arXiv subjects

Shimon Cohen

Publications and source records attributed to Shimon Cohen.

2 recordsLinked to original sources

arXiv2025

Clocking and controlling attosecond currents in a scanning tunnelling microscope

Quantum tunnelling of electrons can be confined to the sub-cycle time scale of strong light fields, contributing decisively to the extreme time resolution of attosecond science. Because tunnelling also enables atomic-scale spatial resolution in scanning tunnelling microscopy (STM), integrating STM with light pulses has long been a key objective in ultrafast microscopy, spanning the picosecond and femtosecond domains, with first signatures of attosecond dynamics. However, while sub-cycle dynamics on the attosecond time scale are routinely controlled and determined with high precision, controlling the direction of attosecond currents and determining their duration have remained elusive in STM. Here, we induce STM tunnelling currents using two-colour laser pulses and dynamically control their direction, relying solely on the sub-cycle waveform of the pulses. Projecting our measurement data onto one-electron and many-body theory descriptions reveals a three-step transport process in the non-adiabatic tunnelling regime as the physical mechanism, with a theory-derived current burst duration of 860 as. Despite working under ambient conditions but free of thermal artifacts, we achieve sub-angstr\"om topographic sensitivity and a lateral spatial resolution of 2 nm. This unprecedented capability to directionally control attosecond bursts will enable triggering and imaging ultrafast charge dynamics at the spatio-temporal microscopy frontier of lightwave electronics.

quant-ph↗

arXiv2017

Artificial Neural Networks that Learn to Satisfy Logic Constraints

Logic-based problems such as planning, theorem proving, or puzzles, typically involve combinatoric search and structured knowledge representation. Artificial neural networks are very successful statistical learners, however, for many years, they have been criticized for their weaknesses in representing and in processing complex structured knowledge which is crucial for combinatoric search and symbol manipulation. Two neural architectures are presented, which can encode structured relational knowledge in neural activation, and store bounded First Order Logic constraints in connection weights. Both architectures learn to search for a solution that satisfies the constraints. Learning is done by unsupervised practicing on problem instances from the same domain, in a way that improves the network-solving speed. No teacher exists to provide answers for the problem instances of the training and test sets. However, the domain constraints are provided as prior knowledge to a loss function that measures the degree of constraint violations. Iterations of activation calculation and learning are executed until a solution that maximally satisfies the constraints emerges on the output units. As a test case, block-world planning problems are used to train networks that learn to plan in that domain, but the techniques proposed could be used more generally as in integrating prior symbolic knowledge with statistical learning

cs.LG↗