High-order exponential solver method for particle-in-cell simulations in cylindrical geometry

This paper introduces a high-order real-space finite-difference exponential time-domain solver for cylindrical particle-in-cell simulations that achieves accuracy comparable to spectral methods like FBPIC while avoiding basis function transformations, as validated through benchmarks and laser wakefield acceleration simulations.

The Gist

Imagine you are trying to simulate a high-speed race between a laser beam and a swarm of tiny particles (electrons) inside a long, cylindrical tube. This is what happens in advanced laser physics, specifically in a process called Laser Wakefield Acceleration (LWFA), where lasers push particles to incredible speeds over very short distances.

To understand this race, scientists use computer simulations called Particle-in-Cell (PIC). Think of these simulations as a massive, digital movie where the computer tracks every single particle and the electromagnetic fields around them.

The Problem: The "3D" Bottleneck

Usually, to get a perfect picture of this race, you need to simulate it in full 3D (like a real movie). However, because the laser and the plasma tube are perfectly round (cylindrical), simulating the whole 3D space is like trying to paint a picture of a round pipe by painting every single square inch of a giant cube around it. It's incredibly slow and requires supercomputers that are hard to find.

Scientists have tried to simplify this by using "cylindrical" math, which is like looking at the pipe from the side and only simulating a slice. The best existing method (used by a famous code called FBPIC) does this by translating the whole problem into a special "Fourier-Bessel" language. It's like translating a book into a secret code to make it easier to read, but then you have to translate it back to understand the results. This translation process is computationally expensive and can sometimes introduce small errors.

The Solution: A New "Real-Space" Solver

The authors of this paper, Szilárd Majorosi and colleagues, have built a new tool that solves the same problem but stays in "real space."

The Analogy:
Imagine you are trying to measure the ripples in a pond.

• The Old Way (FBPIC): You take a photo of the ripples, translate the photo into a complex mathematical code (Fourier-Bessel), solve the math, and then translate the photo back to see the ripples.
• The New Way (This Paper): You measure the ripples directly, right where they are, using a very precise ruler.

They call their method a "High-order exponential solver." Here is how it works in simple terms:

1. High-Order Rulers (Staggered Grids): Instead of using a standard ruler that might be a bit wobbly at the edges, they use a "high-order" ruler. This means they look at a wide area around each point to calculate the slope of the wave, making the measurement incredibly smooth and accurate. They also use "staggered" grids, which is like having two slightly offset rulers working together to catch every tiny detail without missing a beat.
2. Exponential Time Travel: To move the simulation forward in time, they use "exponential operators." Think of this as a time machine that doesn't just take small, shaky steps forward. Instead, it calculates the exact path the wave should take over a time step, skipping the middle ground where errors usually creep in.
3. Handling the Center (The Axis): The hardest part of simulating a cylinder is the very center (the axis), where the math gets tricky because everything converges to a single point. The authors developed special rules (boundary conditions) to handle this center point so the simulation doesn't break or create fake "ghost" particles.

The Laser Envelope Trick

The paper also introduces a shortcut for simulating the laser itself.

• The Full Wave: A laser is a wave that vibrates trillions of times per second. Simulating every wiggle is like trying to record every single frame of a spinning fan.
• The Envelope: Instead of recording every wiggle, the authors simulate the "envelope" (the shape of the fan's blur). They use their exponential method to move this shape forward with high precision. This is much faster and still very accurate, provided the laser beam is symmetrical.

Did it Work? (The Benchmarks)

The team tested their new method against the old "gold standard" (FBPIC) and full 3D simulations:

• Vacuum Test: They sent a laser through empty space. Their method matched the theoretical physics perfectly, with almost no energy loss or distortion.
• Plasma Test: They sent the laser through a gas (plasma). The results were nearly identical to the full 3D simulations and the FBPIC code.
• The "Bubble" Race: They simulated the complex scenario where the laser creates a "bubble" in the plasma that traps and accelerates electrons.
  • Result: Their new method reproduced the results of the full 3D simulation very well.
  • Comparison: Interestingly, the old "Fourier-Bessel" method (FBPIC) produced a slightly "smoother" but less energetic result near the center axis. The authors suggest their new method might actually be capturing the true, slightly "rougher" physics of the center better, while the old method smoothed it out too much.

The Bottom Line

This paper presents a new, highly accurate way to simulate laser-plasma interactions in cylindrical shapes. Instead of translating the problem into a special code and back, it solves the math directly in the real world using very precise, high-order steps.

It is faster than full 3D simulations, more accurate near the center axis than some existing cylindrical methods, and flexible enough to handle both the full laser wave and the simplified "envelope" version. The authors have shown that you can get high-precision results without needing the heavy computational cost of full 3D simulations or the complex translation steps of the old methods.

Technical Summary

Technical Summary: High-order Exponential Solver Method for Particle-in-Cell Simulations in Cylindrical Geometry

Problem Statement
Laser-driven particle acceleration schemes, particularly Laser Wakefield Acceleration (LWFA), rely heavily on Particle-in-Cell (PIC) simulations for microscopic understanding. While full 3D simulations are accurate, they are computationally expensive. To mitigate this, cylindrical geometry simulations utilizing azimuthal mode (AM) decomposition are employed, reducing computational and memory costs by orders of magnitude. However, existing solutions, such as the spectral Fourier-Bessel method used in the FBPIC code, trade off accuracy for efficiency, particularly regarding the representation of particles near the cylindrical axis and the need for special basis functions. Furthermore, standard finite difference methods in cylindrical coordinates often suffer from instability or non-physical spectral behavior at high frequencies. There is a need for a real-space equivalent of spectral methods that maintains high accuracy without relying on complex basis transformations, while also addressing specific numerical artifacts near the axis ($\rho=0$).

