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: Demonstration of fault-tolerant universal quantum gate operations
Organizations: Institut für Experimentalphysik, Universität Innsbruck, Institute for Quantum Information, RWTH Aachen University, Institute for Theoretical Nanoelectronics (PGI-2), Forschungszentrum Jülich, Alpine Quantum Technologies GmbH, Innsbruck, Institut fur Quantenoptik und Quanteninformation,
Österreichische Akademie der Wissenschaften

Quantum computers can be protected from noise by encoding the logical quantum information redundantly into multiple qubits using error correcting codes. When manipulating the logical quantum states, it is imperative that errors caused by imperfect operations do not spread uncontrollably through the quantum register. This requires that all operations on the quantum register obey a fault-tolerant circuit design which, in general, increases the complexity of the implementation. Here, the authors demonstrate a fault-tolerant universal set of gates on two logical qubits in a trapped-ion quantum computer. In particular, they make use of the recently introduced paradigm of flag fault tolerance, where the absence or presence of dangerous errors is heralded by usage of few ancillary ‘flag’ qubits. They perform a logical two-qubit CNOT-gate between two instances of the seven qubit color code, and we also fault-tolerantly prepare a logical magic state. They then realize a fault-tolerant logical T-gate by injecting the magic state via teleportation from one logical qubit onto the other. They observe the hallmark feature of fault tolerance, a superior performance compared to a non-fault-tolerant implementation. In combination with recently demonstrated repeated quantum error correction cycles these results open the door to error-corrected universal quantum computation.

Title: Superconducting-like Heat Current: Effective Cancellation of Current-Dissipation Trade-Off by Quantum Coherence
Organizations: The University of Electro-Communications; Japan Science and Technology Agency; RIKEN
A heat engine absorbs heat from a hot part of its environment, known as the hot bath, performs mechanical work, releases heat to the cold bath and returns to its initial state, ready to repeat the entire process. Operated in reverse, heat engines act as refrigerators. They are everywhere but there is a fundamental trade-off between their power and efficiency. Previous theoretical research into quantum heat engines has shown that they can surpass this classical trade-off provided certain, somewhat challenging, conditions are met. The newly published result shows that coherence between quantum states with the same energy is all that is required to achieve high efficiency and high power at the same time. This is achievable for experimentalists, who have already demonstrated prototype quantum refrigerators, and improves the prospects of developing coolers for nanoscale devices.

Title: Machine learning of high dimensional data on a noisy quantum processor
Organizations: University of Waterloo; Fermi National Accelerator Laboratory; Google Quantum AI; Sandbox@Alphabet
Experimental work on kernel-based quantum machine learning has mostly been performed with a very small number of qubits processing low dimensional data. In this article, 17 of the qubits in Google’s Sycamore quantum processor were remotely accessed and used to classify data related to astrophysics. Specifically, each data point consisted of 67 parameters associated with measurements of light from either a type II or a type Ia supernova. Despite moderate circuit fidelity, the quantum classifier achieved a level of accuracy comparable to that of a competitive classical approach. This shows that a quantum kernel can be efficiently implemented on hardware, perform well in high dimensions and be robust to noise. Interestingly, the attempts that were made to reduce readout errors did not consistently improve this noise robustness.

Title: Adiabatic quantum linear regression
Organization: Oak Ridge National Laboratory
Ordinary computers and adiabatic quantum computers can both solve optimisation problems but which is faster? The authors of this paper considered two specific implementations and found that D-Wave’s adiabatic quantum computer could perform linear regression on 16 million data points up to 2.8 times quicker than a classical Intel i9 processor. This was not a demonstration of quantum advantage because for that, the comparison needs to be against a state of the art supercomputer. Instead, the experiment showed that embedding problems into quantum hardware could be done efficiently, that D-Wave’s devices are improving and that despite not having as many decades of research and development as classical computers, quantum computers can have comparable runtimes.

