1: \begin{abstract}

2: Particle Swarm Optimization (PSO) is a nature-inspired

3: meta-heuristic for solving continuous optimization problems.

4: In \cite{SWa:13,SWc:15}, the potential of the particles of a swarm has been

5: used to show that slightly modified PSO guarantees convergence to local optima.

6: Here we show that under specific circumstances the unmodified PSO,

7: even with swarm parameters known (from the literature) to be ``good'',

8: almost surely does not yield convergence to a local optimum is provided.

9: This undesirable phenomenon is called stagnation.

10: For this purpose, the particles' potential in each dimension is analyzed mathematically.

11: Additionally, some reasonable assumptions on the behavior of the particles' potential are made.

12: Depending on the objective function and, interestingly, the number of particles,

13: the potential in some dimensions may decrease much faster than in other dimensions.

14: Therefore, these dimensions lose relevance,

15: i.\,e., the contribution of their entries to the decisions about attractor updates

16: becomes insignificant and, with positive probability, they never regain relevance.

17: If Brownian Motion is assumed to be an approximation of the

18: time-dependent drop of potential, practical, i.\,e., large

19: values for this probability are calculated.

20: Finally, on chosen multidimensional polynomials of degree two, experiments are provided

21: showing that the required circumstances occur quite frequently.

22: Furthermore, experiments are provided showing that

23: even when the very simple sphere function is processed

24: the described stagnation phenomenon occurs.

25: Consequently, unmodified PSO does not converge to any local optimum of the chosen functions for tested parameter settings.

26: %\end{abstract}

27: