Quantum-centric simulation of hydrogen abstraction by sample-based quantum diagonalization and entanglement forging

This paper demonstrates a quantum-centric simulation of hydrogen abstraction in 2,2-diphenyldipropane using a hybrid approach that combines entanglement forging and sample-based quantum diagonalization on an IBM Heron processor to accurately compute reaction energies across active spaces up to (39e,39o).

The Gist

The Big Picture: Solving a Puzzle with Two Minds

Imagine you are trying to solve a massive, incredibly complex 3D puzzle. The puzzle represents how molecules behave when they react with each other. Specifically, this paper looks at a reaction where a tiny, aggressive "thief" (a free radical) steals a hydrogen atom from a larger molecule. This theft is the first step in a chain reaction that causes airplane parts made of composite materials to rot and peel over time when exposed to sunlight.

Solving this puzzle perfectly requires a supercomputer, but the puzzle is so big that even the world's best classical computers struggle to get the answer right without making mistakes.

The authors propose a new way to solve this: Quantum-Centric Supercomputing. Think of this not as a single machine, but as a team-up between a human mathematician (a classical computer) and a psychic (a quantum computer).

• The Classical Computer is the project manager. It handles the heavy lifting, organizes the data, and checks the math.
• The Quantum Computer is the psychic. It can "feel" the quantum nature of the electrons in a way classical computers can't, but it gets tired easily (it makes noise/errors) and can only hold a small amount of information at once.

The Problem: The "Room" is Too Small

In quantum computing, information is stored in "qubits." To simulate a molecule, you usually need one qubit for every possible way an electron can spin. This is like trying to fit a whole library into a single shoebox. For the large molecules the authors wanted to study, the "shoebox" (the quantum processor) was too small. They didn't have enough qubits to hold the whole picture.

The Solution: "Entanglement Forging" (The Magic Split)

To fix the room size issue, the team used a technique called Entanglement Forging (EF).

The Analogy: Imagine you need to describe a complex dance routine involving 100 dancers, but your camera only has enough memory to record 50 dancers at a time.
Instead of giving up, you split the dance into two groups of 50. You record Group A, then you record Group B. Because the two groups are "entangled" (they are dancing in sync with each other), you can mathematically "forge" the two separate recordings back together to reconstruct the full 100-dancer routine.

In the paper, this allowed them to simulate a molecule using half the number of qubits they normally would have needed. They mapped the problem onto a smaller "shoebox" by splitting the electron pairs and reassembling the results later.

The Method: "Sample-Based Quantum Diagonalization" (SQD)

Even with the smaller room, the quantum computer is noisy. It's like trying to take a clear photo in a dark, shaking room. You might get a blurry picture, or a picture of the wrong thing.

To handle this, they used a method called Sample-Based Quantum Diagonalization (SQD).

The Analogy: Imagine you are trying to find the deepest point in a foggy valley (the lowest energy state of the molecule). You can't see the whole valley at once.

1. Sampling: The quantum computer takes thousands of "snapshots" (samples) of the valley, giving you random points.
2. Classical Processing: The classical computer takes all these snapshots and builds a map. It looks for patterns and calculates the most likely location of the deepest point.
3. Iterating: If the map looks wrong, the quantum computer takes more specific snapshots based on what the classical computer learned, and the process repeats until the map is accurate.

The paper claims this method allows them to correct for the "noise" and "blur" of the quantum computer, effectively cleaning up the data to find the true answer.

The Experiment: Testing the New Tools

The team tested this combined approach (EF + SQD) on a specific chemical reaction: Hydrogen Abstraction.

• The Target: They simulated a simplified version of an epoxy resin (the glue used in airplane wings) reacting with a methyl radical.
• The Scale: They tested this on three different sizes of "active spaces" (different levels of detail):
  1. Small (13 electrons): A quick test.
  2. Medium (23 electrons): A moderate challenge.
  3. Large (39 electrons): A massive challenge that would usually break a standard quantum simulation.

The Results: What They Found

1. Success on Large Scales: For the largest simulation (39 electrons), their new method worked. They were able to calculate the energy of the reaction with high accuracy.
2. The Old Way Failed: When they tried to use the "old" standard method (without Entanglement Forging) on that same large simulation, the quantum computer was too noisy. The data was so corrupted that the classical computer couldn't make sense of it. The "shoebox" was too full, and the "blur" was too strong.
3. Accuracy: Their results matched up very well with the best classical supercomputer simulations available (called DMRG and CCSD(T)), proving that their "team-up" approach is reliable.

The Conclusion

The paper demonstrates that by combining a "splitting" trick (Entanglement Forging) with a "sampling and cleaning" strategy (SQD), scientists can now simulate much larger and more complex chemical reactions on current quantum hardware than was previously possible.

They didn't just simulate a reaction; they proved that this specific combination of tools can handle the "noise" of today's quantum computers to solve problems that are too big for the hardware alone. This is a step toward understanding how airplane materials degrade, which could eventually help engineers design better, longer-lasting materials.

Technical Summary

1. Problem Statement

The paper addresses the challenge of simulating complex chemical reactions relevant to aerospace engineering, specifically the photo-degradation of composite materials (e.g., epoxy resins used in aircraft fuselages).

• Context: Composite materials degrade due to radical chain reactions initiated by UV/visible light, leading to hydrogen abstraction, chain scission, and loss of mechanical integrity.
• Scientific Gap: While experimental accelerated aging tests exist, ab initio computational models for these degradation pathways are lacking. Current computational methods struggle to balance accuracy and cost for large active spaces required to model radical reactions accurately.
• Computational Bottleneck: Standard quantum simulations typically require $2M$ qubits to represent $M$ spatial orbitals (one qubit per spin-orbital). This limits the size of systems that can be simulated on current noisy intermediate-scale quantum (NISQ) devices.

