Accurate calculation of Wannier centers, position matrix, and composite operators using translationally equivariant and higher-order finite differences

This paper introduces a translationally equivariant and higher-order finite-difference scheme to improve the accuracy and symmetry preservation of momentum-space derivatives used in Wannier interpolation for calculating quantum geometric and response properties.

The Gist

Imagine you are trying to map out the exact location of every single person in a massive, crowded music festival using only a few snapshots taken from a drone every few minutes.

If you only have a few snapshots, you have to "guess" where people are moving between the photos. This paper is about a new, much more accurate way to make those guesses in the world of quantum physics.

The Problem: The "Blurry Snapshot" Dilemma

In quantum mechanics, scientists study how electrons move in solid materials (like the silicon in your phone). To do this, they use a mathematical tool called Wannier interpolation.

Think of the "true" movement of an electron as a smooth, continuous movie. However, because of the way computers work, scientists can't see the whole movie; they can only see "snapshots" at specific points (called a $k$-point grid). To fill in the gaps between snapshots, they use a method called Finite Difference.

The problem is that the standard way of "filling in the gaps" is like trying to draw a smooth curve using only straight, jagged lines. It’s often inaccurate, it breaks the "rules" of the system (like symmetry), and if you move the whole festival to a different field, the math suddenly thinks the people have changed shape!

The Solution: Two Major Upgrades

The researchers proposed two big improvements to make these "guesses" much smarter:

1. The "Equivariant" Upgrade (The Symmetry Rule)

Imagine if you moved your entire music festival from a park in New York to a park in London. The people are the same, the music is the same, and the layout is the same—only the GPS coordinates have changed.

The old math was "broken": if you shifted the system, the math would give you different answers, as if the electrons had physically transformed just because you moved the box they were in.

The authors created a "Translationally Equivariant" method. This is like giving the computer a "universal compass." Now, no matter where you move the system in space, the math stays consistent. It respects the natural symmetries of the crystal, ensuring that the "shape" of the electron's movement doesn't get distorted by the math itself.

2. The "Higher-Order" Upgrade (The Smooth Curve)

The old method was like trying to approximate a circle using a square. It’s "okay," but it’s clunky.

The authors introduced "Higher-Order Finite Differences." Instead of using just the immediate neighbors to guess a position, they look at a wider, more sophisticated pattern of points.

The Analogy:

• Old Method (First-Order): Like walking in a city by only ever turning at 90-degree angles. You’ll eventually get to your destination, but your path is jagged and inefficient.
• New Method (Higher-Order): Like a professional driver taking smooth, sweeping curves. You reach the same destination, but the path is much more realistic and requires much less "correction" to get right.

Why does this matter?

By using these two upgrades, scientists can get incredibly accurate results using much less "computer power."

In the paper, they tested this on real materials (like graphene and iron). They found that their new method:

1. Converges faster: You don't need as many "snapshots" to get the right answer.
2. Is more reliable: It doesn't "break" when you move the system or look at complex properties like magnetism or electricity.

In short: They’ve given physicists a high-definition lens to look at the microscopic world, replacing a blurry, jagged view with a smooth, consistent, and mathematically perfect one.

Technical Summary

Technical Summary: Accurate Calculation of Wannier Centers, Position Matrix, and Composite Operators

Title: Accurate calculation of Wannier centers, position matrix, and composite operators using translationally equivariant and higher-order finite differences
Authors: Jae-Mo Lihm, Minsu Ghim, Seung-Ju Hong, and Cheol-Hwan Park
Date: April 24, 2026 (Preprint/Published)

---

1. The Problem: Limitations of Standard Finite-Difference (FD) Methods

In first-principles calculations, momentum-space derivatives of Bloch wavefunctions are essential for studying quantum geometry, electric polarization, orbital magnetization, and Hall effects. These derivatives are typically computed via Wannier interpolation, where the $k$-derivative corresponds to the position operator $\hat{r}$.

Because $k$-space is sampled discretely, the position operator must be approximated using finite-difference (FD) schemes. The authors identify two critical flaws in the standard "Simple FD" (S-FD) approach:

• Broken Translational Equivariance: The S-FD approximation of the position operator is only accurate near the origin. As Wannier functions (WFs) move away from the origin (e.g., due to crystal shifts), the FD error grows significantly. Furthermore, the S-FD method violates the fundamental symmetry that a uniform translation of the system should shift Wannier centers by the same amount while leaving off-diagonal elements invariant.
• Slow Convergence: The error in standard FD approximations converges only quadratically ($O(b^2)$) with respect to the $k$-point grid spacing ($b$). This necessitates extremely dense $k$-point grids to achieve high precision, which is computationally expensive.

2. Methodology: Two Key Improvements

The authors propose a twofold enhancement to the FD scheme:

A. Translationally Equivariant Finite Difference (TEFD)
To restore symmetry, the authors reformulate the position operator approximation. Instead of centering the periodic approximant at the origin, they center it at the midpoint between the two Wannier centers involved in the matrix element:
$$\bar{r}_{ijR} = \frac{r_i + r_j + R}{2}$$
By shifting the approximation to the location where the product of the WFs has the largest magnitude, the error becomes translationally invariant and remains small even when the WFs are far from the origin. This method also ensures that the position matrix is Hermitian and respects the crystal symmetries of the system.

B. Higher-Order Finite Difference (HOFD)
To accelerate convergence, the authors introduce a higher-order scheme. While standard FD is $O(b^2)$, the HOFD method allows for an error scaling of $O(b^{2N})$, where $N$ is the order of the FD.

• The "Multiples" Method: To avoid the rapid growth in the number of $b$-vectors required by traditional "shells" methods, the authors propose a "multiples" scheme. This uses integer multiples of the first-order $b$-vectors, making the implementation computationally efficient and scalable ($O(N)$ vs $O(N^3)$).

C. Generalization to Composite Operators
The authors extend the TEFD framework to composite operators (e.g., $\hat{r}\hat{H}$ or $\hat{s}\hat{r}$), which are required for complex properties like orbital magnetization and spin Hall conductivity. They derive new equivariant formulas for matrix elements of the form $B_{\alpha}^{ijR}[\hat{O}]$ and $C_{\alpha\beta}^{ijR}[\hat{O}]$.

3. Key Contributions

1. TEFD Formalism: A new, robust method for calculating the full position matrix (diagonal and off-diagonal) that is translationally equivariant and Hermitian.
2. HOFD Implementation: A systematic higher-order FD approach using the "multiples" method to achieve rapid convergence.
3. Composite Operator Framework: A generalized equivariant scheme for operators involving position and other physical operators (Hamiltonian, spin).
4. Vector-Coefficient FD: A generalized parametrization that can reduce the number of required $b$-vectors while maintaining accuracy.

4. Results and Validation

The effectiveness of the methods was demonstrated through several benchmark calculations:

• Wannier Centers & Polarization: In monolayer GeS, the TEFD method correctly handled out-of-plane shifts where S-FD failed. In KNbO$_3$, the HOFD method showed significantly faster convergence for macroscopic polarization.
• Orbital Magnetization: For bcc Fe and CrO$_2$, the TEFD method yielded results that were invariant under crystal translation, whereas S-FD produced errors that depended on the atomic position relative to the origin.
• Spin Hall Conductivity: In GaAs, the TEFD/HOFD methods provided much higher precision and faster convergence compared to existing methods (e.g., Ryoo et al.).
• Symmetry Preservation: The TEFD method successfully preserved the $C_3$ rotational symmetry in graphene's interband velocity matrix elements, a symmetry that S-FD violated.

5. Significance

This work provides a mathematically rigorous and computationally efficient foundation for the next generation of Wannier-based electronic structure tools. By solving the fundamental issues of equivariance and convergence, the authors enable the accurate study of topological and geometric properties of solids with much coarser $k$-point grids. The methods have already been integrated into major open-source packages, including Wannier90, WannierBerri, and EPW, ensuring widespread adoption in the condensed matter physics community.