1: \begin{abstract}

2: Fluctuation relations (FRs) are among the few existing general results in

3: non-equilibrium systems. Their verification requires the measurement of

4: the total work (or entropy production) performed on a system. 

5: Nevertheless in many cases only a partial measurement of the

6: work is possible. Here we consider FRs in

7: dual-trap optical tweezers where two different forces (one per

8: trap) are measured. With this setup we perform pulling experiments on

9: single molecules by moving one trap relative to the other. We

10: demonstrate that work should be measured using the force exerted by the

11: trap that is moved. The force that is measured in the trap at rest fails to provide 

12: the full dissipation in the system leading to a (incorrect) work definition that does not 

13: satisfy the FR. The implications to single-molecule experiments

14: and free energy measurements are discussed. In the case of symmetric

15: setups a new work definition, based on differential force measurements,

16: is introduced. This definition is best suited to measure free energies

17: as it shows faster convergence of estimators. We

18: discuss measurements using the (incorrect) work definition as an example

19: of partial work measurement. We show how to infer the full work distribution 

20: from the partial one via the FR. The inference

21: process does also yield quantitative information, e.g. the hydrodynamic

22: drag on the dumbbell. Results are also obtained for asymmetric dual-trap

23: setups. We suggest that this kind of inference could represent a new and

24: general application of FRs to extract information about 

25: irreversible processes in small systems.

26: \end{abstract}

27: