Quantum chemistry based on classical mechanics inspired by simulated bifurcation

The Quantum Chemistry Race: A New Way to Find the Fastest Route

Imagine you are trying to find the absolute best route for a delivery truck to visit 100 different cities. This is a classic "combinatorial optimization" problem. If you try every single possible route, it would take longer than the age of the universe.

In the world of chemistry, scientists face a similar, even harder problem. They want to know exactly how electrons behave inside a molecule to predict its properties (like how strong a bond is or how it reacts). To do this perfectly, they use a method called Full Configuration Interaction (FCI).

The Problem:
Think of the electrons in a molecule as a chaotic crowd of people at a concert. To know exactly where everyone is and how they are moving, you have to track every single possible arrangement of the crowd.

The Old Way (Davidson Method): The standard method used for decades is like a very careful, methodical detective. It checks one clue, then another, slowly narrowing down the search. It's accurate, but for large molecules (big crowds), it takes forever and requires a massive amount of memory (like needing a library the size of a city just to store the clues).
The Quantum Dream: Scientists hoped quantum computers would solve this instantly. But building a reliable quantum computer is like trying to build a spaceship out of toothpicks—it's still under construction.

The New Solution: SBCI (Simulated Bifurcation-based CI)
The authors of this paper, Fumihiko Aiga and Hayato Goto, have invented a new algorithm called SBCI. They didn't wait for the quantum spaceship; instead, they built a super-fast, classical car that drives on a new kind of road.

Here is how it works, using simple analogies:

1. The "Rolling Ball" Metaphor

Imagine you are trying to find the lowest point in a vast, foggy valley (the lowest energy state of a molecule).

The Old Way: You take a step, check the ground, take another step, check again. You are very careful, but you move slowly.
The SBCI Way: Imagine you are a ball rolling down that hill. But this isn't just any ball; it's a ball with momentum.
- In SBCI, the "position" of the ball represents the current guess of the electron arrangement.
- The "momentum" (speed) helps the ball carry itself forward, even if the ground is bumpy.
- The algorithm uses the laws of classical mechanics (like Newton's laws) to simulate this ball rolling. Because the ball has momentum, it doesn't get stuck in small dips; it rolls right over them to find the deepest valley much faster.

2. The "Bifurcation" (The Fork in the Road)

The name "Simulated Bifurcation" comes from the idea of a fork in the road.

Imagine the ball is rolling toward a fork. In a normal search, you might have to stop and decide which way to go.
In SBCI, the algorithm is inspired by a quantum trick where the system naturally "splits" or "bifurcates" to explore both paths simultaneously in a way that helps it settle into the best solution quickly. It's like the ball magically knows which fork leads to the deepest valley without having to stop and think about it.

3. The "Adaptive Restart" (The Pit Stop)

Sometimes, the ball might get a little too fast and overshoot the target, or it might get stuck in a weird loop.

The SBCI algorithm has a clever "pit stop" mechanism. It constantly checks the size of the ball's path. If the path gets too wild (too big) or too small, the algorithm says, "Okay, let's reset!"
It takes the current best guess, cleans it up, and launches the ball again with a fresh burst of energy. This "adaptive restart" prevents the calculation from wasting time on dead ends.

The Results: Faster and Leaner

The authors tested this new "rolling ball" method against the old "careful detective" method (Davidson) on several molecules like Nitrogen ( $N_2$ ), Water ( $H_2O$ ), and Carbon ( $C_2$ ).

Speed: SBCI was significantly faster. For some difficult molecules, it finished the job in half the time.
Memory: It used much less computer memory. If the old method needed a warehouse to store its notes, SBCI only needed a backpack.
Accuracy: Despite being faster, it was just as accurate as the slow, careful method. It found the exact same answers.

Why This Matters

This is a big deal because:

No Waiting for Quantum Computers: We don't have to wait for the "spaceship" (quantum computers) to be built to solve these hard chemistry problems. We can solve them now on regular supercomputers.
Better Materials and Medicine: Faster, cheaper, and more accurate calculations mean scientists can design better batteries, new medicines, and stronger materials much faster than before.
A New Standard: The authors suggest that SBCI could replace the old "Davidson method" as the standard tool for chemists, just like how GPS replaced paper maps for navigation.

In a nutshell: The authors took a concept from quantum optimization (Simulated Bifurcation), turned it into a classical physics simulation (a rolling ball with momentum), and used it to solve chemistry problems faster and cheaper than ever before, all without needing a quantum computer.

Here is a detailed technical summary of the paper "Quantum chemistry based on classical mechanics inspired by simulated bifurcation."

1. Problem Statement

Accurate quantum chemical calculations, particularly Full Configuration Interaction (FCI), are essential for understanding molecular properties and strong electron correlation (e.g., in transition metal complexes or bond dissociation). However, FCI faces a "combinatorial explosion" where the number of configurations grows exponentially with the number of electrons and orbitals.

Current Limitations: The standard method for solving the resulting eigenvalue problems on classical computers is the Davidson method. While effective, it suffers from high computational costs regarding both execution time and memory usage, especially as the system size increases or when calculating multiple excited states.
The Gap: While quantum computers are anticipated to solve these problems, practical, large-scale quantum hardware is not yet available. There is an urgent need for more efficient classical algorithms that can match the accuracy of the Davidson method while significantly reducing computational resources.

