Excited-State Quantum Chemistry on Qumode-Based Processors via Variational Quantum Deflation

This paper introduces the qumode-based variational quantum deflation (QumVQD) framework, which leverages bosonic quantum processors and symmetry-constrained Fock basis filtering to efficiently compute electronic and vibrational excited states with high accuracy and significantly reduced circuit depth compared to qubit-based algorithms.

The Gist

Imagine you are trying to solve a massive, incredibly complex jigsaw puzzle. This puzzle represents a molecule, and your goal is to figure out exactly how its pieces (electrons and atoms) fit together to create different shapes (energy states).

For decades, scientists have tried to solve these puzzles using classical computers (like your laptop). But as the molecules get bigger, the number of possible ways the pieces can fit together explodes exponentially. It's like trying to find a single specific grain of sand on every beach on Earth simultaneously. Classical computers simply run out of time and memory.

Then came Quantum Computers, which promised to solve these puzzles by using the weird rules of quantum mechanics. However, most current quantum computers use "qubits." Think of a qubit like a light switch: it's either ON (1) or OFF (0). To represent a complex molecule with a qubit-based computer, you have to build a giant wall of light switches. This is inefficient because the "switches" are fragile, prone to breaking (noise), and you need thousands of them just to describe a simple molecule.

The New Approach: The "Qumode" Piano

This paper introduces a new way to do quantum chemistry using a different kind of quantum hardware called Qumodes.

Instead of a light switch, imagine a piano key. A qubit can only be "up" or "down." But a piano key (a qumode) can be pressed gently, pressed hard, or pressed anywhere in between. It can vibrate at infinite levels of intensity.

• The Analogy: If a qubit is a binary code (0 or 1), a qumode is a continuous dial that can spin to any number.
• The Benefit: Because a single qumode can hold so much more information than a single qubit, you need far fewer of them to describe a molecule. It's like describing a painting with just a few broad, expressive brushstrokes (qumodes) instead of millions of tiny, rigid pixels (qubits).

The Problem: Finding the "Excited" States

Most quantum chemistry methods are great at finding the molecule's "ground state"—its calmest, most relaxed shape. But chemists often need to know about excited states—what happens when the molecule gets a jolt of energy (like absorbing sunlight).

Finding these excited states is like trying to find the second-lowest, third-lowest, or fourth-lowest valley in a mountain range without accidentally falling back into the deepest valley (the ground state) you already found.

The Solution: "Variational Quantum Deflation" (QumVQD)

The authors created a new algorithm called QumVQD. Here is how it works, using a simple metaphor:

The "No-Go" Zone Strategy:
Imagine you are hiking and you want to find the second-lowest valley.

1. You first find the lowest valley (the ground state).
2. You then put up a giant, invisible "No-Go" fence around that lowest valley.
3. You tell your hiking robot: "Find the lowest point, but you are forbidden from entering the fenced area."
4. The robot is forced to find the next lowest valley.
5. You repeat this, adding more fences around every valley you've already found, until you've mapped out all the important ones.

In the paper, this "fence" is a mathematical rule that forces the computer to ignore solutions it has already found, ensuring it discovers new, unique energy states.

Two Major Tricks to Make it Faster

The authors didn't just use this "fence" strategy; they added two clever tricks to make the computer run much faster and more accurately:

1. The "Guest List" Filter (Particle Number Conservation)
In a molecule, the number of electrons never changes. If you start with 4 electrons, you must end with 4.

• The Old Way: The computer wastes time checking millions of impossible scenarios where the electron count is wrong (e.g., 3 electrons or 5 electrons).
• The New Trick: The authors built a "bouncer" into the algorithm. It checks the "guest list" (the binary code of the state) and immediately kicks out any scenario where the electron count is wrong.
• The Result: This shrinks the search space massively. It's like a library that only lets you look at books about cats, ignoring the millions of books about dogs, cars, and space. This makes the calculation exponentially faster.

