Improvement of performance of Grover's algorithm on three generations of Heron family IBM QPUs without and with topological dynamical decoupling

This paper investigates the performance of Grover's algorithm across three generations of IBM Heron QPUs, demonstrating improved success probabilities compared to previous hardware and showing that topological dynamical decoupling can enhance results, even when using a suboptimal number of iterations.

The Gist

The Quantum Needle in a Haystack: A Story of Better Magnets and Faster Searchers

Imagine you are standing in a massive, dark warehouse filled with millions of identical-looking cardboard boxes. Somewhere in one of those boxes is a single gold coin.

If you were a human (a classical computer), you would have to open the boxes one by one. If there are a million boxes, you might be there for weeks. But if you had a magical "Quantum Searcher" (Grover’s Algorithm), you wouldn't look at them one by one. Instead, you would use a special wave-like trick to make the "gold coin" signal grow louder and louder until it’s the only thing you can hear, allowing you to find it much, much faster.

However, there is a problem: The Warehouse is Shaky.

In the real world, quantum computers aren't perfect. The "warehouse" (the quantum processor) is constantly vibrating, the lights are flickering, and the floor is uneven. These vibrations—which scientists call noise and decoherence—scramble the magical signal of the gold coin. By the time you think you’ve found it, the signal has become so distorted that you’re just grabbing random boxes.

This paper is a report on how researchers tried to fix those vibrations to make the search more successful.

---

1. The Upgraded Warehouse (The IBM Heron QPUs)

The researchers tested three different generations of a specific type of quantum "warehouse" called the IBM Heron family.

Think of this like upgrading from an old, rickety wooden shed to a modern, high-tech laboratory.

• Generation 1 (Torino): An old shed with shaky floors.
• Generation 2 (Marrakesh): A sturdier building.
• Generation 3 (Pittsburgh): A state-of-the-art facility with much more stable floors and better lighting.

The researchers found that as the hardware got better (the "Pittsburgh" generation), the search became significantly more accurate. Even without any extra help, the newest machines were much better at finding the "coin" than the old ones.

2. The "Noise-Canceling Headphones" (Dynamical Decoupling)

Even in a high-tech lab, there is still some noise. To fight this, the researchers used a technique called Dynamical Decoupling (DD).

Imagine you are trying to listen to a faint melody in a noisy room. If you can't stop the noise, you might try to use noise-canceling headphones. In quantum computing, "Dynamical Decoupling" is like hitting the "pause" and "play" buttons on the qubits very rapidly in a specific pattern. This rapid pulsing "averages out" the noise, effectively canceling out the vibrations so the quantum signal can stay clear.

The researchers tested several "patterns" of noise-canceling:

• CPMG and XY4: These are the "old reliable" patterns—standard ways of pulsing the qubits.
• Topological Dynamical Decoupling ($T_n$): This is a new, fancy, mathematical pattern designed to be even more robust against different types of noise.

3. The Results: Finding the Needle

The researchers pushed the machines to their limits, searching for "coins" in increasingly larger haystacks (from 3-qubit problems to 6-qubit problems).

• For small haystacks (3-5 qubits): The new machines were superstars. They found the answer much more reliably than previous generations of computers.
• For the big haystack (6 qubits): This is where it gets tricky. A 6-qubit search is much more complex and requires many more "steps" (iterations). In a perfect world, you’d take a specific number of steps to find the coin. But in the real world, if you take too many steps, the "shakiness" of the machine builds up and ruins the search.
• The Breakthrough: On the best machine (Pittsburgh), the researchers found that if they used the "noise-canceling" patterns ($T_4$ or $XY4$), they could actually find the correct answer in a 6-qubit search, even though it was incredibly difficult. They were able to distinguish the "gold coin" from a random guess, which is a huge win for a machine of this size.

Summary in a Nutshell

The researchers proved that better hardware (the Heron chips) combined with smarter noise-canceling tricks (Topological Dynamical Decoupling) allows us to perform complex quantum searches that were previously impossible due to "noise." We are moving from a world where the quantum signal is lost in the static to a world where we can actually hear the music.

Technical Summary

Technical Summary: Improvement of Grover’s Algorithm Performance on IBM Heron QPUs

1. Problem Statement

The primary objective of this research is to evaluate and enhance the performance of Grover’s algorithm—a quantum algorithm providing a quadratic speedup for searching unsorted databases—on contemporary superconducting quantum processing units (QPUs).

As quantum hardware scales, the algorithm faces significant challenges from decoherence (loss of quantum information due to environmental interaction) and gate errors. Specifically, the researchers aim to determine if the latest generation of IBM’s Heron family QPUs can outperform previous generations and whether Topological Dynamical Decoupling (TDD) can effectively mitigate noise to allow for larger problem sizes (up to six qubits) and more iterations.

2. Methodology

The study utilizes a comparative approach across three generations of IBM Heron QPUs: Torino (r1), Marrakesh (r2), and Pittsburgh (r3).

• Algorithm Implementation: The researchers implemented Grover’s algorithm using IBM Qiskit, targeting problem sizes of 3, 4, 5, and 6 qubits. They focused on "balanced" bitstrings (equal numbers of 0s and 1s) to account for potential decay in excited states.
• Error Mitigation Strategy: The study focuses on Dynamical Decoupling (DD)—a technique that applies sequences of pulses to idle qubits to suppress decoherence. They compared three types of sequences:
  1. CPMG: A standard, well-established sequence.
  2. XY4: A robust sequence utilizing rotations around different axes.
  3. $T_n$ (Topological Dynamical Decoupling): A recently proposed family of sequences characterized by specific phase patterns.
• Experimental Design:
  • Runs were executed with 10,000 shots per experiment.
  • The researchers carefully selected execution times based on calibration data (T1, T2, readout error, and 2Q gate error) to ensure results reflected the best possible hardware performance.
  • For the 6-qubit case, they tested varying numbers of Grover iterations to find the practical limit of the hardware.

3. Key Contributions

• Benchmarking Heron QPUs: The paper provides a systematic performance comparison across three generations of the Heron architecture, demonstrating a clear upward trend in hardware fidelity.
• Validation of TDD: The study provides empirical evidence for the effectiveness of Topological Dynamical Decoupling ($T_n$) in practical quantum algorithms, specifically showing that certain $T_n$ sequences (like $T_2$ or $T_8$) can outperform classical sequences like CPMG.
• Scaling Analysis: The research pushes the boundaries of current NISQ (Noisy Intermediate-Scale Quantum) capabilities by successfully demonstrating a 6-qubit Grover search that distinguishes the target state from a random guess.

4. Key Results

• Hardware Improvement: Success probabilities for 3, 4, and 5 qubits on the Heron family (without error correction) are higher than those reported for previous generations of superconducting QPUs. The Pittsburgh (r3) QPU showed significantly higher success rates than its predecessors.
• Iteration Trade-offs: Theoretically, the optimal number of Grover iterations is $\approx \frac{\pi}{4\theta}$. However, the study found that practical optimal iterations are lower than theoretical ones. This is due to the accumulation of 2Q gate errors and decoherence; as iterations increase, the noise eventually outweighs the algorithmic gain.
• DD Performance:
  • For 5 qubits, $T_n$ sequences showed an "oscillatory behavior" regarding the number of pulses. Smaller pulse counts (e.g., $T_2, T_4$) generally provided better enhancement than longer sequences ($T_{10}, T_{12}$).
  • On the Pittsburgh (r3) QPU, the $T_2$ sequence improved the 5-qubit success probability to approximately 0.4.
• 6-Qubit Breakthrough: While the theoretical optimal number of iterations (6) resulted in success probabilities near the "random guess" threshold ($1/2^6$), using $T_4$ or XY4 DD sequences with a reduced number of iterations (e.g., 2 or 3) allowed the algorithm to clearly identify the target bitstring with a probability significantly higher than random chance.

5. Significance

This work is significant because it demonstrates that algorithmic success on near-term quantum hardware is a delicate balance between algorithmic depth and hardware noise. By combining the improved physical properties of newer QPU generations (higher $T_1/T_2$ times and lower gate errors) with advanced error suppression techniques like Topological Dynamical Decoupling, the researchers successfully extended the reach of Grover's algorithm to larger bitstrings. This provides a roadmap for maximizing the utility of NISQ-era devices before full-scale fault-tolerant quantum computing is realized.