Determining Molecular Ground State with Quantum Imaginary Time Evolution using Broken-Symmetry Wave Function

This paper proposes enhancing Quantum Imaginary Time Evolution (QITE) for open-shell molecular systems by replacing the standard Hartree-Fock wave function with a spin- and spatial-symmetry broken wave function augmented by an $S^2$ penalty term, which demonstrates superior convergence and initial overlap with exact solutions in multi-configurational regimes such as diradical character and bond dissociation.

The Gist

The Big Picture: Finding the "Coziest" Spot in a Stormy Room

Imagine you are trying to find the absolute lowest, most comfortable spot in a room filled with rolling hills and valleys. In the world of quantum chemistry, this "lowest spot" is the Ground State—the most stable, lowest-energy arrangement of electrons in a molecule. Finding this spot is crucial for designing new drugs, better batteries, and understanding how chemical reactions work.

The paper tackles a specific problem: How do we find this spot quickly and accurately when the molecule is "messy"?

The Problem: The "Perfectly Organized" Map vs. The Real Chaos

Traditionally, scientists use a method called Hartree-Fock (HF) to start their search. Think of HF as a map drawn by someone who loves order. It assumes every electron has a perfect partner and sits neatly in a specific seat. This works great for calm, stable molecules (like a tidy bedroom).

However, when a molecule is under stress—like when a chemical bond is stretching or breaking (think of pulling a rubber band until it snaps)—the electrons get chaotic. They stop pairing up neatly and start acting like wild, unpaired kids running around. This is called an Open Shell System.

If you try to use the "tidy map" (HF) to navigate this "chaotic room," you get lost. The map doesn't match reality, so the computer algorithm takes forever to find the bottom of the valley, or it gets stuck in a shallow dip that isn't the true lowest point.

The Solution: The "Broken Symmetry" Map and the "Bouncer"

The authors propose a clever two-part fix to speed up the search:

1. The "Broken Symmetry" (BS) Map

Instead of forcing the electrons to look tidy, the authors suggest starting with a map that admits the chaos. They use a Broken Symmetry (BS) wave function.

• The Analogy: Imagine you are looking for a lost dog in a park. The HF map says, "The dog is sitting perfectly still in the center." The BS map says, "Okay, the dog is actually running around wildly, maybe sniffing the left side and the right side at the same time."
• Why it helps: By starting with a map that acknowledges the electrons are "broken" or unpaired, the computer is already much closer to the real answer when the bond starts breaking.

2. The "Spin Penalty" (The Bouncer)

Even with the better map, the computer might still get confused between the "true ground state" and a "fake ground state" that looks similar but has the wrong spin (like a dog wearing a hat when it shouldn't).

To fix this, they add a Penalty Term (specifically the $\hat{S}^2$ operator) to the rules of the game.

• The Analogy: Imagine the computer is a hiker trying to find the lowest valley. There are two valleys: the real one (Singlet) and a fake one (Triplet) that looks almost the same but is slightly higher up.
• The Penalty: The authors put a "bouncer" at the entrance of the fake valley. If the hiker tries to enter the fake valley (which has the wrong spin), the bouncer charges a heavy fine (adds energy).
• The Result: The fake valley suddenly becomes a mountain. The hiker (the algorithm) is forced to ignore the fake path and zoom straight down into the real, lowest valley. This creates a bigger gap between the right answer and the wrong answers, making the search much faster.

The Experiments: Testing the New Strategy

The team tested this idea on three different "rooms" (molecules):

1. The Hydrogen Molecule ($H_2$): A simple two-electron system.

   • Finding: When the bond is short and tight (like a fresh rubber band), the old "Tidy Map" (HF) is actually faster. But as soon as you stretch the bond and the electrons get "diradical" (unpaired, like a stretched rubber band about to snap), the "Broken Symmetry" map with the "Bouncer" wins by a landslide. It found the answer in 260 steps instead of 440 steps.
   • The Threshold: They found a "tipping point" (a diradical character of 0.21). Below this, use the old map. Above this, switch to the new map.

