Two-Qubit Implementation of QAOA for MAX-CUT on an NV-Center Quantum Processor

This paper reports a proof-of-principle demonstration of the Quantum Approximate Optimization Algorithm (QAOA) for a two-qubit MAX-CUT problem on a room-temperature NV-center quantum processor, successfully reconstructing the cost landscape despite non-projective optical readout limitations.

The Gist

The Big Picture: Solving a Puzzle on a Tiny Diamond

Imagine you have a very difficult puzzle. In the world of computers, this is called the MAX-CUT problem. It's like trying to split a group of friends into two teams (Team A and Team B) so that the maximum number of friendships are broken between the teams.

Doing this with a regular computer is slow and hard, especially as the group gets bigger. This paper is about testing a new, super-fast way to solve this puzzle using a Quantum Computer.

But here's the twist: instead of using a giant, freezing-cold machine (like the ones from IBM or Google), these researchers used a tiny diamond that works at room temperature (just like your living room).

The Star of the Show: The "Diamond Atom"

The researchers didn't use a whole diamond; they used a single defect inside a diamond called an NV Center (Nitrogen-Vacancy center).

• The Analogy: Think of the NV center as a tiny, magical atom trapped inside the diamond. It has two "hands" it can use to do math:
  1. An Electron Spin (a fast, energetic hand).
  2. A Nitrogen Nucleus Spin (a slower, steadier hand).
• Together, these two hands act as two qubits (the basic units of quantum information). It's like having a tiny two-person team inside a diamond that can hold two pieces of information at once.

The Algorithm: QAOA (The "Guess and Check" Machine)

The team used an algorithm called QAOA (Quantum Approximate Optimization Algorithm).

• The Metaphor: Imagine you are trying to find the best route through a maze. A normal computer checks every path one by one. QAOA is like a hiker who starts at a random spot, looks around, takes a step toward a better spot, and repeats this process.
• The "Layers": The researchers only used one layer of this hiker's journey. It's a very simple version, but it's enough to prove the concept works.
• The Process:
  1. Mix: They put the two "hands" (qubits) into a superposition (a state where they are both Team A and Team B at the same time).
  2. Cut: They apply a specific "cut" operation to see which team split is better.
  3. Repeat: They tweak the settings (like turning a dial) to see if the result gets better.

The Challenge: Reading the Answer

Here is where it gets tricky. In a normal computer, you just look at the screen to see the answer (0 or 1). In this diamond quantum computer, you can't just "look" at the atom without messing it up.

• The Problem: When you shine a green laser on the diamond to read the answer, the atom glows red. But the brightness of the glow isn't a perfect "Yes" or "No." It's more like a dimmer switch. Sometimes a "Team A" glows a little, and sometimes "Team B" glows a little. You can't tell the difference in a single flash.
• The Solution: The researchers acted like statisticians. They ran the experiment 300,000 times.
  • They took the average brightness of the glow.
  • They used a clever math trick (called a "Hadamard transform") to reverse-engineer the data.
  • Analogy: Imagine trying to guess the color of a coin by flipping it 300,000 times and measuring the average light it reflects, rather than looking at one flip. By averaging out the noise, they could reconstruct the true probability of the answer.

The Results: A Rough Draft, But a Good One

The researchers mapped out the "Cost Landscape."

• What is that? Imagine a hilly terrain where the bottom of the valley is the perfect solution (the best team split). The researchers wanted to see if their quantum diamond could "feel" the shape of this terrain.
• The Outcome:
  • Success: The shape of the hills and valleys they measured looked very similar to the perfect, theoretical shape. They proved that the diamond can do the math.
  • The Noise: The measured hills weren't as smooth as the perfect ones. There was some "static" or fuzziness. This is because the diamond isn't perfect yet; the controls aren't 100% precise, and the environment causes some "jitter."
  • The Takeaway: Even with the fuzziness, the main features were clear. They could see where the best solution was.

Why Does This Matter?

This paper is a proof-of-principle. It's like the Wright Brothers' first flight. It wasn't a transatlantic journey, but it proved that "heavier-than-air flight" is possible.

1. Room Temperature: Most quantum computers need to be colder than outer space. This one works in a normal room. That's a huge deal for making quantum tech practical.
2. Scalability: They started with just two qubits. The paper suggests that by adding more "hands" (using other atoms in the diamond), they could solve bigger, more complex puzzles.
3. Future Steps: The next steps are to clean up the "noise" (make the controls more precise) and add more qubits to solve real-world problems like traffic routing or financial optimization.

Summary in One Sentence

The researchers successfully used a tiny defect in a diamond at room temperature to run a quantum optimization algorithm, proving that even with some noise and a simple setup, we can start solving complex puzzles using light and diamond atoms.

Technical Summary

1. Problem Statement

The paper addresses the challenge of implementing the Quantum Approximate Optimization Algorithm (QAOA) on Noisy Intermediate-Scale Quantum (NISQ) devices, specifically focusing on Nitrogen-Vacancy (NV) centers in diamond.

• Target Problem: The authors tackle the MAX-CUT problem, a standard NP-hard combinatorial optimization problem. They focus on the smallest non-trivial instance: a two-vertex graph with a single edge.
• Hardware Constraints: The experiment is conducted on a room-temperature NV-center processor. Key challenges include:
  • Non-projective Readout: Unlike superconducting or trapped-ion qubits, NV centers do not yield single-shot projective measurements in the computational basis. Instead, they provide fluorescence signals that overlap for different states.
  • Noise and Decoherence: The system operates at room temperature with limited coherence times and control noise.
  • Connectivity: The specific hardware uses the electron spin and the $^{14}$N nuclear spin of a single NV center as a two-qubit register.

