A few days before last Christmas, EuroHPC JU announced new efforts to expand the list of quantum computers in the European ecosystem. The call will be open until the 31st of March 2024.
In March 2022, EuroHPC JU published the first call to host quantum computers in HPC centers. As a result, almost half a year later, EuroHPC issued a press release announcing the hosted agreements signed for the installation of the new EuroHPC quantum computers. Each of the six HPC centers will host a quantum computer with a different physical implementation of qubits: superconducting qubits in a star-shaped topology (IT4), superconducting qubits in a square-lattice topology (LRZ), photonic quantum-dot (TGCC), neutral atoms (CINECA), trapped ions (PSNC) and superconducting analog quantum computer (BSC). A description of these implementations is beyond the scope of this article, but it may be the topic of a future one.
If you are interested in the call or curious what it’s about, but you are clueless when it comes to “quantum language”, read on.
In summary, the call for expression of interest provides €20M from EuroHPC JU to procure at least two quantum computers. Those quantum computers will be co-funded with the EuroHPC JU participating states. EuroHPC JU will cover up to 50% of the total cost, considered as up to 50% of the acquisition costs, up to 50% of the costs to integrate the quantum computer with the existing supercomputer of the hosting entity, and up to 50% of the operating costs of these quantum computers. However, the quantum computer will be 100% owned by EuroHPC JU. The most relevant requirements are that the hosting entity should have experience working with quantum technologies or installing and operating integrated HPC systems, and ensure connectivity among the other hosting facilities inside the EuroHPC ecosystem.
As you may know from the previous call (summarized at the beginning of this article), due to the nature and maturity of quantum computing hardware, a wide assortment of physical implementations of qubits exist. The call follows the same path as its older sibling: the hosting entity is in charge of choosing the quantum hardware (without singling out any one vendor) and how it will be implemented with the HPC system. However, due to the novelty and complexity behind quantum computing, choosing the best option could be a hard decision that can be simplified after contacting experts, quantum computing vendors, and/or consulting companies.
Anyhow, despite contacting external people, the minds behind the proposal should be aware of the basic technicalities, such as the specifications to be fulfilled by the installation date.
The quantum chip should have:
- At least 10 physical qubits: Physical qubits are the physical implementation of the two-state quantum system. In other words, it is the number of “pieces” that can stay in the state defined as 0 or 1, such as the energy levels of an atom, polarization of a photon, presence or absence of electrons in a vacancy or in certain positions, electron spin, among others.
- 2-qubit gate fidelity of 99%: the most attractive quantum operations are those that operate two qubits at the same time, so the call specifies that the error in operating 2-qubit gates must be less than 1%. One example of these operations is the CNOT gate, used to entangle two qubits.
- Qubit read-out fidelities of at least 95%, which, as the reader can imagine, means getting a 0 or a 1 correctly 95% of the time a circuit is read.
Are these numbers too high or too low?
Well, it depends. Let’s start with the number of qubits. The qubit architecture can favor one kind of operation over others. One of the oldest quantum computer vendors provides a 5k qubit (500x more than the call), but the chip (a quantum annealer) can only afford optimization operations. Then, for general-purpose, also called universal quantum computers, this number dramatically decreases and depends on the physical implementation of the qubit. However, 10 qubits is an affordable number and enough to work with small circuits. The other two specifications relate to the quality of the qubits. The number of quality parameters to differentiate between a “good” or a “bad” quantum chip is not small, but the call specifies only the minimum fidelities of 2-qubit gates and qubit readout. Considering the explanation given above, those numbers can sound reasonably large compared to the classical bit rate errors of a processor, but they are quite stringent considering the current status of the technology. In summary, this call asks for hosting quantum computers which will be powerful enough to encourage further investigations into the co-location of quantum computers with HPC clusters, and the development of new algorithms thanks to the economic effort made by EuroHPC JU and the participating states.
Is such an effort worth it?
I’m pretty sure the Wright brothers got the same question from their family and friends before they built the first powered airplane. The experimental machine they made in 1903 doesn’t have much in common with commercial airplanes such as the Boeing 787 Dreamliner, but everything has to start somewhere, powered by the hands and minds of talented scientists who want to push the boundaries of technology a little further.
Finally, to learn more about the quantum computing nomenclature used by EuroHPC, see this PRACE article