1: \begin{abstract}

2: When studying high-dimensional dynamical systems such as macromolecules,

3: quantum systems and polymers, a prime concern is the identification

4: of the most probable states and their stationary probabilities or

5: free energies. Often, these systems have metastable regions or phases,

6: prohibiting to estimate the stationary probabilities by direct simulation.

7: Efficient sampling methods such as umbrella sampling, metadynamics

8: and conformational flooding have developed that perform a number of

9: simulations where the system\textquoteright{}s potential is biased

10: such as to accelerate the rare barrier crossing events. A joint free

11: energy profile or stationary density can then be obtained from these

12: biased simulations with weighted histogram analysis method (WHAM).

13: This approach (a) requires a few essential order parameters to be

14: defined in which the histogram is set up, and (b) assumes that each

15: simulation is in global equilibrium. Both assumptions make the investigation

16: of high-dimensional systems with previously unknown energy landscape

17: difficult. Here, we introduce the transition matrix unweighting (TMU)

18: method, a simple and efficient estimation method which dismisses both

19: assumptions. The configuration space is discretized into sets, but

20: these sets are not only restricted to the selected slow coordinate

21: but can be clusters that form a partition of high-dimensional state

22: space. The assumption of global equilibrium is replaced by requiring

23: only local equilibrium within the discrete sets, and the stationary

24: density or free energy is extracted from the transitions between clusters.

25: We prove the asymptotic convergence and normality of TMU, give an

26: efficient approximate version of it and demonstrate its usefulness

27: on numerical examples.\end{abstract}