Wigner-Seitz truncated TDDFT approach for the calculation of exciton binding energies in solids

This paper demonstrates that while hybrid-TDDFT approaches improve upon standard long-range corrected kernels for calculating exciton binding energies, the persistent discrepancy between accurate spectra and binding energies stems from the numerical difficulties in treating the long-range Coulomb singular term, highlighting the need for a more robust description of electron-hole interactions.

The Gist

The Big Picture: Catching the "Dancing Pairs" in a Crystal

Imagine a solid material, like a piece of silicon or a diamond, as a giant, crowded dance floor. The dancers are electrons. Usually, they dance independently, moving to their own rhythm. But sometimes, if you hit the material with light (like a flash of sunshine), an electron gets excited, jumps up to a higher energy level, and leaves a "hole" behind where it used to be.

Here's the magic part: The excited electron and the hole it left behind are attracted to each other, like two magnets. They form a temporary, dancing pair called an exciton.

Scientists want to know exactly how tightly these pairs are holding hands. This is called the exciton binding energy. If you know this number, you can design better solar panels, LEDs, and computer chips.

The Problem: The "Mathematical Glitch"

To predict how these pairs behave, scientists use a powerful computer tool called TDDFT (Time-Dependent Density Functional Theory). It's like a super-accurate weather forecast for electrons.

However, there is a major glitch in the math. The force that holds the electron and hole together is the Coulomb force (the same force that makes your hair stand up when you rub a balloon on it). In a crystal, this force has a "singular" point—a mathematical singularity—where the numbers try to divide by zero.

Think of it like trying to calculate the speed of a car by dividing the distance by zero time. The computer crashes, or worse, it gives you a wrong answer that looks plausible but is actually nonsense.

The paper argues that for years, scientists have been trying to fix this glitch by tweaking a "knob" (a parameter called $\alpha$) to make the results match experiments. But the authors say: "We are just turning the knob to hide the fact that our math is broken at the singularity."

The Investigation: Two Ways to Fix the Glitch

The authors decided to investigate this "broken math" in two different ways.

1. The "Pure" Approach (TDDFT)

In this method, they tried to solve the problem using standard rules.

• The Analogy: Imagine you are trying to measure the distance between two people in a room. You use a rule that says, "Distance = Speed × Time." But in a crystal, the "Time" part is tricky because the room is infinite and repeats itself.
• The Discovery: The authors found that when scientists convert the math from "position" to "momentum" (a common trick to avoid the division-by-zero error), they accidentally throw away a tiny "surface term."
• The Result: This thrown-away term is actually huge! It's like ignoring the weight of the floor when calculating how much a building weighs. Because they ignored this term, the math was off by a factor of 10 to 30. They realized that the "knob" ($\alpha$) scientists were turning was just compensating for this missing weight.

2. The "Hybrid" Approach (Screened Exact Exchange)

Since the pure method was messy, they tried a hybrid method. This is like mixing a standard recipe with a secret ingredient (Exact Exchange) to get a better taste.

• The Innovation: They used a technique called Wigner-Seitz Truncation.
• The Analogy: Imagine the crystal is an infinite ocean. Calculating the force between two points in an infinite ocean is impossible. So, they built a fence around a single "cell" of the ocean (a Wigner-Seitz cell). They calculated the force only inside that fenced area and ignored the rest.
• The Result: This worked beautifully for some materials (like semiconductors GaAs or GaN). It was like finding a shortcut that gave the right answer without the heavy math. However, for very hard insulators (like Magnesium Oxide), the fence was too small. The "local field effects" (the neighbors peeking over the fence) were too important to ignore, so the method underestimated the binding energy.

The Conclusion: Why This Matters

The paper concludes that the "singular term" (the division-by-zero problem) is the most important part of the equation.

• The Takeaway: You cannot just guess the right answer by tweaking a knob. You have to handle the mathematical singularity correctly.
• The Future: The authors suggest that if we want to design perfect new materials, we need to stop ignoring the "surface terms" and the "singularities." We need a better way to describe how the electron and hole interact when they are right next to each other in a repeating crystal.

In short: The authors found that the math used to predict how electrons dance in solids has a hidden "bug" at the center of the calculation. They showed that fixing this bug (or at least acknowledging it) is the key to getting accurate predictions for the next generation of technology.

Technical Summary

1. Problem Statement

The accurate calculation of exciton binding energies in solids is a persistent challenge in computational materials science. While Time-Dependent Density Functional Theory (TDDFT) offers a computationally efficient alternative to Many-Body Perturbation Theory (MBPT/BSE) for optical absorption spectra, it struggles to simultaneously reproduce accurate spectra and correct exciton binding energies.

The core issues identified are:

• Parameter Discrepancy: Long-Range Corrected (LRC) kernels, which successfully model optical spectra, require an empirical parameter ($\alpha$) that differs by orders of magnitude when fitted for spectra versus binding energies.
• Numerical Singularity: The primary contribution to exciton binding energies comes from the singular long-range Coulomb term ($q \to 0, G=0$) in the exchange-correlation kernel. In periodic solids, this term is ill-defined (0/0 indeterminate form) because valence and conduction Bloch states are orthogonal.
• The $p-r$ Relation Flaw: Standard approaches handle this singularity using the commutator relation (momentum-position relation, $p-r$) to transform the position matrix element into a momentum matrix element. The authors argue this method neglects a crucial surface term ($C_{ck,vk}$) that arises in infinite periodic systems, leading to significant underestimation of matrix elements and erroneous binding energy predictions.

