True randomness is surprisingly hard to come by: even the best generators harbour small systematic biases that can weaken cryptography. Physicists at ETH Zurich led by Renato Renner of the Institute for Theoretical Physics and Andreas Wallraff of the Laboratory for Solid-State Physics have now produced random numbers that are certifiably free of bias, the first time a quantum experiment has achieved this, and shown how imperfect randomness can be amplified into certified perfect randomness. The work appears in Nature.
The experiment linked two superconducting qubit chips, cooled to near absolute zero, through a 30-metre cooled tube along which microwave photons travel and entangle the qubits. The team then ran a Bell test on the entangled pair, choosing measurement settings with an imperfect random number generator, and processed the outcomes with a special algorithm that amplifies the results into provably unbiased random bits.
Crucially, the certification holds independently of the analysis methods used to test the output, giving users a physical guarantee rather than a statistical assurance. The researchers compare the potential role of such devices in digital security to that of atomic clocks in timekeeping: a trusted source others can rely on. Applications range from encrypting communications and digital identities to public randomness services for lotteries and blockchains.