IBM has made a significant advance on the road to larger and more powerful quantum computers by linking two such devices together and performing calculations beyond the capabilities of either alone. The result is a positive sign that the company鈥檚 bet on a modular approach to scaling up quantum computers is feasible.
Quantum computers promise to be able to solve certain problems much faster than conventional devices, but major hurdles remain, including making the computers large enough while reducing errors. Various research groups and companies are taking different approaches to tackling these hurdles, and IBM is using superconducting chips that can be produced by the same machines that make existing computer hardware.
One issue with this choice is that the input and output wires for the chips are far larger than the quantum bits, or qubits, that do the calculating. This means the qubits must be spaced further apart than the transistors in a conventional processor, reducing the number that can be squeezed onto a chip. That is why IBM wants to be able to link up its chips.
Advertisement
鈥淚t鈥檚 a game that you win by sort of breaking it up into chunks,鈥� says at IBM. 鈥淭his is a central choice, [an] engineering choice we鈥檝e made in terms of how we鈥檙e scaling up our systems. And there are scientific questions to answer as well as engineering questions to answer about the consequences of that choice.鈥�
That is because getting quantum chips to communicate with each other is much more difficult than doing so with conventional chips, where data is represented by the absence or presence of an electrical signal. Quantum computers instead rely on the strange physics of quantum entanglement, which can鈥檛 simply be transmitted by a wire.
To get around this, researchers have developed a method to entangle a pair of qubits and teleport one to a second chip, forging a quantum link between the two devices that must be mediated by a conventional computer. IBM has now successfully demonstrated this by connecting two of its Eagle quantum processing units, each with 127 qubits, to perform a calculation requiring a total of 142 qubits 鈥� more than could fit at one time on either chip alone.
at the University of Oxford says a modular approach is considered the most likely path by which current prototypes can be scaled up into useful quantum computers. Until now, however, it hasn鈥檛 been clear that this would be a simple thing to achieve.
鈥淚f you don鈥檛 have entanglement, then this cannot work,鈥� says Gogioso. 鈥淏ut once you do have an entangled resource, you can always use teleportation to swap a quantum channel for a classical channel 鈥� that鈥檚 a fairly well understood trick. [But] obviously, doing it in practice is different than doing it in theory.鈥�
at the University of Texas at Austin says the idea of tying together multiple quantum chips has been discussed for decades, but IBM has now demonstrated an early step towards that goal 鈥� although there are still hurdles to overcome.
鈥淩eally scaling up superconducting quantum computers will require a much more high-fidelity version of the same thing, one that would let hundreds or thousands of superconducting chips operate as one,鈥� he says.
Nature
