In this blog post — the first in a new series! — we share our favourite papers released in the first quarter of 2024. The selection is based on relevance to quantum algorithms and applications; these are results that we admire and that have been influential to our research. Xanadu papers won’t appear due to an obvious conflict of interest, but we take the opportunity to share our latest work at the end.
Contents
- The Top 5
- 1. Programmable Simulations of Molecules and Materials with Reconfigurable Quantum Processors
- 2. Quantum Circuit Optimization with AlphaTensor
- 3. Computational supremacy in quantum simulation
- 4. Quantum-centric Supercomputing for Materials Science: A Perspective on Challenges and Future Directions
- 5. Language models for quantum simulation
- Honourable mentions
- Xanadu papers from winter 2024
The Top 5
1. Programmable Simulations of Molecules and Materials with Reconfigurable Quantum Processors
- This is exemplary work showing how methods from quantum chemistry, materials science, quantum algorithms and quantum hardware can be expertly combined to advance applications of quantum computing.
- The paper is not just an application of known techniques; they developed better quantum algorithms based on clever innovations like employing methods from dynamical decoupling to project onto valid symmetry subspaces, and performing time evolution over "batches" of spin subsystems.
- We found this work persuasive in its argument that it is possible to use effective models to learn properties of the spin ladder in realistic materials.
2. Quantum Circuit Optimization with AlphaTensor
- We were impressed by the claim that AlphaTensor-Quantum “can save hundreds of hours of research by optimizing relevant quantum circuits in a fully automated way.”
- For example, their model can reportedly reproduce compilation performance for tensor hypercontraction algorithms for quantum chemistry. 🤯
- There are caveats however; the Hadamard gadgetization results in extra qubits, and to assess its actual usefulness in practice it should be tested independently by the community.
3. Computational supremacy in quantum simulation
- Any work claiming quantum advantage is likely to be controversial, but we found this to be a serious and remarkable paper presenting compelling evidence that the quantum annealer is capable of qualitatively correct simulation of quenched spin systems with hundreds of qubits.
- They perform simulations with 1000 measurements per second for a variety of topologies; for example, we noted the use of a biclique graph, which was also studied in a recent Xanadu paper.
- It contains a thorough comparison to powerful classical simulation techniques, likely performed by academic (and thus more independent) co-authors. Their conclusion that DMRG outperforms PEPS and NQS was especially interesting to us.
4. Quantum-centric Supercomputing for Materials Science: A Perspective on Challenges and Future Directions
-
This review it contains deeply valuable information, especially Section V on applications, although it is not always easy to find.
-
The paper is a collection of perspectives by a conglomerate of top scientists working at the interface of quantum computing, materials science, and high-performance supercomputing. It is a very useful resource that we likely will keep referring back to.
-
Here are some key quotes where the authors propose application areas (slightly paraphrased for readability, with emphases added):
-
While our classical toolkit for simulating ground-state physics is rather mature, the study of excitations within linear response is much more demanding and often requires high-performance computing resources. The study of time-dependent phenomena in real materials is still in its infancy. These science domains therefore offer promising application fields for future quantum computers.
-
There are two ways of making feasible progress toward the solution of challenging instances of the electronic structure problem: quantum embedding theories and effective low-energy models.
-
One concrete example of a fermionic lattice model that can be connected to a technological application is the three-band Hubbard model in two dimensions, also known as the Emery model. This model is generally believed to contain the essential properties of high-Tc cuprate superconductors.
-
Polyyne molecules, characterized by alternating carbon single and triple bonds, are challenging to stabilize. They belong to high symmetry point groups and feature numerous silent modes, undetectable through infrared or Raman spectroscopy. A precise depiction of their low-frequency vibrations is crucial for understanding their reactivity.
-
5. Language models for quantum simulation
-
A paper that stands out by signaling the emergence of a new field — watch out for the co-appearance of “transformer” and “quantum” in the title of many new papers. 😉
-
We found this paper to be eloquently written, clear in explaining technical concepts, and comprehensive in summarizing prior and future work. We found the presented argument persuasive, that it is important to study how large language models can impact quantum computing.
-
Some key quotes (emphases added):
-
So far, models the size of today’s commercial LLMs have not been trained on quantum computing data. It is intriguing to ask whether, with sufficient scaling, LLMs could learn accurate representations of quantum computing hardware well into the future.
-
Will LLMs be able to learn digital copies of quantum computers, given sufficient training data?
-
Can LLMs display emergence of macroscopic quantum phenomena such as superconductivity if scaled sufficiently large?
-
While it may be hard to predict how the collision of quantum computers and LLMs will play out, what is clear is that the fundamental transformation enabled by the interplay of these technologies has already begun.
-
Honourable mentions
Fast emulation of fermionic circuits with matrix product states
Strong work describing an upgrade to a fermionic quantum emulator to allow MPS representations of the underlying states. This sacrifices accuracy but results in massive gains in performance.Quantum eigenvalue processing
Perhaps the most important new quantum algorithm on this list — a tour de force by superstars of the field, extending techniques of quantum singular value transformations to work directly with eigenvalues and with non-normal matrices.Reducing the runtime of fault-tolerant quantum simulations in chemistry through symmetry-compressed double factorization
This work uses a “shifting technique” that displaces eigenvalues of a Hamiltonian to reduce its one-norm, without changing the cost of implementing the final quantum algorithm. The result is arguably the new state-of-the-art, but unfortunately it only improves on previous methods by less than a factor of 5.On the need for effective tools for debugging quantum programs
Probably the best-written paper on this list. It masterfully presents a call to action to develop better debugging methods for quantum computing. Our favourite quote from the paper:One can argue that the author should just “be better at quantum programming”, or “not make so many mistakes”. But, everyone makes mistakes, and we all benefit from better debugging tools.
We hope you enjoyed this selection of top papers. Stay tuned for the Spring 2024 edition! You can sign up to the Xanadu newsletter or follow PennyLane on LinkedIn or Twitter/X to get notified.
Xanadu papers from winter 2024
Here we share the list of our publications from this winter, but you can find the full list of publications on our website.
Weight Reduced Stabilizer Codes with Lower Overhead
A method to reduce the weight of parity checks in stabilizer codes. This is important because such checks are also noisy, so reducing them can lead to better error-correction performance, as demonstrated in this work.
Better bounds for low-energy product formulas
Product formulas like Trotter–Suzuki decompositions are ubiquitous in quantum algorithms. This work considerably tightens error bounds for applications tailored to low-energy states.
Hybrid quantum programming with PennyLane Lightning on HPC platforms
Large-scale simulations using PennyLane Lightning on high-performance supercomputers. This paper establishes Lightning’s often superior performance compared to other software platforms. See also our blog post for an overview.
Hardware-efficient ansatz without barren plateaus in any depth
Building on prior work by some of the authors, this work argues that it is possible to avoid issues of barren plateaus for variational circuits aimed at many-body physics simulations.
Better than classical? The subtle art of benchmarking quantum machine learning models
A sobering and influential paper introducing a framework for benchmarking in quantum machine learning. It also presents an accompanying software library and introduces 160 datasets, made available through PennyLane Datasets: Hidden manifolds, Hyperplanes, Linearly separable, and Two curves. See also the accompanying blog post.
Quantum simulation of time-dependent Hamiltonians via commutator-free quasi-Magnus operators
This paper introduces a new quantum simulation technique for time-dependent Hamiltonians, which is arguably the new state-of-the-art.
About the author
Juan Miguel Arrazola
Making quantum computers useful