2. Methodology: Simulated Bifurcation-based CI (SBCI)

The authors propose a new algorithm called Simulated Bifurcation-based CI (SBCI). This approach maps the eigenvalue problem of the Configuration Interaction (CI) coefficients onto the dynamics of a classical mechanical system, inspired by the Simulated Bifurcation (SB) algorithm originally developed for combinatorial optimization.

Core Concepts

Classical Mechanical Mapping: Instead of treating CI coefficients as static vectors to be iteratively updated, SBCI treats them as the positions ( $\mathbf{x}$ ) of a classical system. Conjugate momenta ( $\mathbf{y}$ ) are introduced to construct a classical Hamiltonian ( $H_{SBCI}$ ).
Hamiltonian Formulation: The potential energy of this classical system is proportional to the Rayleigh quotient (the energy expectation value), which the algorithm seeks to minimize. The kinetic energy depends on the momentum vectors.
Dynamics: The system evolves according to Hamilton's equations of motion, solved numerically using the symplectic Euler method.
Variational Parameter Control: Unlike the original SB algorithm which uses adiabatic time evolution, SBCI variationally determines the time-dependent parameters ( $b_t, c_t$ ) at each step. These parameters are tuned to ensure the updated trial vector aligns with the lowest (or next lowest) eigenvector of the Hamiltonian within a specific subspace.
Adaptive Restarts: A distinctive feature of SBCI is that the trial wavefunction is not normalized during the evolution. The norm of the wavefunction is monitored; if it deviates too far from unity, an "adaptive restart" is triggered to reset the parameters, accelerating convergence.

Algorithm Variants

The authors implemented two versions to handle different state requirements:

SBCI1 (Single-State Update): Updates one state ( $\alpha$ ) at a time, determining states sequentially from the ground state upward. It stores only two vectors ( $\mathbf{y}$ and $\mathbf{z}$ ) per state during updates.
SBCI2 (Two-State Update): Updates a pair of states ( $\alpha, \alpha+1$ ) simultaneously. This is designed to handle nearly degenerate states more efficiently than sequential updates. It stores four vectors per pair.

Implementation

Implemented within the PySCF (Python-based Simulations of Chemistry Framework).
Replaces the standard Davidson solver (davidson1) in the library.
Uses the standard preconditioner $(D - E_0 I)^{-1}$ (where $D$ is the diagonal of $H$ ) as the inverse mass matrix.

3. Key Contributions

Novel Algorithm: Introduction of SBCI, the first application of Simulated Bifurcation dynamics to solve electronic structure eigenvalue problems directly, distinct from Ising machine approaches or Density Matrix Renormalization Group (DMRG).
Memory Efficiency: SBCI significantly reduces memory footprint. Unlike the Davidson method, which must store the history of residual vectors for all unconverged states, SBCI only requires storing a small, fixed number of vectors (2 for SBCI1, 4 for SBCI2) regardless of the number of iterations.
Adaptive Restart Mechanism: The non-normalized evolution allows for dynamic restarts based on the wavefunction norm, a feature not present in standard iterative eigensolvers, leading to faster convergence.
High Accuracy: The method achieves accuracy comparable to the standard Davidson method and literature benchmarks, validating its reliability for exact diagonalization.

4. Results

The authors tested SBCI1 and SBCI2 on various molecular systems (N $_2$ , CN, H $_2$ O, HF, BH, C $_2$ , Ne, H $_2$ O $^+$ , F $_2$ ) using Full Configuration Interaction (FCI).

Accuracy:
- The total energies and excitation energies obtained by SBCI matched those of the Davidson method within numerical precision (differences typically $< 10^{-9}$ Hartree).
- Results were consistent with high-accuracy literature values.
Performance (Execution Time):
- Ground States: SBCI1 was consistently faster than the Davidson method across all bond lengths tested for N $_2$ and CN. The performance advantage increased as bond lengths increased (a regime where electron correlation is stronger).
- Excited States: SBCI2 was the fastest method for all excited state instances tested (H $_2$ O, HF, N $_2$ , BH, C $_2$ ). In many cases, SBCI2 reduced execution time by 30–50% compared to Davidson.
Performance (Memory Usage):
- SBCI1 and SBCI2 demonstrated significantly lower memory consumption. For example, in the C $_2$ excited state calculation, memory usage dropped from ~359 GB (Davidson) to ~133 GB (SBCI2).
Scalability:
- The algorithms showed favorable scaling with the number of CPU cores, with SBCI2 often achieving the shortest computation times in multi-core environments.

5. Significance

Immediate Impact: SBCI offers a practical, high-performance alternative to the Davidson method for classical quantum chemistry calculations. It enables larger-scale FCI calculations that were previously limited by memory constraints.
Bridge to Quantum Computing: By leveraging concepts from quantum-inspired optimization (Simulated Bifurcation) on classical hardware, this work demonstrates that classical mechanics-inspired algorithms can rival traditional linear algebra methods in the quantum chemistry domain.
Future Directions: The authors suggest that SBCI can be extended to Multiconfigurational Self-Consistent Field (MCSCF) calculations (e.g., CASSCF) and that multi-node implementations using MPI could further accelerate large-scale computations, potentially replacing the Davidson method as the standard for CI calculations in the near future.

In conclusion, the paper presents a robust, efficient, and accurate classical algorithm that successfully accelerates high-precision electronic structure calculations without compromising reliability, addressing a critical bottleneck in computational chemistry.