By Dr. Chris Mansell

Shown below are summaries of a few interesting research papers in quantum computing and communications that have been published over the past month.


Title: Multipartite entanglement of billions of motional atoms heralded by single photon
Organizations: Shanghai Jiao Tong University; University of Science and Technology of China; TuringQ Co., Ltd.
Over the course of a few years, experiments with atomic vapours have gone from entangling thousands of atoms to millions, and now, in this latest work, to billions. The vapour of atoms didn’t even need to be cooled below room-temperature and remained entangled for over a microsecond. While this may be long enough for certain applications, such as in quantum metrology, there are ideas to transfer the entanglement to noble-gas nuclear spins with lifetimes of up to several hours, which would be tremendous for the development of the quantum internet.

Title: Optomechanical quantum teleportation
Organizations: Delft University of Technology; University of Campinas; Zhejiang University
Tiny solid beams – about 10 micrometers long, consisting of about 10 billion silicon atoms – can interact with light, oscillate in their motional ground states and even become entangled. Incredibly, researchers have now quantum mechanically teleported the optical state of a single photon into a pair of these resonators. Scientifically, this work could allow for tests of fundamental physics by probing the boundary between quantum and classical physics. On a more practical level, the experiment was performed with light of the same wavelength that is employed in conventional telecommunications, showing that optomechanical quantum memories could serve as repeater nodes in a quantum network.

Title: Quantum Enhanced Precision Estimation of Transmission with Bright Squeezed Light
Organization: University of Bristol
When performing repeated measurements, two crucial quantities are the sensitivity, which is the smallest detectable value, and the precision, which is the inverse of the values’ variance. For optical measurements, both of these can be enhanced beyond the classical limit using amplitude squeezed light, where the fluctuations in the number of photons are at sub-Poissonian levels. Before this latest research, achieving such enhancements was only possible when restrictively low intensity light was employed. The current work demonstrates an 8 order of magnitude improvement in the power of the probe beam, from picowatts to almost a milliwatt, and allows a fundamentally quantum phenomenon, squeezed light, to compete with classical methods in a new regime. 

Title: Quantum computer-aided design: digital quantum simulation of quantum processors
Organizations: University of Toronto; Harvard University; Massachusetts Institute of Technology; Intel Labs; Vector Institute for Artificial Intelligence; Canadian Institute for Advanced Research
In the September Research Roundup, a paper was highlighted where the authors used a superconducting quantum computer to design an upgraded superconducting qubit. Hot on its heels, this article from another group of prominent researchers also focusses on using existing quantum computers to design candidates for the next generation of superconducting quantum processors. They argue that quantum algorithms for finding the energy levels of transmon qubits and for simulating the time-evolution of their gate operations have more scalable resource requirements than comparable classical approaches involving huge supercomputers. They point out that this fundamental idea is not restricted to devices composed of transmons but is applicable to other complex quantum computing architectures 


Title: Exponentially Many Local Minima in Quantum Neural Networks
Organization: University of Maryland
Classical neural networks with linear activation functions are easy to train because they have saddle points which are easy to escape from and local minima that are almost as good as the global minimum. Since quantum mechanical systems evolve linearly with time, it occurred to the authors of this paper that quantum neural networks (QNNs) could share these properties. Unfortunately, they discovered that quantum interference takes place in under-parameterised QNNs and leads to lots of bad local minima. This calls for future research into sufficiently parameterised QNNs as well as gradient-free training methods.

Title: Shadow process tomography of quantum channels
Organizations: NIST / University of Maryland; Massachusetts Institute of Technology; Harvard University; Lawrence Berkeley National Laboratory
Quantum process tomography is central to the verification, validation, performance assessment and debugging of quantum computers. Unfortunately, the number of measurements it requires scales exponentially with the number of qubits. A breakthrough in a related procedure, focused on states rather than processes, inspired the authors of this work to see if the concept of a classical shadow could improve the scaling of these measurement requirements. It turns out that by thinking in terms of succinct classical descriptions, the bounds on process tomography can be made similar to the most recent ones for state tomography. Overall, this research enhances our understanding of quantum processes and provides improved ways to study them experimentally.
Link for a recent, related paper:

Title: Absence of Barren Plateaus in Quantum Convolutional Neural Networks
Organizations: Los Alamos National Laboratory; University College London; Imperial College London
Before neural networks were all the rage, they were considered hard to train with the possible exception of the now well-established architecture, the convolutional neural network. Currently, the training of quantum neural networks is quite challenging but just as before, the interleaving of convolutional layers and pooling layers associated with a quantum convolutional neural network (QCNN) may provide a way forwards. In this paper, the authors consider the barren plateau phenomenon in which the optimisation landscape becomes exponentially more shallow as the problem size increases, greatly hindering the training. They rigorously prove that the gradients encountered in training a QCNN never become shallow more than polynomially quickly, making them one of the leading architectures for quantum machine learning.

Title: Free versus bound entanglement, a NP-hard problem tackled by machine learning
Organization: University of Vienna
Entanglement is an important resource for quantum information processing but even determining whether or not a given state is entangled is an NP-hard problem. Furthermore, there are at least two types: if the entanglement can be distilled using local operations and classical communication, it is called free and if it cannot, it is called bound. The former is considered more useful than the latter. In this work, the author uses machine learning to classify the entanglement present in a large family of quantum states, finding that bound entangled states occur more often than unentangled ones. Future work will involve applying these methods to systems with larger Hilbert spaces so that more can be discovered about how common the different types of entanglement are.

Title: Symmetry-resolved entanglement detection using partial transpose moments
Organizations: University of Innsbruck; The Abdus Salam International Center for Theoretical Physic; SISSA; Institute for Quantum Optics and Quantum Information of the Austrian Academy of Sciences; INFN; Université Grenoble Alpes; Johannes Kepler University Linz
The positive partial transpose test, also known as the Peres-Horodecki criterion, is a celebrated method to detect the presence of bipartite entanglement in quantum systems. Despite ongoing theoretical developments in the field of entanglement witnesses, experimental investigations are still extremely time-consuming. Fortunately, measuring informative quantities called the ‘partial transpose moments’ is more tractable for experimentalists. In this article, the authors develop the techniques for processing these numbers to reliably and confidently detect entanglement across a bipartition. Finding that their methods can be enhanced by exploiting symmetry, they go on to analyse the effects of experimental imperfections and suggest several exciting avenues for future exploration.
Link for a recent, related paper:

Title: Faster Coherent Quantum Algorithms for Phase, Energy, and Amplitude Estimation
Organization: The University of Texas at Austin
Phase, energy and amplitude estimation are important tasks for which there are already several quantum algorithms but these have their drawbacks. For example, they can be cumbersome, rely on assumptions that are not always met or require many copies of the input quantum state. In this work, block-encoding and singular value transformation techniques allow a new algorithm to avoid all these issues. It runs 14 times faster than the prior state-of-the-art for phase estimation and 20 times faster for energy estimation. This has positive implications for the protocols in which it could be used as a subroutine, such as quantum Metropolis sampling and quantum Bayesian inference. 

Title: Limitations of optimization algorithms on noisy quantum devices
Organization: University of Copenhagen
This research considers whether quantum computers that do not employ error correction have the potential to answer practical problems faster than classical computers. The main result, which applies to both quantum annealers and to variational quantum algorithms, is that classical optimisation will not admit a quantum speed-up without dramatic improvements in the noisiness of today’s quantum devices. Even then, the connectivity of the qubits will need to be tailored to the problem at hand. The work rigorously confirms the intuition that error-prone quantum circuits cannot have too many layers of logic gates before their output is no better than that of a classical solver. Finally, there is an interesting discussion in the supplementary information about what happens when the optimisation problem being tackled is quantum in nature.

Title: Quantum evolution kernel: Machine learning on graphs with programmable arrays of qubits
Organization: Pasqal, Sorbonne Universite
The study details how a Quantum Evolution Kernel (QEK) can serve as a more versatile and scalable procedure for building graph kernels and analyzing graph-structured data on quantum devices as compared to classical computers.  Benchmarking expected performances on a neutral-atom quantum computer, researchers found that QEK is stable against detection error and on par with state-of-the-art graph kernels on classical systems in terms of accuracy.

October 28, 2021