A force field for water has been developed entirely from first principles, without any fitting to experimental data. It contains both pairwise and many-body interactions. This force field predicts the properties of the water dimer and of liquid water in excellent agreement with experiments, a previously elusive objective. Precise knowledge of the intermolecular interactions in water will facilitate a better understanding of this ubiquitous substance.

Molecular scattering behavior has generally proven difficult to study at low collision energies. We formed a molecular beam of OH radicals with a narrow velocity distribution and a tunable absolute velocity by passing the beam through a Stark decelerator. The transition probabilities for inelastic scattering of the OH radicals with Xe atoms were measured as a function of the collision energy in range of 50 to 400 wavenumber, with an overall energy resolution of about 13 wavenumbers. The behavior of the cross sections for inelastic scattering near the energetic thresholds was accurately measured, and excellent agreement was obtained with cross-sections derived from coupled-channel calculations on ab initio computed potential energy surfaces.The study of collisions between gas-phase atoms and molecules is a well-established method of gathering detailed information about their individual structures and mutual interaction (1 ). The level of detail obtained by these studies depends on the quality of preparation of the collision partners before the collision (2-4 ) and on how accurately the products are analyzed afterward (5-7 ). In recent years, it has become increasingly possible to control the internal and external degrees of freedom of the scattering partners, allowing the potential energy surfaces that govern the molecular collisions to be probed in ever greater detail. The most detailed information is obtained when crossed molecular beams are used to produce intense jets of molecules with a well-defined velocity, confined to only a few internal quantum states. Further state selection can be achieved by optical preparation of a single quantum state or by purification of the beam with the use of electrostatic or magnetic multipole fields (2, 3 ). These methods allow the orientation of the molecules to be controlled before the collision (8, 9 ) and the orientation of the scattered products can be measured (10 ).One of the most important parameters describing a scattering event is the collision energy of the scatterers. However, control over the collision energy has been a difficult experimental task. Since the 1980's, ingenious crossed beam machines have been engineered to vary the crossing angle of the intersecting beams, allowing variation of the collision energy while maintaining particle densities high enough for scattering (11 ). It was thereby possible to measure threshold behavior of rotational energy transfer (12, 13 ), or to tune the collision energy over the reaction barrier for reactive scattering (14, 15 ). These methods led to considerable improvement in the control over collision energy at high energiesfor example to probe short-range interactions. However, a similar level of control over collisions at low energies, which are sensitive probes for long-range interactions, is generally lacking. The angle of the 1

A new ab initio pair potential for water was generated by fitting 2510 interaction energies computed by the use of symmetry-adapted perturbation theory ͑SAPT͒. The new site-site functional form, named SAPT-5s, is simple enough to be applied in molecular simulations of condensed phases and at the same time reproduces the computed points with accuracy exceeding that of the elaborate SAPT-pp functional form used earlier ͓J. Chem. Phys. 107, 4207 ͑1997͔͒. SAPT-5s has been shown to quantitatively predict the water dimer spectra, see the following paper ͑paper II͒. It also gives the second virial coefficient in excellent agreement with experiment. Features of the water dimer potential energy surface have been analyzed using SAPT-5s. Average values of powers of the intermolecular separation-obtained from the ground-state rovibrational wave function computed in the SAPT-5s potential-have been combined with measured values to obtain a new empirical estimate of the equilibrium O-O separation equal to 5.50Ϯ0.01 bohr, significantly shorter than the previously accepted value. The residual errors in the SAPT-5s potential have been estimated by comparison to recent large-scale extrapolated ab initio calculations for water dimer. This estimatetogether with the dissociation energy D 0 computed from SAPT-5s-leads to a new prediction of the limit value of D 0 equal to 1165Ϯ54 cm Ϫ1 , close to but significantly more accurate than the best empirical value.

Nearly exact six-dimensional quantum calculations of the vibration-rotation-tunneling ͑VRT͒ levels of the water dimer for values of the rotational quantum numbers J and K р2 show that the SAPT-5s water pair potential presented in the preceding paper ͑paper I͒ gives a good representation of the experimental high-resolution far-infrared spectrum of the water dimer. After analyzing the sensitivity of the transition frequencies with respect to the linear parameters in the potential we could further improve this potential by using only one of the experimentally determined tunneling splittings of the ground state in (H 2 O) 2 . The accuracy of the resulting water pair potential, SAPT-5st, is established by comparison with the spectroscopic data of both (H 2 O) 2 and (D 2 O) 2 : ground and excited state tunneling splittings and rotational constants, as well as the frequencies of the intermolecular vibrations.