2. Methodology

A. Algorithm Formulation

• Cost Function: The authors use a minimization formulation of MAX-CUT. The cost Hamiltonian is defined as $H_C = -\frac{1}{2} \sum_{i<j} A_{ij} (I - Z_i Z_j)$.
• QAOA Ansatz: A single-layer ($p=1$) ansatz is implemented:
  $$|\psi(\beta, \gamma)\rangle = U_B(\beta) U_C(\gamma) H^{\otimes 2} |00\rangle$$
  • Cost Unitary ($U_C$): Implemented as an $R_{ZZ}(\gamma)$ gate.
  • Mixing Unitary ($U_B$): Implemented as $RX(2\beta)^{\otimes 2}$.
• Hardware Mapping:
  • Qubits: The electron spin ($e^-$) and the $^{14}$N nuclear spin serve as the two qubits.
  • Gate Decomposition: The $R_{ZZ}$ gate is decomposed into native operations: $CNOT \cdot (I \otimes R_Z) \cdot CNOT$. The $CNOT$ is realized via conditional radio-frequency (RF) pulses on the nuclear spin controlled by the electron spin state.
  • Initialization: The system is initialized to $|00\rangle$ (electron $m_s=0$, nuclear $m_I=+1$) using optical pumping and hyperfine-mediated polarization transfer.

B. Measurement and Population Reconstruction

Since the optical readout is not projective, the authors developed a specific protocol to reconstruct computational basis populations:

1. Fluorescence Readout: The system measures the mean photon count ($\bar{n}$) over $N$ shots ($3 \times 10^5$ shots per parameter set).
2. Calibration: Calibration coefficients ($I_s$) are determined by preparing each of the four basis states ($|00\rangle, |01\rangle, |10\rangle, |11\rangle$) directly and measuring their mean fluorescence.
3. Linear Inversion: To reconstruct the populations of an arbitrary state $|\psi\rangle$, the authors measure four expectation values of the observable $O$ after applying specific bit-flip operations ($I, X_1, X_2, X_1X_2$) before readout.
   • The populations are recovered via a Walsh-Hadamard transform (linear inversion) of these four measurements.
   • This allows the calculation of the expected cost function $F(\beta, \gamma)$ without full quantum state tomography.

C. Experimental Setup

• Platform: A single NV center in diamond operated at room temperature ($27.5^\circ$C).
• Control: Microwave pulses control the electron spin; RF pulses control the nuclear spin.
• Scan: A grid search was performed over variational parameters $\gamma \in [0.1\pi, 2.1\pi]$ and $\beta \in [0.1\pi, 0.6\pi]$ to map the cost landscape.

3. Key Contributions

1. First Proof-of-Principle on NV Centers: This is the first demonstration of QAOA for combinatorial optimization on a room-temperature NV-center quantum processor.
2. Novel Readout Protocol: The paper establishes a robust method for reconstructing computational basis populations from non-projective fluorescence signals using linear inversion and bit-flip calibration. This bypasses the need for projective measurement, which is a major bottleneck for NV centers.
3. Hardware-Adapted Circuit: The successful mapping of the QAOA circuit to the native gate set of the NV center (electron-nuclear flip-flops and virtual Z-rotations).
4. Empirical Cost Landscape: The generation of an experimental cost landscape that, despite noise, retains the structural features necessary for variational optimization.

4. Results

• Landscape Reconstruction: The experimentally reconstructed cost landscape (Fig. 4b) successfully reproduces the qualitative structure of the ideal noise-free simulation (Fig. 4a).
  • It correctly identifies low-cost regions (optimal solutions) and high-cost regions.
  • The periodicity in $\gamma$ and $\beta$ matches theoretical predictions.
• Accuracy:
  • The average error between the experimental and ideal landscapes is 12.3%.
  • The reconstructed populations sum to $\approx 0.989$, indicating high fidelity in the reconstruction process.
• Error Sources:
  • Asymmetry and Contrast: The experimental landscape is less symmetric and has lower contrast than the simulation.
  • $\beta$ Sensitivity: Deviations are larger at higher $\beta$ values, attributed to the fact that the mixing layer ($RX$) requires driven rotations, which are more susceptible to calibration errors and decoherence.
  • $\gamma$ Stability: The dependence on $\gamma$ is more stable because the $Z$-rotation is implemented virtually (phase update), incurring no physical pulse time or decoherence.
• Convergence: The population reconstruction converges stable after approximately $10^5$ shots.

5. Significance and Future Outlook

• Feasibility: The study proves that the core elements of QAOA (state preparation, entangling gates, and variational optimization) are feasible on room-temperature solid-state platforms.
• Baseline for Improvement: The work establishes a baseline protocol for future improvements, specifically highlighting the need for:
  • Phase Tracking: To mitigate phase accumulation during long conditional gates.
  • Refined Calibration: To reduce the 12.3% error margin.
  • Scaling: The authors propose scaling to larger problem sizes by coupling to $^{13}$C nuclear spins or using circuit cutting techniques.
• Broader Impact: This research validates NV centers as a viable platform for small-scale quantum information processing and networked quantum architectures, offering a room-temperature alternative to cryogenic quantum processors.

In conclusion, this paper demonstrates that despite the limitations of non-projective readout and room-temperature noise, NV-center processors can successfully execute variational quantum algorithms, provided that appropriate reconstruction techniques are employed.