---

As physical implementations of small quantum computers become a reality, we investigate two common bottlenecks: limited memory size and noise caused by heat dissipation. In this talk I will go over the theoretical fundamentals behind heat production in quantum information processing, and how it can be alleviated without a large cost in terms of quantum memory requirements. The talk will be accessible to non-experts, familiar with quantum mechanics. Landauer posited that irreversible information-processing operations (like the erasure of a classical or quantum memory at the end of a computation) have an intrinsic thermodynamic work cost, which is associated with heat dissipation to the environment [1]. In the case of a quantum computer this heat dissipation is particularly undesirable, as the computer relies on cooling to keep coherence of quantum states. Bennett proposed a solution: if we run the circuit up until the final measurements and then reverse it, we keep heat production to a theoretical minimum [2]. However, this proposal has a cost in terms of quantum memory: none of the quantum registers that are only needed for parts of the computation can be freed until the very end. In the intermediate term, when we have only a few hundreds of qubits in a quantum memory, this is a bottleneck for parallelization of quantum computations: we have to wait to finish an algorithm before we can free the memory to run another. In recent work, we try to ameliorate this problem by investigating "online erasure" of quantum registers that are no longer needed for a given algorithm. That way, we could free up quantum memory already halfway through a computation. We study the minimal cost of erasure in these scenarios, using results on the erasure of entangled quantum registers [3]. For a specific class of algorithms (the hidden subgroup problem) we find optimal online erasure protocols, and see how we could use them to optimize the algorithms [4].

---

Lidia Del Rio, ETHZ
Seminar Room 0.03, ETP
Contact: Sebastian Diehl