Benchmarking the plasmon-pole and multipole approximations in the Yambo Code using the GW100 dataset

This paper evaluates the numerical accuracy and convergence of the Yambo code's GW implementation by comparing the Godby-Needs plasmon-pole model and the multipole approximation against the GW100 benchmark dataset.

The Gist

The "Digital Microscope" Challenge: Making Better Predictions for New Materials

Imagine you are a master chef trying to invent a brand-new type of chocolate. You want to know exactly how it will melt, how sweet it will be, and how it will feel on the tongue. However, you aren't allowed to actually cook it yet—you have to predict everything using a complex mathematical recipe on your computer.

In the world of science, researchers do this with atoms and molecules. They use "digital recipes" (called computational models) to predict how new materials—like better batteries for electric cars or more efficient solar panels—will behave.

This paper is essentially a "Master Chef Competition" where scientists are testing a specific digital recipe called Yambo to see how accurate it is.

---

1. The Problem: The "Blurry Lens" of Physics

When scientists simulate how electrons (the tiny particles that make electricity flow) move around a molecule, they run into a massive problem: Complexity.

If you try to calculate every single tiny interaction of every electron perfectly, your computer would need to run for a thousand years. It’s like trying to predict the exact movement of every single drop of water in a crashing ocean wave. It’s too much data!

To solve this, scientists use "Approximations"—mathematical shortcuts. Think of these like looking at a distant mountain through a pair of binoculars. You don't see every pebble, but you get a very good idea of the mountain's shape.

2. The Contenders: Two Different Binoculars

The researchers in this paper are testing two specific types of "mathematical binoculars" within the Yambo software:

• The Plasmon-Pole Approximation (The "Sketch Artist"): This is an old, fast shortcut. It’s like a sketch artist who draws a person using just a few quick lines. It’s very fast, but sometimes it misses the subtle details of the face.
• The Multipole Approximation (The "High-Def Camera"): This is a newer, smarter shortcut. Instead of just a few lines, it uses several "points of focus" to capture the shape. It’s more detailed and much closer to the "real" picture, but it requires a bit more brainpower from the computer.

3. The Test: The GW100 "Gold Standard"

To see who is better, the researchers used the GW100 dataset. Think of this as the "Ultimate Taste Test." It is a collection of 100 different molecules that have already been tested by the world's best "chefs" (other scientists) using the most expensive, slow, and perfect methods possible.

The Yambo team ran their "recipes" on all 100 molecules and compared their results to the "Gold Standard" results.

4. The Results: A New Champion Emerges

The researchers found something very exciting:

1. The Old Way (Plasmon-Pole) is good, but not perfect. It gets the general idea right, but it occasionally misses the mark, especially with tricky molecules (like those containing heavy metals).
2. The New Way (Multipole) is a game-changer. It was significantly more accurate. It brought the digital simulation much closer to the "real" truth. It’s like upgrading from a blurry sketch to a high-definition photograph.
3. Speed vs. Accuracy: The best part? The Multipole method gives you "high-def" accuracy without needing a supercomputer that runs for a century. It’s the "sweet spot" of efficiency and precision.

Why does this matter to you?

We are currently in a race to discover materials that can solve climate change, create faster computers, and build better medical tools. We can't afford to waste years in a real lab testing things that don't work.

By proving that the Yambo code (specifically using the Multipole method) is highly accurate and reliable, this paper gives scientists a faster, more trustworthy "digital microscope." It means we can design the future of technology on a computer screen with much higher confidence before we ever step foot in a laboratory.

Technical Summary

Technical Summary: Benchmarking Plasmon-Pole and Multipole Approximations in Yambo

1. Problem Statement

In computational materials science, the GW approximation within Many-Body Perturbation Theory (MBPT) is the gold standard for calculating excited-state properties like quasiparticle energies. However, evaluating the frequency-dependent screened Coulomb interaction ($W$) is computationally expensive. To mitigate this, researchers use various frequency-integration schemes, ranging from Full-Frequency (FF) methods (highly accurate but costly) to Plasmon-Pole Approximations (PPA) (efficient but potentially less accurate).

While Density Functional Theory (DFT) is well-benchmarked, systematic code-to-code verification and validation (V&V) for GW methods are relatively recent. Specifically, there has been a lack of systematic benchmarking for the widely used Godby-Needs (GN-PPA) model against the newer, more sophisticated Multipole Approximation (MPA) using large-scale molecular datasets.

2. Methodology

The authors performed a high-throughput benchmarking study using the GW100 dataset, which consists of 100 closed-shell molecules covering a diverse range of chemical bonds and ionization potentials (IP).

• Code & Framework: The study utilized the Yambo code, implemented within a plane-wave pseudopotential framework. The workflow was fully automated using the aiida-yambo plugin, performing over 9,000 quasiparticle (QP) evaluations.
• Approximations Tested:
  • GN-PPA: A single-pole model that interpolates the dielectric response.
  • MPA: A recently developed method that generalizes the PPA by using multiple complex poles (typically $\sim10$) to better capture frequency dependence.
• Numerical Setup:
  • DFT Starting Point: Single-shot $G_0W_0$ based on PBE functionals.
  • Basis Set: Plane waves (PW) with optimized norm-conserving Vanderbilt (ONCV) pseudopotentials.
  • Convergence Strategy: To handle the interdependence of the number of empty states ($N_b$) and the kinetic-energy cutoff ($G_{cut}$), the authors employed a 2D simultaneous extrapolation scheme to the infinite limit.
  • Supercell: A face-centered cubic (FCC) geometry was used to optimize computational efficiency while managing periodic image interactions via the Martyna-Tuckerman method.

3. Key Contributions

• First Systematic Assessment: Provided the first large-scale comparison of GN-PPA and MPA within the Yambo plane-wave framework using the GW100 set.
• Validation of MPA: Demonstrated that the Multipole Approximation can achieve accuracy comparable to Full-Frequency (FF) methods at a significantly lower computational cost.
• Workflow Validation: Validated the reliability of the aiida-yambo high-throughput automation ecosystem for complex MBPT calculations.
• Benchmarking GN-PPA: Showed that the Godby-Needs model, contrary to the Hybertsen-Louie (HL) model used in other codes, is in good agreement with FF data.

4. Results

• Ionization Potentials (IP):
  • GN-PPA showed a Mean Absolute Error (MAE) of 168 meV relative to other codes.
  • MPA improved this accuracy, reducing the MAE to 114 meV and providing a narrower, more symmetric error distribution.
  • The MPA results were found to be highly consistent with other high-level implementations like VASP and WEST.
• Electron Affinities (EA):
  • The MPA approach also outperformed GN-PPA for EAs, with an MAE of 163 meV compared to 217 meV for GN-PPA, confirming its superior description of frequency dependence for unoccupied states.
• Outliers & Discrepancies: Large deviations (e.g., in $Xe$ or $F_2$) were attributed to the challenges of treating strongly localized $2p$ or $3d$ electrons in one-shot $G_0W_0$ and the inherent limitations of the finite supercell size.

5. Significance

This work establishes the Multipole Approximation (MPA) as a robust and highly efficient alternative to full-frequency integration for large-scale GW calculations. By proving that MPA provides "FF-quality" results with a fraction of the cost, the study paves the way for more accurate, high-throughput screening of electronic properties in complex molecular and material systems. Furthermore, it reinforces the reliability of the Yambo code for state-of-the-art excited-state simulations.