2. Methodology

The authors investigated the numerical treatment of the Coulomb singularity in two frameworks: Pure TDDFT and Hybrid TDDFT (incorporating Screened Exact Exchange, SXX).

A. Pure TDDFT Analysis

• Re-evaluation of the $p-r$ Relation: The authors derived the correct relation between momentum and position matrix elements in periodic systems, explicitly including the surface term $C_{ck,vk}$:
  $$\langle ck | \hat{p} | vk \rangle = i(\epsilon_{ck} - \epsilon_{vk}) \langle ck | \hat{r} | vk \rangle + C_{ck,vk}$$
• Direct Computation: Instead of relying on the $p-r$ trick, they developed a custom code to directly compute matrix elements for a finite, small $q$ grid (avoiding the $q=0$ limit directly) to evaluate the magnitude of the correction term $C_{ck,vk}$.
• Testing: They applied this to GaAs, $\beta$-GaN, and MgO using Cohen-Bergstresser pseudopotentials and Quantum ESPRESSO for ground states.

B. Hybrid TDDFT (WS-SXX)

• Wigner-Seitz Truncation: To address the divergence in the Screened Exact Exchange (SXX) kernel ($1/|q+G|^2$), the authors implemented a Wigner-Seitz (WS) truncated kernel. This method truncates the Coulomb interaction within a WS supercell, providing a well-defined analytical value at $G=0$.
• Convergence Study: They analyzed the convergence of exciton binding energies with respect to:
  • $k$-point sampling ($N_k$) in the Brillouin Zone.
  • Dielectric screening parameter ($\gamma$).
  • Number of valence/conduction bands.
• Materials Tested: A set of semiconductors (GaAs, CdS, GaN, AlN) and wide-bandgap insulators (MgO, LiF, Ar, Ne).

3. Key Contributions

1. Identification of the Surface Term Error: The paper demonstrates that the correction term $C_{ck,vk}$, often neglected in the $p-r$ relation, is non-trivial in periodic solids. For GaAs, this term is ~31 times larger than the position matrix element itself. Neglecting it leads to underestimation of matrix elements and necessitates artificially high $\alpha$ parameters to fit experimental data.
2. Numerical Origin of Discrepancy: The authors conclude that the discrepancy between spectral fitting and binding energy fitting in LRC-TDDFT is primarily numerical, stemming from how the $q \to 0$ singularity is handled, rather than a fundamental failure of the kernel form.
3. WS-Truncated SXX Kernel: They propose and validate a Wigner-Seitz truncated SXX kernel that yields a finite, well-behaved analytical term at the optical limit ($q=0$).
4. Convergence Insights: They highlight that exciton binding energies are extremely sensitive to $k$-point sampling ($N_k$), decaying as $N_k^{-1/3}$, and that the dielectric screening parameter $\gamma$ is linearly proportional to the binding energy.

4. Results

• Pure TDDFT:
  • When the singular term ($G=0$) is ignored, fitted $\alpha$ parameters are 15–30, showing no trend with material type.
  • Including the singular term (Head-only or Diagonal) drastically reduces the required $\alpha$ values.
  • The parabolic trend of $\alpha$ vs. bandgap observed in previous literature (using the $p-r$ relation) is attributed to the neglect of the surface term $C_{ck,vk}$.
• Hybrid WS-SXX:
  • Semiconductors: The WS-SXX approach yields binding energies in good agreement with BSE and experimental values for semiconductors (e.g., GaAs, GaN, AlN). It performs particularly well for Wurtzite structures (w-GaN, AlN) due to efficient handling of the elongated $c$-axis symmetry.
  • Insulators: The method significantly underestimates binding energies for wide-bandgap insulators (e.g., MgO, LiF, Ar).
  • Cause of Underestimation: The authors attribute this to the WS-SXX kernel's focus on long-range interactions while neglecting local field effects, which are critical for localized Frenkel excitons in insulators.
  • Convergence: Results are highly dependent on the $k$-grid. Fully converged results may vary significantly from current practical grid sizes, presenting a numerical bottleneck.

5. Significance and Conclusion

• Theoretical Impact: The work challenges the standard practice of using the $p-r$ relation without surface corrections in periodic systems, suggesting that current TDDFT implementations may be systematically flawed for binding energy calculations.
• Methodological Advancement: The Wigner-Seitz truncated kernel offers a robust, computationally affordable alternative to full BSE for semiconductors, bridging the gap between TDDFT and MBPT.
• Future Directions: The authors conclude that while the WS-SXX approach is promising, a "better description of the electron-hole interaction" is needed, particularly for insulators where local field effects cannot be ignored. The findings urge a re-examination of how Coulomb singularities are treated in all excitonic calculations to ensure numerical accuracy.

In summary, the paper identifies that the difficulty in calculating exciton binding energies is largely a numerical artifact related to the treatment of the Coulomb singularity and the neglect of surface terms in periodic boundary conditions, rather than an inherent limitation of the TDDFT formalism itself.