A friend of mine shared a link recently in LinkedIn of the news about QuEra’s 256-qubit quantum computer and this attracted my attention. It is true that nowadays we mostly follow the progress of superconducting qubits, trapped ions or photonic systems, so cold atoms is not something I was familiar about. But this technology made some promising advances, and reading a couple of research papers was enough to realize that cold atoms have some compelling advantages.

So, to start with, these computers use atoms instead of ions. Atoms are neutral as they don’t have positive or negative charge like ions, and this is why this technology is sometimes referred to as ‘neutral atoms’. They need to be cooled to temperatures close to absolute zero, and this is why they are sometimes referred to as ‘cold atoms’. These cold atoms are usually obtained from materials with one electron in their outer shell, like lithium (Li), potassium (K), rubidium (Rb) or cesium (Cs) called alkali metals, or with two electrons in the outer shell, like strontium (Sr) or beryllium (Be) called alkaline earth metals.

The first advantage of these atoms comparing to ions is that they can be isolated from the environment more easily. And all atoms are identical, which is not the case with superconducting qubits. Another difference with superconducting qubits is that the cooling is being done with vacuum and laser beams, so they don’t need dilution refrigerators.

Excitation of cold atoms brings them to so-called Rydberg states (this is why they are sometimes refered to as ‘Rydberg qubits’). These excited Rydberg states have very large radiuses that can be used to efficiently control neighboring atoms and can be exploited for multi-qubit entanglement (with demonstrations of GHZ states involving 20 qubits), multi-qubit Toffoli and CNOT / CZ gates with multiple control and multiple target qubits. For example, Toffoli gate can be implemented with 7 pulses, much less than on superconducting qubits which would need 30 pulses. This feature is unique and opens huge new possibilities to create some new algorithms and quantum computational approaches.

The coherence time of atoms in Rydberg state is 100μs, enough for depth of 40 gates. Initialization of the system requires to prepare and pump new atoms in the processor every time you run the circuit. But this works fast. The whole execution with preparation, qubit manipulations and readout takes 200ms, so this animation with Super-Mario provided by QuEra is realistic. What they do here is actually they pump the atoms to the cavities (called tweezers) of their quantum register in form of a grid, then excite some of them to the Rydberg states, move these atoms around with lasers and after each move, they take fluorescent images.

Rydberg qubits are scalable, and with current laser mechanisms to move atoms around, the number of qubits can grow to 1’000. For NIQS era, this is impressive. To go beyond these limits would require some new ideas.

Talking about the applications, Rydberg qubits have mostly been used for quantum simulations to simulate other physical or chemical systems (new molecules for example) by modeling their Hamiltonians and performing controlled experiments. But, their use is now going beyond quantum simulators. They can be used for quantum annealing for optimization use cases, and one can also run universal set of quantum gates to implement variational quantum algorithms (VQE or QAOA). And here they have unique advantages. For example, resolving combinatorial optimization problems would require the problem to be expressed in a quadratic unconstrained binary optimization model (QUBO), so that we can use quantum approximate optimization algorithm (QAOA) to simulate Hamiltonian that corresponds to this QUBO model. QAOA requires entangling every qubit to all others, but the limited interconnectivity of superconducting qubits involves a lot of swap instructions, which increases the depth of the circuit. With the current issues of qubit decoherence and noise, the size of the problem that we can resolve becomes very limited. The large connectivity of neutral atoms (Rydberg qubits) allows this kind of entanglement required by QAOA to be done more easily.

In addition to all this, there have been experiments of using Rydberg qubits as source of quantum memory (QRAM), where information encoded in photons is transferred to atoms in a form of polaritons. This memory can be used to create quantum repeaters for quantum communication networks. Another interesting practical application of cold atoms is when we pass two photons through them, and the interaction between polaritons will create non-linear interactions between these two photons. This can be used for photonic computers to build non-linear Gaussian boson samplers.

In addition to QuEra, French startup Pasqal has very advanced program based on Rydberg qubits. They got investment of €25M in 2021 and have already developed a set of use cases for optimization problems like MaxCut and Maximal Independent Set (MIS), and algorithms to implement them on their Pasqal processor including a small-scale implementation for electrical vehicle charging. And one can use Google CIRQ to code on their processor.

December 5, 2021