### Speaker

### Description

Stochastic Thermodynamics usually investigates and describes Markovian systems where memory is absent. A typical example is Brownian Motion, where dynamics are overdamped with a delta-correlated noise term. Yet this Markovian hypothesis is very strong, and only valid for a limited range of time and length scales. More generally, a wide range of physical systems exhibit memory with finite relaxation times, which cannot be captured by Markovian models. Such memory effects not only derive from nonidealities, but can also be actively sought, to develop new physical properties and applications. In this work we apply stochastic Thermodynamics to a model non-Markovian bath made of a viscoelastic fluid. This complex fluid is characterized by a stress memory: energy injected in the bath is locally trapped and needs time to fully diffuse and be absorbed. In theory, this energy can be interpreted as a local excess heat which should be usable to produce deterministic work.

In our experiments, we measure the stochastic work, heat and internal energy of a colloidal particle inside a moving optical potential. We first show that the first law still applies, and that work is fully converted into heat and potential energy. Next we highlight that after applying work on the system, heat (and therefor entropy) can temporarily decrease, confirming the previous hypothesis. This negative heat is still limited by the amount of work that was initially spent into the system, and therefore doesn't violate the second law of thermodynamics. Finally, we discuss the implications of this finding and highlight some consequences for thermodynamics cycles, which could optimize such effects.