Title: Programmable and sequential Gaussian gates in a loop-based single-mode photonic quantum processor
Organization: The University of Tokyo
The wave-particle duality of light means that different approaches to optical quantum information processing can be taken. The discrete number of photons in a beam of light could take the primary role or the continuous phases and amplitudes of the electromagnetic wave could take centre stage. The final alternative that is explored in this paper is to get the best of both worlds: Gaussian logic gates that rely on the continuous variables of waves are deterministic and when the waves are very short, discrete pulses, each one can act as a qubit. A set-up where time-domain multiplexing allows the qubit pulses to go around a loop of optical fibre until they are needed has previously been used for entangled state generation. In the latest upgrade, the experimenters can choose which operation to perform based on the result of an earlier measurement, which gives them the ability to perform universal, scalable and programmable quantum computation.

Title: Cross-Cross Resonance Gate
Organizations: IBM; The University of Tokyo
Qubits decohere due to interactions with their environment even when they are not being acted upon by any quantum logic gates. The goal of a logic gate is therefore to perform its operation without further increasing this decoherence. A gate that achieves this is said to reach the coherence limit. In this paper, the researchers propose and demonstrate an extension of the cross resonance gate for superconducting qubits. The use of microwave pulses with the same intensity, but at two frequencies instead of one, allows their gate to be faster and for the chance of the quantum state leaking out of the computational basis to be halved. Overall, the fidelity of the new gate is close to the coherence limit and could be improved further by adding more microwave frequencies.

Title: Using graphene conductors to enhance the functionality of atom chips
Organizations: University of Nottingham; Durham University; University of Sussex; Humboldt-Universität zu Berlin; University of Applied Sciences Wiener Neustadt
By coherently manipulating a uniform, homogenous ensemble of cold atoms, it is possible to keep track of time with incredibly high accuracy and measure magnetic fields with high sensitivity, spatial resolution and field of view. Trapping the atoms closer to current-carrying surfaces and for longer periods of time would greatly improve these quantum sensing and timing technologies. In this theoretical paper, detailed calculations of the current noise and the atom-surface attraction show that replacing metal wires with graphene would have both these effects. In addition to improving one of the leading quantum metrology platforms, it could also enable hybrid quantum information processing devices with atomic and solid-state elements to be designed.


Title: Generalization in Quantum Machine Learning: A Quantum Information Standpoint
Organizations: University of Florence; INFN Sezione di Firenze; University of York
Information theory was a great triumph of the twentieth century that originated at Bell Labs and still has innumerable applications today. Quantum information theory has aided the progress of many quantum technologies and in this paper, it is used to analyse how data, presented in the form of quantum states, can be classified into different categories. The authors find that the ability to correctly classify new, previously unseen data points is improved by discarding the irrelevant aspects of the data. They explore the idea of an information bottleneck that only lets through the information that is essential for choosing the right category. This theory, backed up by their numerical experiments, is a useful guide for further design iterations of related quantum algorithms.

Title: Scalable Mitigation of Measurement Errors on Quantum Computers
Organization: IBM
In a quantum computation, errors occur when preparing quantum states, implementing logic gates and performing measurements. One way to mitigate the effect of the measurement errors in particular is to construct a matrix that describes them, use a classical computer to invert the matrix and see what the output would be without them. Unfortunately, the size of the matrix is exponential in the number of qubits, which means this approach isn’t scalable. In this latest article, the authors show that measurement error mitigation can be performed by explicitly constructing a much smaller version of this matrix. This tractable solution could be employed in numerous settings and will be especially important as the qubit count of modern experiments increases.

Title: Qubit readout error mitigation with bit-flip averaging
Organization: Imperial College London
Continuing with the same topic as the previous Research Roundup paper, namely the mitigation of measurement errors, this work shows several benefits of a remarkably simple procedure. For each qubit, decide with 50% probability whether or not to perform a bit-flip Pauli X gate on it immediately before it gets measured. This averages out the biases in the measurements, dramatically reduces the amount of calibration that needs to take place and makes the matrix describing the errors highly symmetric and easier to invert. The procedure works even when the errors are correlated and can be straightforwardly combined with other error mitigation schemes. 

November 26, 2021