2. The "Lego Block" Strategy (Hamiltonian Fragmentation)
For vibrating molecules (like how atoms wiggle), the math is incredibly hard.

• The Old Way: Trying to solve the vibration of the whole molecule as one giant, tangled knot.
• The New Trick: They break the molecule's vibration into smaller, manageable "Lego blocks" (fragments). They solve each small block individually using the unique strengths of the qumode (which is naturally good at handling vibrations) and then snap the answers together.
• The Result: This reduced the number of complex "gates" (steps the computer must take) by 10 to 100 times compared to qubit computers.

Why This Matters: The "Noise" Factor

Quantum computers are currently "noisy." Imagine trying to listen to a whisper in a hurricane. The errors (noise) can ruin the answer.

• Qubit computers are very sensitive. Because they need so many steps (gates) to solve these problems, the noise accumulates quickly, and the answer gets garbled.
• Qumode computers (in this study) need far fewer steps. It's like listening to that same whisper, but you only have to listen for a split second before the hurricane passes. The answer stays clear.

The Bottom Line

The authors tested this new method on simple molecules like Hydrogen ($H_2$), Carbon Dioxide ($CO_2$), and Hydrogen Sulfide ($H_2S$).

• For Electrons: They found the energy levels with "chemical accuracy" (perfect enough for real-world chemistry).
• For Vibrations: They found the energy levels with "spectroscopic accuracy" (precise enough to identify the molecule just by its sound/vibration).

In summary: This paper shows that using "piano keys" (qumodes) instead of "light switches" (qubits), combined with smart filtering and breaking problems into smaller pieces, allows us to solve complex chemical puzzles much faster and with fewer errors. It suggests that the future of quantum chemistry might not be built on the qubits we see today, but on these more powerful, continuous-wave machines.

Technical Summary

1. Problem Statement

Quantum chemistry faces significant challenges in simulating many-body wavefunctions, particularly for excited states and strongly correlated systems.

• Limitations of Qubit-Based Architectures: Current qubit-based quantum computers suffer from limited qubit counts, short coherence times, and high error rates. Simulating molecular vibrations on qubits requires costly "boson-to-qubit" mappings (e.g., truncating Fock spaces), which introduce substantial overhead and circuit depth.
• Need for Excited States: While ground-state algorithms (like VQE) are well-developed, calculating excited-state energies (crucial for spectroscopy and photochemistry) requires specialized extensions like Variational Quantum Deflation (VQD).
• Hardware Gap: There is a lack of established frameworks for running excited-state calculations on bosonic quantum processors (qumode-based devices), which naturally align with molecular structures and vibrational dynamics.

2. Methodology: QumVQD Framework

The authors introduce QumVQD (Qumode-based Variational Quantum Deflation), a framework designed to compute both electronic and vibrational excited-state energies on bosonic hardware.

A. Core Algorithm: Variational Quantum Deflation (VQD)

• The method extends the Variational Quantum Eigensolver (VQE) by adding a penalty term to the cost function.
• Orthogonality Constraint: To find the $k$-th excited state, the algorithm minimizes the energy while enforcing orthogonality to all previously computed $k-1$ eigenstates. This is achieved by classically computing state overlaps and adding weighted penalty terms ($\beta_n$) to the energy expectation value.

B. Electronic Structure: Particle Number Enforcement

• Fock Basis Encoding: The system uses a Fock basis where computational states are labeled by a composite index.
• Hamming Weight Filtering: The authors enforce particle number conservation by restricting the Hilbert space to states with a specific Hamming weight (the count of '1's in the binary representation of the Fock index).
• Dimensionality Reduction:
  • Unrestricted Hilbert space dimension: $O(2^M)$ for $M$ spin orbitals.
  • Restricted dimension (fixed electron count $n_e$): $O(\binom{M}{n_e})$.
  • Impact: This reduces the required number of qumodes logarithmically and eliminates spurious non-physical eigenstates, significantly lowering computational overhead.

C. Vibrational Structure: Hamiltonian Fragmentation

• Bogoliubov Transforms: Instead of mapping vibrational modes to qubits, the authors utilize the natural bosonic representation. They decompose anharmonic vibrational Hamiltonians into solvable fragments using Cartan subalgebra approaches.
• Decomposition: The Hamiltonian $H$ is split into fragments $H_k = U_k D_k U_k^\dagger$, where $D_k$ are diagonal and solvable.
• Gate Implementation: The unitary $U_k$ is implemented using native bosonic gates: Displacement ($D$), Squeezing ($S$), and Beam Splitter ($BS$) gates. This avoids the overhead of boson-to-qubit mapping.

3. Key Contributions

1. QumVQD Framework: The first implementation of Variational Quantum Deflation specifically tailored for qumode-based processors, applicable to both electronic and vibrational excited states.
2. Symmetry Enforcement: Introduction of Hamming-weight-based filtering in the Fock basis to enforce particle number conservation, reducing the effective Hilbert space dimension from exponential to combinatorial scaling.
3. Vibrational Fragmentation: Integration of Hamiltonian fragmentation (via Bogoliubov transforms) with VQD, allowing for the calculation of vibrational eigenenergies with minimal circuit depth.
4. Noise Analysis: A systematic comparison of noise resilience between qubit and qumode circuits, utilizing both gate fidelity models and Kraus operator-based amplitude damping models.

4. Results

The framework was validated through simulations on small molecules ($H_2$, $H_4$, $LiH$, $CO_2$, $H_2S$).

• Electronic Structure ($H_2$):

  • Achieved agreement with Full Configuration Interaction (FCI) benchmarks within chemical accuracy ($< 1$ kcal/mol or $\approx 1.6$ mEh) across potential energy surfaces.
  • The method successfully reproduced excited-state energies with errors well below $10^{-3}$ Hartree.
  • Particle number enforcement reduced the Hilbert space dimension significantly (e.g., a compression ratio $R \approx 573$ for LiH/6-31G).

• Vibrational Structure ($CO_2$, $H_2S$):

  • Achieved spectroscopic accuracy ($< 1$ cm$^{-1}$) for the lowest vibrational eigenstates.
  • Gate Count Reduction: The qumode approach required 1–2 orders of magnitude fewer entangling gates compared to analogous qubit-based algorithms (e.g., UVCC or CHC).
    • Example: Solving $CO_2$ required only ~26 Beam Splitter gates, whereas qubit-based methods required thousands of CX gates.

• Noise Resilience:

  • Analysis showed that the reduced circuit depth of qumode algorithms translates to significantly lower error accumulation.
  • Under amplitude damping models, the qumode approach demonstrated enhanced resilience, provided photon loss rates remain on the order of $10^{-4}$ or lower per gate to maintain chemical accuracy.

5. Significance and Future Outlook

• Paradigm Shift: The work demonstrates that bosonic quantum devices are not just alternatives but potentially superior platforms for specific quantum chemistry problems, particularly those involving vibrational dynamics and excited states.
• Scalability: The combination of symmetry enforcement and Hamiltonian fragmentation suggests that bosonic devices could tackle systems (like $C_3H_8$) that are currently intractable for classical FCI and require prohibitive resources on qubit-based machines.
• Hardware Suitability: By leveraging the native infinite-dimensional Hilbert space of qumodes, the method avoids the "mapping overhead" inherent in qubit approaches, offering a more direct and efficient path to chemical accuracy.
• Future Work: The authors note that while current hardware cannot yet support the required number of qumodes for large molecules (e.g., 11 qumodes for $C_3H_8$), the theoretical resource requirements are within reach of near-future hardware improvements. Further work is needed to map reduced Hamiltonians to specific bosonic gates and optimize circuit complexity for larger systems.

In summary, QumVQD establishes a robust, noise-resilient pathway for solving excited-state quantum chemistry problems on bosonic hardware, offering a distinct advantage over qubit-based methods in terms of circuit depth, gate count, and natural alignment with molecular physics.