Methodology
The authors present a high-order exponential time-domain solver adapted for cylindrical geometry, building upon their previous work in Cartesian coordinates. The methodology involves several key components:

1. Exponential Maxwell Solver in Real Space:

   • The Maxwell equations are solved using a real-space finite difference approach combined with exponential time-stepping operators.
   • Spatial derivatives are approximated using high-order staggered finite differences (ranging from 6th to 32nd order).
   • Time evolution is handled via explicit Taylor expansions of exponential operators ($\exp(\Delta t \hat{H}_m)$), where $\hat{H}_m$ represents the Maxwell operator for each azimuthal mode $m$.
   • The method utilizes a real Fourier-series decomposition for the azimuthal variable, pairing sine and cosine coefficients to form real vectors, avoiding complex spectral transforms.

2. Cylindrical Discretization and Boundary Conditions:

   • A staggered grid (Yee-grid) is employed in the radial direction to ensure stability and suppress numerical Cherenkov radiation.
   • Special attention is paid to the behavior of field modes near the axis ($\rho=0$). The authors derive specific symmetric and antisymmetric boundary conditions for scalar and vector field modal coefficients based on their asymptotic behavior (e.g., $mS$ for scalar-like, $mA$ for polar-like modes).
   • High-order Lagrange interpolation is used to stagger and destagger fields between grid types, ensuring minimal accuracy loss.

3. Particle-in-Cell (PIC) Routine:

   • Current Deposition: A modified Esirkepov algorithm is adapted for the $z$-component. For transverse components ($J_\rho, J_\theta$), a current correction step is introduced. This involves solving a planar Poisson equation to correct divergence errors near the axis, which are otherwise difficult to define due to radial-angular coupling.
   • Charge Conservation: The authors introduce a 4th-order accurate quadrature scheme for density deposition to improve charge conservation near $\rho=0$, addressing issues where standard methods yield fluctuations exceeding 10%.
   • Angular Sampling: A scheme is implemented to allow random angular offsets for particles within a cell to reduce ordered grid artifacts and improve angular phase-space sampling.

4. Laser Envelope Model:

   • An exponential solver is developed for the scalar laser potential envelope equation. This allows for the propagation of the laser envelope without resolving the carrier frequency, significantly reducing computational cost for long pulses.
   • The model incorporates the ponderomotive guiding center approximation, where particles interact with cycle-averaged fields.

Key Contributions

• Real-Space Exponential Solver: The paper introduces a finite difference exponential time-domain (FDETD) method for cylindrical geometry that serves as a real-space equivalent to spectral methods, eliminating the need for Fourier-Bessel transforms.
• High-Order Accuracy: By utilizing high-order staggered finite differences (up to 32nd order) and exponential time-stepping, the method achieves spectral-like accuracy in real space.
• Axis Handling: The work provides a rigorous treatment of near-axis boundary conditions and particle representation, including a specific current correction algorithm to handle divergence errors at $\rho=0$.
• Laser Envelope Integration: The development of an exponential solver for the laser envelope potential allows for highly accurate and efficient simulations of laser propagation in underdense plasma, particularly when the envelope is cylindrically symmetric.

Results and Benchmarks
The authors verified the method through several benchmarks:

• Vacuum Propagation: Simulations of laser pulse propagation in vacuum demonstrated that the energy conservation and dispersion errors scale linearly with propagation distance and match the accuracy characteristics of their previous Cartesian solver. The method can achieve negligible dispersion and norm loss over long distances.
• Linear Propagation in Plasma: In underdense plasma, the solver accurately reproduced the analytical group velocity of the pulse centroid. The laser envelope model showed superior accuracy in group velocity and wakefield generation compared to the full Maxwell-AM decomposition, even at lower resolutions.
• Laser Wakefield Acceleration (LWFA): In a bubble regime simulation involving electron self-injection, the cylindrical AM solver reproduced the spatial density distribution of injected electrons with good agreement to 3D Cartesian simulations.
  • Comparison with FBPIC: While the spatial density distributions were similar, the spectral solution (FBPIC) produced less energetic electrons and a smoother spatial distribution near the axis compared to the high-resolution 3D reference. The authors attribute the FBPIC results to its binomial current smoothing, which suppresses artifacts but alters the high-energy tail.
  • Numerical Cherenkov: The method successfully suppressed the zero-order numerical Cherenkov effect in real space, a common issue in cylindrical PIC codes.

Significance
The paper claims that this method provides a very high-accuracy alternative to spectral methods for cylindrical PIC simulations without relying on special basis functions. It demonstrates that high-order finite differences combined with exponential operators can achieve accuracy comparable to spectral codes while offering the flexibility of real-space methods. The authors highlight that the laser envelope model, in particular, yields superior accuracy for group velocity and linear wakefields in underdense plasma compared to standard AM or 3D simulations. The work lays the foundation for future extensions, including quantum-electrodynamics modules (e.g., betatron radiation, pair creation) and field ionization, within the framework of relativistic laser-plasma interactions.