Theor Chem Account (2008) 120:307–312 DOI 10.1007/s00214-007-0300-z REGULAR ARTICLE Assessment of Two Surface Monte Carlo (TSMC) method to find stationary points of (H 2 O) 15 and (H 2 O) 20 clusters Pradipta Bandyopadhyay Received: 15 November 2006 / Accepted: 15 February 2007 / Published online: 1 May 2007 © Springer-Verlag 2007 Abstract The Two Surface Monte Carlo (TSMC) technique reduces computational cost by using a computa- tionally cheap biasing potential, which guides the molecular system to explore the potential energy surface of interest. It was shown earlier that the Effective Fragment Potential (EFP) can be a good choice for this biasing potential (Bandy- opadhyay, J Chem Phys 122:091102, 2005) when the poten- tial energy surface of interest is quantum mechanical. This may help in expanding the applicability of TSMC, since find- ing a good biasing potential is a major challenge. In the pres- ent work, the viability of TSMC method in finding stationary points of large molecular system is investigated using EFP as the biasing potential and RHF theory as the potential of inter- est. TSMC is applied to find the stationary points of water clusters of size 15 and 20. A semi-automated method starting from random geometries, without using any chemical intui- tion, found several stationary points. The simulated anneal- ing method was used to refine the structures obtained from TSMC. Among the several low-energy structures obtained for 15 water cluster, one minimum, about 1 kcal/mol higher than the global minimum, was found. However, for 20 water cluster, no structure very close to the global minimum was obtained. Several strategies, learned from the experience of the present work, are discussed for improving the TSMC method, including the acceptance between the two energy surfaces. Keywords Monte Carlo · Effective fragment potential · Stationary points · Water clusters · Global minimum Contribution to the Mark S. Gordon 65th Birthday Festschrift Issue. P. Bandyopadhyay (B ) Department of Biotechnology, Indian Institute of Technology, Guwahati 781039, Assam, India e-mail: pradipta@iitg.ernet.in; praban07@gmail.com 1 Introduction Computational investigation of floppy molecular systems, such as water clusters and flexible proteins, is challenging due to the presence of a large number of stationary points, many of which may be close in energy and accessible during experiments at room temperature. In quantum chemistry, sta- tionary points are typically determined by employing local minimization methods from reasonable starting structures constructed by chemical intuition. However, because of diver- sity of different conformations of large fluxional systems, finding a reasonable starting structure becomes increasingly difficult with increase in size and flexibility of the molecule. Molecular simulation techniques such as Molecular Dynam- ics (MD) and Monte Carlo (MC) can also be used to find stationary points. In this method, simulations are done to explore different structures of molecules and then local min- imization methods are used to minimize selected low-energy structures from the simulation trajectory. One major problem with the simulation techniques is the quality of the potential function used. Ideally, one would like to use high quality energy function, including the quantum mechanical (QM) or quantum mechanical/molecular mechanical (QM/MM) ones. However, the use of QM or QM/MM energy functions in molecular simulation is prohibitively expensive for investi- gations in a routine manner. The use of approximate energy functions can make the simulation feasible; however, the quality of structures obtained from such simulations could be poor and may be far from the stationary points on the QM energy surface. The Two Surface Monte Carlo (TSMC) technique is a promising method to circumvent this problem. In this method, a cheap potential function is used as a biasing potential to propagate the more expensive potential function of inter- est. Jumps between the cheap and the expensive potential 123