{A_n}{A_0})$ .
- LQ Implementation: The mean emergence latitude $\lambda_0$ shifts poleward based on cycle strength: $\lambda_0 = \lambda_{base} + b_{lat} (\frac{A_n}{A_0})$ .
- Stochasticity: Cycle amplitudes follow a lognormal distribution to simulate realistic fluctuations.
PINN Architecture:
- Inputs: Latitude ( $\lambda$ ) and Time ( $t$ ).
- Output: Magnetic field $B(\lambda, t)$ .
- Loss Function: A weighted sum of Initial Condition (IC), Boundary Condition (BC), and PDE Residual losses. The PDE residual is computed using automatic differentiation to enforce the SFT equation directly within the network training.
- Optimization: A two-step strategy using ADAM (for rapid convergence) followed by L-BFGS (for high-precision fine-tuning).
Diagnostic Metric: The Residual Dipole Moment ( $D_{res}$ ) is used to isolate cycle-to-cycle amplification by removing the exponential memory of previous cycles. The study analyzes the deviation in dipole gain ( $\Delta D$ ) caused by TQ and LQ relative to a linear (no-quenching) baseline.

3. Key Contributions

PINN Framework for Solar Dynamo: Demonstrates the application of PINNs to the SFT problem, achieving mesh-independent solutions with superior numerical stability and flux conservation compared to traditional upwind schemes.
Quantification of Nonlinear Interplay: Systematically isolates the relative contributions of TQ and LQ across a wide parameter space (meridional flow $u_0$ and diffusivity $\eta$ ).
Refined Empirical Scaling: Establishes a smooth inverse-square dependence for the ratio of LQ to TQ effects ( $\Delta D_{LQ} / \Delta D_{TQ}$ ) as a function of the Dynamo Effectivity Range ( $\lambda_R = u_0 R / \eta$ ). This refines previous empirical fits with reduced scatter.
Decay Term Analysis: Reveals that the explicit decay term ( $\tau$ ) is not essential for stabilizing the PINN solution, as the training process inherently introduces a regularization effect that mimics decay, making the system more dependent on transport parameters ( $\lambda_R$ ) than on $\tau$ .

4. Key Results

Regime Dependence:
- Advection-Dominated Regimes (Low $\lambda_R$ , High $u_0$ , Low $\eta$ ): LQ is the dominant quenching mechanism.
- Diffusion-Dominated Regimes (High $\lambda_R$ , Low $u_0$ , High $\eta$ ): TQ becomes increasingly effective and eventually dominates.
- Transition Point: The crossover where TQ and LQ contribute equally ( $\Delta D_{LQ} / \Delta D_{TQ} \approx 1$ ) occurs at $\lambda_R \approx 10^\circ$ .
Combined Effects: The combined LQ+TQ model produces the strongest suppression of polar dipole buildup, particularly during strong cycles, leading to lower dipole amplitudes than either mechanism alone.
Accuracy Improvements:
- The PINN model achieves a coefficient of determination ( $R^2$ ) of 0.9841, significantly outperforming the traditional scheme ( $R^2 \approx 0.973$ ).
- Error metrics (MAE, MSE, RMSE, MAPE) are substantially lower for PINNs, indicating better recovery of nonlinear trends and reduced numerical scatter.
Cycle Modulation Mechanism: The study provides a physical explanation for the Gnevyshev–Ohl rule (even-odd cycle alternation):
- Strong Cycles: TQ reduces tilt (increasing intra-hemisphere cancellation) and LQ shifts emergence poleward (reducing cross-equatorial transport). This suppresses polar field buildup, leading to a weak subsequent cycle.
- Weak Cycles: TQ is weak (larger tilts) and LQ shifts emergence equatorward (enhancing cross-equatorial cancellation). This allows more flux to reach the poles, building a stronger polar field and seeding a strong subsequent cycle.

5. Significance

Predictive Capability: The PINN framework offers a robust, accurate, and physically consistent tool for solar cycle prediction. It can be extended to incorporate real-time data assimilation (e.g., Wilcox Solar Observatory maps) for forecasting.
Physical Insight: The study clarifies that the interplay between TQ and LQ, governed by the transport regime ( $\lambda_R$ ), is the primary driver of solar cycle amplitude modulation, rather than simple decay terms.
Methodological Advancement: By embedding physical laws directly into the neural network, this work demonstrates that machine learning can solve complex astrophysical PDEs with higher fidelity than traditional numerical methods, particularly in capturing nonlinear feedback loops essential for dynamo stability.
Future Applications: The framework paves the way for data-constrained solar cycle forecasts using high-resolution observations from instruments like DKIST, bridging the gap between theoretical dynamo models and observational constraints.