# When can a Computer Simulation act as Substitute for an Experiment? A Case-Study from Chemisty

Johannes Kästner and Eckhart Arnold

## 1 Introduction

Since computer simulations have caught the attention of philosophers of science, there are ongoing debates concerning their ontological status (Barberousse et al. 2009), their function in science and their epistemic reach (Humphreys 2004), whether they are a novelty or merely introduce a more powerful but not essentially different form of modeling (Frigg/Reiss 2009, Humphreys 2009), how they affect scientific practice (Winsberg 2010) and how they are related to other tools of science like “Gedankenexperimente” on the one hand

or “material” experiments on the other hand.

In this paper we pick up the debate whether simulations are experiments. This debate appears to have reached a stalemate: The arguments pro and contra have been exchanged (Guala 2002, Morgan 2003, Parker 2009, Winsberg 2009, Morrison 2009) and obvious misunderstandings have been clarified, but no universally accepted solution has been reached. We believe that this stalemate is in part due to the nature of the question. For, given that simulations and experiments have some features in common but also differ in certain respects, the answer to the question depends on whether the common features or the differences are considered essential, which ultimately depends on the point of view of the philosopher investigating the question.

However, because experiments play a very distinctive role in science, the question is too important to be left just at that. In order to break the stalemate, we therefore suggest shifting the emphasis of this question. Quite frequently simulations are used in situations where experiments are not feasible for practical or ethical reasons. Instead of asking the question whether these simulations are truly experiments, we should ask:

1. Under what conditions can a simulation serve as a surrogate for an experiment?

2. What requirements must be met so that the results of a simulation are at least as reliable as the the results of an analogous experiment would be?

We will investigate this question with a case study from theoretical chemistry, the simulation of $H_2$-formation in outer space on the basis of quantum-mechanics (Goumans/Kaestner 2010). Different aspects of the simulation process will be discussed with respect to the epistemic reliability of the simulation. We identify the existence of comprehensive and empirically well-confirmed background theories as a crucial requirement for the reliability of simulations of experimentally inaccessible phenomena.