2. Methodology

The authors propose a Quantum-Centric Supercomputing (QCSC) approach, combining classical High-Performance Computing (HPC) with a superconducting quantum processor (IBM Heron family). The core methodology integrates two specific techniques:

A. Entanglement Forging (EF)

• Concept: A qubit-reduction technique that maps a qubit to a spatial orbital rather than a spin-orbital. This reduces the required qubit count by half (from $2M$ to $M$).
• Wavefunction Representation: Instead of a standard $2M$-qubit circuit, EF represents the wavefunction as a sum of tensor products of two $M$-qubit states:
  $$|\Psi\rangle = \sum_\mu c_\mu |u_\mu\rangle \otimes |v_\mu\rangle$$
  where $|u_\mu\rangle$ and $|v_\mu\rangle$ represent the $\alpha$ (spin-up) and $\beta$ (spin-down) sectors, respectively.
• Generalization: The authors move beyond previous EF implementations that used simple "bitstring states" and "hopgates." They introduce a more general ansatz based on Unitary Coupled Cluster Doubles (uCCD) and Resonating Hartree-Fock (ResHF).
  • The states $|u_\mu\rangle$ and $|v_\mu\rangle$ are constructed using Local Unitary Cluster Jastrow (LUCJ) circuits and non-orthogonal Slater determinants.
  • This allows for a more accurate representation of electron correlation while maintaining the qubit reduction.

B. Sample-Based Quantum Diagonalization (SQD)

• Concept: A quantum subspace method where a quantum device samples electronic configurations (bitstrings) from a probability distribution $p(x) = |\langle x|\Psi\rangle|^2$.
• Classical Post-Processing: A classical computer solves the Schrödinger equation within the subspace spanned by these sampled configurations:
  $$\sum_n H_{mn} c_n = E c_m$$
• Configuration Recovery: To handle noise and symmetry breaking (e.g., incorrect particle numbers), SQD employs an iterative self-consistent procedure. It "recovers" configurations by flipping bits to match target spin-resolved particle numbers and updates the occupation number distribution based on the lowest energy found in batches.

C. Integration of EF and SQD

• The authors derive a method to combine EF with SQD. Since the joint probability distribution of an EF wavefunction is not a simple compound distribution, they approximate it using a weighted sum of compound distributions.
• They sample from this approximate distribution on the quantum device and feed the resulting bitstrings into the classical SQD solver.

3. Key Contributions

1. Methodological Innovation: The first successful combination of Entanglement Forging with Sample-Based Quantum Diagonalization. This allows for the simulation of larger active spaces (up to 39 electrons in 39 orbitals) on current hardware by halving the qubit requirement.
2. Generalized EF Ansatz: The introduction of a more flexible family of quantum circuits for EF, moving from simple bitstring/hopgate combinations to uCCD-based LUCJ circuits and non-orthogonal Slater determinants. This improves accuracy and better captures electron correlation.
3. Hybrid Workflow: A robust QCSC workflow that uses classical HPC to correct for decoherence-induced symmetry breaking and solve large sparse eigenvalue problems, while offloading the sampling of complex electronic configurations to the quantum processor.
4. Resource Efficiency: Demonstrated that EF circuits require significantly fewer two-qubit gates and lower circuit depth compared to standard LUCJ circuits (which would require double the qubits), making them more resilient to hardware noise.

4. Results

The method was applied to simulate hydrogen abstraction from 2,2-diphenylpropane (a model for DGEBA epoxy resin) by a methyl radical ($\text{CH}_3^\bullet$).

• Active Spaces Tested: Simulations were performed across three active spaces:
  • (13e, 13o)
  • (23e, 23o)
  • (39e, 39o)
• Performance Metrics:
  • (13e, 13o) & (23e, 23o): The SQD+EF results agreed with high-level classical benchmarks (CCSD(T), DMRG) within 1 mEh (milli-Hartree) for ground state energies. The extrapolated energies (zero variance) were highly accurate.
  • (39e, 39o): This is the largest active space simulated to date using this approach.
    • Standard LUCJ ansatz failed on hardware due to excessive noise (the Hartree-Fock bitstring was not recovered).
    • SQD+EF succeeded, providing results within ~2 mEh of DMRG benchmarks after energy-variance extrapolation.
• Reaction Energies:
  • The calculated activation energy ($E_{TS} - E_R$) and reaction energy ($E_P - E_R$) using SQD+EF were consistent with DMRG and CCSD(T) within 1–2 kcal/mol across all active space sizes.
  • The method successfully captured the energetics of the transition state and product formation, which are critical for understanding degradation mechanisms.

5. Significance

• Scalability: This work demonstrates that qubit-reduction techniques like Entanglement Forging are essential for scaling quantum chemistry simulations on near-term hardware. It effectively doubles the accessible system size for a given number of qubits.
• Practical Application: It provides a pathway to ab initio modeling of polymer degradation, a problem previously intractable for high-accuracy quantum methods. This could lead to the rational design of more durable aerospace materials.
• QCSC Paradigm: The paper validates the QCSC model, showing that classical supercomputers can effectively mitigate quantum noise and solve the linear algebra problems, while quantum processors handle the specific task of sampling complex, high-dimensional probability distributions that are hard for classical computers.
• Future Outlook: The authors suggest this framework can be extended to systems with multireference character and combined with circuit knitting techniques to further scale up simulations.

In summary, the paper presents a breakthrough in hybrid quantum-classical simulation, successfully using entanglement forging to simulate a chemically complex radical reaction on a 39-qubit active space, achieving accuracy comparable to state-of-the-art classical methods.