Symmetry-adapted qubit encoding with complete active space and Bravyi--Kitaev mapping for quantum chemistry on a quantum computer

This paper introduces a symmetry-adapted qubit encoding with complete active space (SAE-CAS) that integrates approximate Z-symmetries and Bravyi-Kitaev mapping to significantly reduce qubit counts and circuit complexity for quantum chemistry simulations, demonstrating superior convergence and resource efficiency over standard methods on both near-term and fault-tolerant quantum processors.

The Gist

Imagine you are trying to solve a massive, incredibly complex jigsaw puzzle to understand how a molecule works. In the world of quantum chemistry, this puzzle is the "electronic structure" of a molecule. To solve it on a quantum computer, we usually need to assign a tiny piece of the computer (a "qubit") to every single possible spot where an electron could be.

The problem? For even small molecules, this requires thousands of puzzle pieces (qubits), and the instructions to assemble them (the circuit) become so long and tangled that current computers can't handle them, and even future ones will struggle.

This paper introduces a clever new way to solve the puzzle called SAE-CAS. Here is how it works, using simple analogies:

1. The "Freeze and Ignore" Strategy (The CAS Part)

Think of a molecule like a busy office building.

• The Frozen Core: The basement and the top floor are always full of people who never leave and never interact with the rest of the building. In quantum terms, these are "frozen-core" electrons. They are boring and predictable.
• The Virtuals: The attic is completely empty and will likely stay empty. These are "virtual" orbitals.
• The Active Space: The middle floors are where the real action happens. People are moving around, talking, and changing things. This is the "active space."

Traditional methods try to assign a qubit to every floor, even the boring basement and the empty attic. SAE-CAS says: "Let's just ignore the basement and the attic." We only assign qubits to the middle floors where the interesting chemistry happens. This immediately shrinks the size of the puzzle we need to solve.

2. The "Symmetry Shortcut" (The SAE Part)

Even within the busy middle floors, there are rules. For example, in a water molecule, the left side is a mirror image of the right side. If you know what's happening on the left, you automatically know what's happening on the right.

Usually, computers calculate both sides separately, wasting time and resources. SAE-CAS uses a mathematical "magic trick" (called an affine Clifford transformation) to realize that because of these mirror rules, we don't need separate qubits for both sides. We can mathematically "fold" the puzzle in half. This removes even more qubits, making the puzzle even smaller and easier to solve.

3. The "Translation" (The Bravyi–Kitaev Part)

Once we have our tiny, folded puzzle, we need to translate it into a language the quantum computer understands. There are two main translators:

• Jordan-Wigner (JW): The standard translator. It's simple but makes the instructions very long (like reading a book where every word is repeated).
• Bravyi–Kitaev (BK): A smarter translator. It organizes the information more efficiently, so the instructions are shorter and less tangled.

The authors show that you can use their "Folded Puzzle" method (SAE-CAS) with either translator. They created a version called SAE-CAS-BK that uses the smarter translator. It doesn't change the final answer, but it often makes the path to get there smoother and faster.

What Did They Find?

The authors tested this method on nine small molecules (like water, oxygen, and nitrogen) using two different "search strategies" (algorithms) to find the molecule's energy:

1. UCCSD: A chemically precise but complex search.
2. HE-SCA: A simpler, hardware-friendly search.

The Results:

• Fewer Qubits: By ignoring the boring parts and folding the symmetrical parts, they needed significantly fewer qubits (sometimes cutting the number in half or more).
• Shorter Circuits: The instructions to run the simulation were much shorter and less tangled.
• Faster Success: When using the simpler search strategy (HE-SCA), their method found the correct answer for every molecule tested. The old method (JW-CAS) got stuck and failed to find the answer for oxygen and carbon monoxide within the time limits.
• No Loss of Accuracy: Even though they ignored the "boring" electrons and folded the puzzle, the final energy numbers were just as accurate as the standard, massive calculations.

The Bottom Line

The authors have built a "resource-efficient" toolkit. They proved that you can safely throw away the parts of the molecule that don't change and fold the parts that are symmetrical, without losing the correct answer. This makes it possible to run these complex chemical simulations on quantum computers that are much smaller and less powerful than previously thought necessary.

They have also made the code for this "magic trick" available for free (in a package called QuantumSymmetry) so others can use it to simulate molecules on their own quantum computers.

Technical Summary

Technical Summary: Symmetry-Adapted Qubit Encoding with Complete Active Space and Bravyi–Kitaev Mapping

Problem Statement
Quantum chemistry simulations on near-term (NISQ) and fault-tolerant quantum processors face significant resource constraints, primarily driven by the number of qubits required to encode molecular electronic structures and the complexity (Pauli weight and circuit depth) of the resulting Hamiltonians. While the Complete Active Space (CAS) approximation reduces the problem size by freezing core orbitals and discarding virtual orbitals, standard implementations often fail to fully exploit symmetries to minimize qubit counts further. Additionally, different fermion-to-qubit mappings, such as Jordan–Wigner (JW) and Bravyi–Kitaev (BK), offer trade-offs between locality and qubit requirements that are not always optimally integrated with symmetry reduction techniques.

Methodology
The authors propose a unified framework called Symmetry-Adapted Encoding with Complete Active Space (SAE-CAS). This approach treats the CAS approximation not merely as a truncation of the Hilbert space, but as the imposition of approximate $Z$-symmetries on frozen-core and virtual orbitals.

1. Theoretical Framework:

   • Approximate Symmetries: The method extends exact symmetry-adapted encodings (SAEs) to include approximate symmetries corresponding to fixed orbital occupancies (frozen-core and virtual). These are modeled as $Z$-symmetries where specific qubits are constrained to eigenvalues of $-1$ (frozen) or $+1$ (virtual).
   • Affine Clifford Transformations: The reduction is achieved via affine Clifford maps. The authors define a binary orbital-character matrix $S$ representing the symmetries. By constructing an invertible affine map $T$ over $\mathbb{F}_2$, they transform the computational basis such that the symmetry constraints correspond to fixed values (typically 0) on specific qubits.
   • Qubit Removal: Once the Hamiltonian is conjugated by the Clifford operator, terms containing $X$ or $Y$ on the constrained qubits are discarded, and the constrained $Z$ operators are replaced by scalars. This permanently removes the redundant qubits, reducing the system from $n$ spin-orbitals to $n-k$ qubits, where $k$ is the number of independent symmetries.
   • Equivalence Proof: The authors prove that the resulting active-space qubit Hamiltonian is exactly equivalent to the canonical CAS Hamiltonian derived via frozen-core and virtual-orbital projection in molecular electronic structure theory. The approximation error arises solely from the choice of the active space, not the encoding.

2. Integration with Bravyi–Kitaev (SAE-CAS-BK):

   • The framework is extended to the Bravyi–Kitaev (BK) mapping. Since the BK transformation is itself an affine Clifford basis change acting on the computational bitstrings, it can be composed with the SAE-CAS Clifford map.
   • The resulting SAE-CAS-BK encoding is unitarily equivalent to SAE-CAS. It preserves the qubit count, the number of Pauli terms, and the variational parameters but alters the locality structure (Pauli weight per term) and the resulting circuit depth/CNOT counts.

3. Numerical Benchmarking:

   • The method was tested on nine small molecules (including H$_2$O, C$_2$H$_4$, O$_2$, N$_2$, and C$_4$H$_6$) using a minimal STO-3G basis.
   • Two variational ansätze were employed: Unitary Coupled Cluster with Single and Double excitations (UCCSD) and a hardware-efficient shifted-circular-alternating (HE-SCA) ansatz.
   • Comparisons were made against standard JW, JW-CAS (CAS with JW but no symmetry reduction), and JW-CAS with symmetry filtering (SF) at the circuit level.

Key Results

• Resource Reduction: SAE-CAS consistently reduces the number of qubits, Pauli operator count, circuit depth, CNOT count, and variational parameters compared to JW-CAS. For example, in the C$_4$H$_4$ (4,4) CAS, SAE-CAS reduced the qubit count from 8 (JW-CAS) to 4 and the Pauli count from 97 to 46.
• Convergence and Accuracy:
  • UCCSD: SAE-CAS reached converged energies comparable to JW-CAS and JW-CAS (SF) but required significantly fewer optimization iterations (e.g., 21 vs. 109 for N$_2$). Notably, for C$_4$H$_4$ and C$_4$H$_6$, the JW-CAS and JW-CAS (SF) UCCSD ansätze failed to run due to memory exhaustion from the large number of excitations, whereas SAE-CAS succeeded by filtering the excitation list at the mapping stage.
  • HE-SCA: SAE-CAS demonstrated superior convergence properties. For all benchmarked molecules, SAE-CAS converged to CAS reference energies within the tested layer budgets. In contrast, JW-CAS failed to converge for O$_2$ and CO within 200 layers. The authors attribute this to the "barren plateau" effect and sector leakage in JW-CAS, where the variational manifold spans incorrect spin-parity sectors, whereas SAE-CAS restricts the search to the correct symmetry sector natively.
• SAE-CAS vs. SAE-CAS-BK: The two encodings produced identical converged energies and iteration counts, confirming their unitary equivalence. Differences were observed only in circuit depth and CNOT counts, which varied by less than 6% for most molecules (except CO, which saw a +20% increase for BK). The authors note that the asymptotic $O(\log n)$ locality advantage of BK is not yet realized at these small active space sizes.

Significance and Claims
The paper claims that SAE-CAS offers a systematic and composable pathway to resource-efficient molecular simulations. By treating the CAS approximation as a set of approximate symmetries, the method unifies exact symmetry reduction (point-group and spin-parity) with active-space truncation. The primary significance lies in:

1. Native Symmetry Enforcement: Enforcing symmetries at the encoding level (rather than post-hoc in the ansatz) removes redundant search directions and prevents the optimizer from drifting into incorrect particle-number or spin sectors.
2. Scalability: The method enables simulations on larger active spaces that would otherwise be intractable due to memory constraints in constructing the ansatz (as seen in the C$_4$H$_4$ and C$_4$H$_6$ cases).
3. Flexibility: The framework is compatible with any fermion-to-qubit mapping expressible as an affine Clifford basis change, demonstrated here with both JW and BK.

The authors provide an open-source implementation in the Python package QuantumSymmetry and conclude that SAE-CAS is a viable route for reducing qubit counts and operator complexity on both near-term and fault-tolerant quantum processors.