2. The Nitrogen Molecule ($N_2$): A tougher molecule with triple bonds.

   • Finding: As the triple bonds break, the "Broken Symmetry" approach was significantly better at finding the ground state, especially when the bonds were stretched far apart.

3. The Tetrahydrogen Cluster ($P_4$): A square of four hydrogen atoms.

   • Finding: This system is very messy (strongly correlated). The old method took 1,500 steps to converge, while the new method took only 950 steps.

Why This Matters

In the world of quantum computing, time is money (or rather, "qubits" are expensive because they are fragile). Every extra step the computer takes increases the chance of errors.

• The Takeaway: By starting with a "messier" but more realistic map (Broken Symmetry) and adding a "bouncer" to push away wrong answers (Spin Penalty), we can find the ground state of difficult molecules much faster.
• The Future: This doesn't just help scientists understand chemistry; it paves the way for quantum computers to actually solve real-world problems like designing new medicines or better battery materials, where these "messy" electron interactions are the norm, not the exception.

Summary in One Sentence

The paper teaches us that when searching for the lowest energy state of a chaotic molecule, it's better to start with a realistic, "messy" map and add a penalty for wrong turns, rather than sticking to a perfect but unrealistic map that gets you stuck in traffic.

Technical Summary

1. Problem Statement

The paper addresses the challenge of Ground State Energy Estimation (GSEE) for open-shell systems (molecules with unpaired electrons) using quantum computing algorithms, specifically Quantum Imaginary Time Evolution (QITE).

• The Core Issue: Standard QITE and Adiabatic State Preparation (ASP) rely heavily on the initial wave function's overlap with the true ground state. For open-shell systems (e.g., during bond dissociation), the ground state exhibits strong multi-configurational character.
• Limitation of Hartree-Fock (HF): The conventional starting point, the spin-restricted Hartree-Fock (RHF) wave function, is a single-determinant approximation. It fails to accurately describe open-shell systems where electrons are unpaired, leading to:
  • Poor initial overlap with the true ground state.
  • A narrowing energy gap between the ground state and excited states (quasi-degeneracy).
  • Slow convergence or failure to reach the ground state within a feasible number of iterations, especially in the Noisy Intermediate-Scale Quantum (NISQ) era.

2. Methodology

The authors propose a hybrid approach combining Broken-Symmetry (BS) wave functions with a Spin Penalty term in the Hamiltonian to accelerate QITE convergence.

A. Broken-Symmetry (BS) Wave Function

Instead of using the RHF state, the authors utilize a BS wave function.

• Definition: A BS state is a single-determinant wave function that breaks spin and spatial symmetry to approximate a multi-configurational state. It is constructed as a linear combination of high-spin and low-spin eigenfunctions.
• Advantage: BS states can be prepared on a quantum computer using only Pauli-X gates (state preparation depth is shallow), unlike complex multi-reference states (like CASSCF) which require deep circuits.
• Construction: The coefficients for the linear combination are derived using a branching diagram (Appendix A), which maps the superposition of spin eigenstates ($|N, S, M\rangle$) required to form the specific occupancy pattern (e.g., $|\alpha\beta\rangle$ for H$_2$).

B. Spin Penalty Hamiltonian

To further accelerate convergence, the authors modify the system Hamiltonian ($\hat{H}$) by adding a penalty term proportional to the total spin operator squared ($\hat{S}^2$):
$$\hat{H}' = \hat{H} + c\hat{S}^2$$

• Mechanism: The penalty term raises the energy of high-spin excited states (where $S$ is large) relative to the low-spin ground state.
• Effect: This artificially widens the energy gap between the ground state and the excited states. Since the convergence rate of QITE is inversely proportional to this energy gap, the evolution converges significantly faster.
• Synergy with BS: While RHF is an eigenfunction of $\hat{S}^2$ (making the penalty ineffective for lifting degeneracy in RHF-based evolution), the BS wave function is a superposition of different spin multiplicities (e.g., singlet and triplet). The penalty term effectively lifts the quasi-degeneracy of the triplet excited state, allowing the system to evolve rapidly toward the singlet ground state.

C. Simulation Techniques

• Algorithms: The study employs both Direct Matrix Exponentiation (for exact simulation of N$_2$) and Measurement-Assisted Unitary Approximation (for QITE on H$_2$ and P$_4$).
• Mapping: Fermion-to-qubit mapping was performed using the Symmetry-Conserving Bravyi–Kitaev Transformation (SCBKT).
• Systems Studied:
  • H$_2$: Bond dissociation (STO-6G basis).
  • N$_2$: Triple bond dissociation (STO-3G basis), analyzing $\pi$ and $\sigma$ bond breaking.
  • P$_4$ (Square Tetrahydrogen): A cluster with strong open-shell character.

3. Key Contributions

1. Novel Initial State Strategy: Demonstrated that replacing the standard RHF initial state with a BS wave function significantly improves QITE performance for open-shell systems.
2. Penalty Term Integration: Validated the use of the $\hat{S}^2$ penalty term specifically in conjunction with BS wave functions to widen the energy gap and accelerate convergence.
3. Diradical Character Threshold: Identified a critical diradical character ($y$) threshold that dictates the optimal choice of initial wave function:
   • Low $y$ (Closed Shell/Near Equilibrium): RHF is superior (faster convergence).
   • High $y$ (Open Shell/Bond Dissociation): BS wave functions are superior.
4. Scalability Insight: Showed that for larger systems (like N$_2$), BS wave functions provide a significantly higher initial overlap with the Complete Active Space Configuration Interaction (CAS-CI) ground state compared to RHF, particularly in multi-configurational regimes.

4. Results

• N$_2$ Molecule (Matrix Exponentiation):
  • At a bond length of 2.7 Å, BS-based QITE converged faster than RHF-based QITE.
  • Initial Fidelity: BS wave functions showed higher initial overlap with the CAS-CI ground state ($\approx 0.14$ for BS2, $\approx 0.25$ for BS3) compared to RHF ($\approx 0.08$).
  • Thresholds:
    • $y_\pi < 0.48$ (Bond length < 1.62 Å): RHF is faster.
    • $0.48 < y_\pi < 0.35$ (1.62 Å < Length < 2.08 Å): BS2 is optimal.
    • $y_\sigma > 0.35$ (Length > 2.08 Å): BS3 is optimal.
• H$_2$ Molecule (Unitary Approximation):
  • At a bond length of 2.0 Å ($y = 0.56$), BS-QITE reached chemical precision ($\Delta E \le 1.0$ kcal/mol) in 260 steps, whereas HF-QITE required 440 steps.
  • Transition Point: The optimal switch from HF to BS occurs at a bond length of 1.56 Å ($y = 0.21$). Below this, HF is better; above it, BS is superior.
• P$_4$ Cluster:
  • For the square geometry (2.0 Bohr), BS-QITE achieved chemical precision in ~950 steps, while HF-QITE required ~1500 steps.
  • Although HF started with higher fidelity, the BS trajectory rose above it quickly due to the penalty term's effect on the energy gap.

5. Significance and Implications

• Practical Quantum Chemistry: This method provides a pathway to simulate strongly correlated open-shell systems (crucial for catalysis, magnetic materials, and reaction pathways) on near-term quantum devices where circuit depth and iteration count are limited.
• Resource Efficiency: By reducing the number of QITE iterations required to reach convergence, the approach directly reduces the total circuit depth and the number of measurements needed, mitigating noise accumulation.
• Adaptive Strategy: The paper establishes a clear protocol for selecting the initial wave function based on the diradical character of the system, offering a practical heuristic for quantum chemists.
• Overhead Management: While the $\hat{S}^2$ penalty adds terms to the Hamiltonian (increasing circuit depth per step), the authors argue this overhead is outweighed by the significant reduction in the total number of iterations required.

In conclusion, the paper successfully demonstrates that Broken-Symmetry wave functions combined with a spin penalty offer a robust, efficient, and scalable strategy for ground state energy estimation in challenging open-shell molecular systems using Quantum Imaginary Time Evolution.