1: \begin{abstract}    

2:  We present a study of numerical effects in dissipationless

3:  cosmological simulations. The numerical effects are evaluated and

4:  studied by comparing results of a series of $64^3$-particle

5:  simulations of varying force resolution and number of time steps,

6:  performed using three of the $N$-body techniques currently used for

7:  cosmological simulations: the Particle Mesh (PM), the Adaptive

8:  Particle-Particle Particle-Mesh (\ap3m), and the newer Adaptive

9:  Refinement Tree (ART) code. This study  can therefore be interesting

10:  both as an analysis of numerical effects and as a systematic comparison

11:  of different codes. 

12: 

13: We find that the \ap3m and the ART codes produce similar results,

14: given that convergence is reached within the code type. We also

15: find that numerical effects may affect the high-resolution simulations

16: in ways that have not been discussed before. In particular, our study

17: revealed the presence of two-body scattering, effects of which can

18: be greatly amplified by inaccuracies of time integration. This process

19: appears to affect the correlation function of matter, mass function

20: and inner density of dark matter halos and other statistics at scales

21: much larger than the force resolution, although different statistics

22: may be affected in different fashion. We discuss the conditions at

23: which strong two-body scattering is possible and discuss

24: the choice of the force resolution and integration time

25: step. Furthermore, we discuss recent claims that simulations with

26: force softening smaller than the mean interparticle separation are

27: not trustworthy and argue that this claim is incorrect in general

28: and applies only to the phase-sensitive statistics. Our conclusion is

29: that, depending on the choice of mass and force resolution and

30: integration time step, a force resolution as small as 

31: $0.01$ of the mean interparticle separation can be justified.

32: \end{abstract}

33: