The researchers demonstrated directional photon emission, the first step toward measurable quantum interconnects

However, delivering on this promise requires flexible and resilient hardware. One of the challenges of building a large-scale quantum computer is that researchers must find an efficient way to connect quantum information nodes – separate smaller processing nodes on a computer chip. Because quantum computers are fundamentally different from classical computers, the conventional methods used to communicate electronic information do not translate directly to quantum devices. However, one requirement is certain: whether through a classical or quantum interconnect, the information carried must be sent and received.

To this end, MIT researchers have developed a quantum computing architecture that will enable scalable, high-fidelity communication between superconducting quantum processors. In a book published in Natural Physics, MIT researchers demonstrate the first step, the deterministic emission of single photons – carriers of information – in a direction specified by the user. Their method ensures that quantum information flows in the right direction more than 96% of the time.

Linking several of these modules allows for a larger network of quantum processors to be interconnected, regardless of their physical separation on a computer chip.

“Quantum interconnects are an important step toward larger, modular machine implementations built from smaller, individual components,” said Bharath Kannan PhD ’22, co-lead author of a research paper describes the technique.

“The ability to communicate between smaller subsystems will enable a modular architecture for quantum processors, and this may be an easier way to scale to larger system sizes compared to the brute force approach of using a single large and complex chip,” added Kannan.

Kannan wrote the paper with co-lead author Aziza Alanakly, a graduate student in electrical engineering and computer science in the Engineering Quantum Systems group at MIT’s Research Laboratory of Electronics (RLE). The lead author is William D. Oliver, Professor of Electrical Engineering, Computer Science, and Physics, Fellow of the MIT Lincoln Laboratory, Director of the Center for Quantum Engineering, and Associate Director of the RLE.

Removal of information altogether

In a typical classical computer, different components perform different functions, such as memory, calculation, etc. which transfers electrons to a computer processor.

But the amount of information is more complex. Instead of containing only a value of 0 or 1, quantum information can also be both 0 and 1 at the same time (a phenomenon known as superposition). Moreover, quantum information can be carried by particles of light, called photons. These additional complexities make quantum information fragile and cannot be transported using standard protocols alone.

A quantum network connects processing nodes using photons that travel through special interconnects called waveguides. The waveguide can be unidirectional, and move a photon to the left or right only, or it can be bidirectional.

Most existing architectures use unidirectional waveguides, which are easier to implement because the direction of photon travel can be easily established. But since each waveguide only transfers photons in one direction, more waveguides are required as the quantum network grows, making this technique difficult to scale. In addition, unidirectional waveguides typically incorporate additional components to enhance directivity, which introduces communication errors.

“We can eliminate these missing components if we have a waveguide that can support propagation in both left and right directions, and a way to choose the direction at will. This ‘directional transmission’ is what we demonstrated, and it the first step towards two-way communication with higher fidelity,” says Kannan.

Thanks to their architecture, many processing modules can be attached to a single waveguide. A unique feature of the architectural design is that the same module can be used as both a transmitter and a receiver, he said. And photons can be sent and received by any two modules in a common waveguide.

“We only have one physical connection that can have any number of modules along the way. That’s what makes it measurable. Having demonstrated the directional emission of a photon from a module, we now work on capturing the photon that that’s downstream in a second module,” Alanakly added.

Take advantage of quantum properties

To do this, the researchers built a module consisting of four qubits.

Qubits are the building blocks of quantum computers and are used to store and process quantum information. But qubits can also be used as photon emitters. Adding energy to a qubit causes the qubit to become excited, then when it de-excites, the qubit releases energy in the form of a photon.

However, simply connecting a qubit to a waveguide does not guarantee directivity. A single qubit emits a photon, but whether it moves left or right is completely random. To avoid this problem, the researchers use two qubits and a property known as quantum interference to ensure that the emitted photon travels in the right direction.

The method consists of preparing two qubits in a single entangled excitation state called the Bell state. This quantum mechanical state has two aspects: the left qubit is excited and the right qubit is excited. Both aspects exist simultaneously, but which qubit is excited at any given time is unknown.

When the qubits are in this entangled bell state, the photon is effectively emitted into the waveguide at both qubit locations simultaneously, and these two “emission pathways” interfere with each other. Depending on the relative phase in the Bell state, the resulting photon emission must shift left or right. By preparing the Bell state with the correct phase, the researchers select the direction in which the photon travels in the waveguide.

They can use the same technique, but in reverse, to receive the photon in another module.

“The photon has a certain frequency, a certain energy, and you can prepare a module to receive it by tuning it to the same frequency. If they are not at the same frequency, then the photon will just pass . It is analogous to tuning the radio to a particular station. If we select the right radio frequency, we will receive the music transmitted on that frequency,” explained Alanakly.

The researchers found that their method achieved more than 96% fidelity, meaning that if they intended to emit a photon to the right, 96% of the time it would go to the right.

Now that they have used this technique to efficiently emit photons in a specific direction, the researchers want to connect multiple modules and use the process to emit and absorb photons. This would be a big step towards building a modular architecture that combines many smaller processors into one larger and more powerful quantum processor.

The research was funded, in part, by the AWS Center for Quantum Computing, the US Army Research Office, the Department of Energy Office of Science National Quantum Information Science Research Centers, the Co-design Center for Quantum Advantage, and the Department of Defense. .