This work is part of ongoing research on stationary natural gas engines equipped with unscavanged prechambers. Spark ignition inside the prechamber has been successfully proven with natural gas as well as biogas, reducing emissions while keeping good efficiencies. The conversion of the prechamber system from spark ignition to auto-ignition will lead to longer maintenance intervals in the case of stationary natural gas engines and to a probable more efficient and cleaner combustion. The latter expectation is based on the combustion mode: HCCI like combustion inside the prechamber and faster combustion in the main chamber due to an earlier arrival of the flame front, compared to prechamber spark ignition. To the authors knowledge no work following this approach has been conducted. From this new approach one of the major difficulties arises. While for standard prechamber systems either the spark plugs or injecting gas directly into the prechamber creating rich conditions guarantee to well know the point of ignition, auto-ignition of a premixed gas does not do so. The basic idea of this prechamber is to create as homogeneous conditions as possible. Then, the reacting mixture streams through the orifices into the main chamber. The difficulty here is to create a state that triggers auto-ignition only inside the prechamber. As temperature is the main influencing factor on the reaction chemistry the prechamber walls are heated, aiming at higher gas temperatures inside the prechamber. However, auto-ignition will take place wherever the energy level is sufficiently elevated. In particular, this causes difficulties in the region of the exhaust valve. In this work the three- dimensional flow conditions in a mono-cylinder test engine were numerically simulated in order to better understand the conditions inside the engine, with the new prechamber configuration. Therefore, a geometric model was created, disregarding intake and exhaust valves to simplify matters. A suitable multi-block mesh for parallel computing was generated and models for the initial and boundary conditions derived. The chemical reactions were incorporated to the flow field, using a 54 species mechanism developed for natural gas combustion by two approaches. In a first step the flow field was superposed with zero-dimensional Chemkin calculations whereas in a second step the species conservation equations were solved together with the Navier Stokes equations, which is much more expensive in calculation time. Finally, the numerical results were compared to experimentally gained data. Results show the sensitivity of the computations to varied initial and boundary conditions and provide a guideline for modifications in order to enhance the prechamber configuration. Numerical results show the risk of auto-ignition in the main chamber. Experimental results could be reproduced numerically in regard to the time dependency of first ignition.