Subspace Selected Variational Quantum Configuration Interaction with a Partial Walsh Series

The paper proposes a new Variational Quantum Eigensolver (VQE) ansatz that uses subspace superposition and diagonal Walsh operators to efficiently estimate ground-state energies for both full and selected configuration interaction (CI) wavefunctions across various molecular systems.

The Gist

Imagine you are trying to find the exact recipe for the perfect sourdough bread. The "recipe" is the ground-state energy of a molecule—the most stable, lowest-energy way its electrons can arrange themselves.

The problem is that for a complex molecule, there are trillions of possible ways to arrange those electrons. Trying to check every single one is like trying to taste every single grain of flour in a giant warehouse to find the perfect one. It’s impossible for even the fastest computers.

This paper introduces a new "smart tasting" method called Subspace Selected Variational Quantum Configuration Interaction with a Partial Walsh Series. Here is how it works, broken down into three simple steps.

1. The "VIP List" (Subspace Selection)

Instead of wandering aimlessly through the entire warehouse of flour, the researchers use a trick to create a VIP List.

In chemistry, we know that most electron arrangements are useless—they are high-energy and unstable. Instead of looking at everything, the algorithm identifies a "subspace"—a small, elite group of electron arrangements (called Slater Determinants) that are actually physically relevant. It’s like saying, "Don't look at all the grains in the warehouse; only look at the ones in these ten specific premium bags." This makes the problem much smaller and manageable.

2. The "Musical Remix" (The Walsh Ansatz)

Now that we have our VIP list, we need to figure out exactly how much of each "ingredient" to use to get the perfect recipe. Usually, in quantum computing, this part is messy and can lead to a problem called "Barren Plateaus"—which is like trying to find the bottom of a valley while standing on a perfectly flat, infinite plain. You just wander around forever without knowing if you're getting closer to the bottom.

To avoid this, the researchers use something called a Walsh Series.

Think of the electron recipe as a complex song. Instead of trying to adjust every single tiny note one by one (which is exhausting), they use "Walsh operators," which act like master volume sliders for different frequencies of the song. By adjusting these broad "frequency sliders," they can shape the entire wavefunction much more efficiently. It’s like adjusting the bass, the treble, and the mid-tones on a stereo rather than trying to tweak the vibration of every individual speaker wire.

3. The "Smart Shortcut" (Partial Walsh Series)

Even with the sliders, if you have too many sliders, you get overwhelmed (this is called "overparameterization"). To fix this, they use a Partial Walsh Series.

They don't use every possible slider available. Instead, they randomly pick a clever subset of sliders that are mathematically guaranteed to cover all the important parts of the song. It’s like having a soundboard with 1,000 knobs, but realizing you can get 99% of the same sound using just 50 specific, well-chosen ones. This keeps the quantum computer from getting "confused" and keeps the math fast.

Why does this matter?

The researchers tested this on real quantum hardware (an IBM processor) and simulators using molecules like Hydrogen ($H_2$) and Water ($H_2O$).

The result? They were able to find the "perfect recipe" (the ground-state energy) with incredible accuracy, often hitting "chemical accuracy"—the gold standard in chemistry.

In short: They found a way to stop searching the whole universe for an answer and instead created a "VIP list" of candidates and a "set of master sliders" to fine-tune the result. This makes quantum computers much more practical for solving real-world chemistry problems today, even before we have "perfect," error-free quantum computers.

Technical Summary

Technical Summary: Subspace Selected Variational Quantum Configuration Interaction with a Partial Walsh Series

Problem Statement

Estimating the ground-state energy of many-body electronic systems is a primary goal for quantum computing. While the Variational Quantum Eigensolver (VQE) is a leading candidate for Noisy Intermediate-Scale Quantum (NISQ) devices, standard VQE Ansätze often suffer from two major issues:

1. Overparameterization and Barren Plateaus: Large, unstructured parameter spaces can lead to flat cost landscapes where optimization becomes impossible.
2. Scaling Limits: Full Configuration Interaction (FCI) provides exact solutions but scales combinatorially, making it classically intractable for large systems. Conversely, Selected Configuration Interaction (SCI) reduces the space but can still lead to costly classical diagonalizations or errors in Hamiltonian construction.

Methodology

The authors propose a new VQE Ansatz that combines subspace selection with a diagonal Walsh-series-based wavefunction encoding. The algorithm consists of two primary stages:

1. Subspace Superposition Preparation ($U_S$):
   Instead of preparing a state across the entire $2^r$ Hilbert space, the algorithm prepares a uniform superposition only over a selected set of $D$ physically relevant Slater Determinants (SDs).

   • For number-conserving spaces, they use Dicke state preparation ($O(rN)$ CNOT scaling).
   • For more complex subspaces (preserving spin and particle number), they utilize a quantum walk algorithm ($O(rD)$ CNOT scaling).

2. Diagonal Walsh Ansatz Application:
   To encode the CI coefficients ($c_k$) into the superposition, the authors use a diagonal operator $U$. Because this operator is generally non-unitary, they employ an ancilla qubit and a dilated diagonal unitary construction.

   • The operator is implemented using Walsh operators ($\hat{w}_j$), which are diagonal in the computational basis.
   • Partial Walsh Series: To avoid the complexity of a full Walsh-Fourier Transform (WFT), they use a "partial" series. They oversample the Walsh basis by selecting $O(D \log D)$ functions randomly. This ensures the transformation remains full-rank with high probability while keeping the number of variational parameters and gates manageable.
   • Implementation: The circuit uses CNOT and $R_z$ gates, with gate counts scaling linearly with the number of determinants ($O(D)$).

Key Contributions

• Avoidance of Barren Plateaus: Because the Ansatz is diagonal and commutative, it does not form a 2-design, which helps mitigate the emergence of barren plateaus during optimization.
• Controlled Parameterization: The number of variational parameters is directly tied to the number of selected determinants ($D$), preventing the overparameterization common in hardware-efficient Ansätze.
• Hybrid Scalability: The algorithm offers a middle ground between exact FCI and approximate SCI. It allows for "systematically improvable" results—by increasing the number of selected SDs, the accuracy increases.
• Efficient Classical Preprocessing: By using random oversampling or QR decomposition, the classical overhead is kept to $O(D^2 \log D)$ or $O(D^3)$, respectively.

Results

The authors benchmarked the algorithm using quantum simulators and IBM’s Torino (Heron-based) quantum hardware:

• Molecular Dissociation: For $H_2$ and $H_6$, the algorithm successfully reproduced dissociation curves. For $H_2$, the results on the noisy Torino device and simulator were in excellent agreement with the exact FCI solution, often surpassing chemical accuracy ($1.6 \times 10^{-3}$ Hartree).
• Complex Molecules: For $H_2O$ in a 6-31G basis, the algorithm demonstrated that increasing the SCI subspace (via a threshold $\epsilon$) systematically improved the fidelity and energy accuracy compared to the exact solution.
• Scaling Analysis: They confirmed that for large systems, the required oversampling rate for the Walsh basis actually decreases, following the theoretical $O(1 - D^{-1})$ scaling.

Significance

This work provides a robust framework for performing high-accuracy quantum chemistry on NISQ-era hardware. By mapping the CI wavefunction directly to a diagonal Walsh series, the authors have created an Ansatz that is physically motivated, resource-efficient, and optimization-friendly. It bridges the gap between classical SCI methods and quantum eigenvalue estimation, offering a scalable path toward simulating strongly correlated electronic systems that are currently beyond the reach of classical computers.