The intensity correlation spectrum of an ultracold 20 Ne atomic beam in the 1s 3 ͓3s : 1 P 0 ͔ state is experimentally studied. The spectrum shows a peak around the origin with the width corresponding to the kinetic energy spread of atoms in the beam source. [S0031-9007(96) PACS numbers: 32.80.Pj Various interferometric effects of neutral atomic beams have been demonstrated in recent years [1][2][3][4][5][6][7][8][9]. However, those experiments deal with the wave nature of a single atom. This is in sharp contrast to the case of optical beams, in which interesting results on many-photon correlation effects have been studied. This is partly due to the difference in the technical level of preparing samples. The invention of lasers and following development in nonlinear optics enabled us to work with optical beams with various statistical characteristics, whereas for particles with mass the only sample available for us was a beam of uncorrelated particles. For random particles the first order correlation is constant in time, and a nontrivial spectrum can be observed only in higher order correlations. Furthermore, to observe a many-particle correlation effect, at least two particles have to be found in a single external quantum state, which was prohibitively small with a conventional particle beam. This situation has changed considerably for a neutral atomic beam owing to the laser cooling technique. In a Bose-Einstein condensate of alkali gases, which has been reported recently by several groups [10][11][12], atoms are in a highly degenerate state. Even with commonly used laser-cooling techniques it is possible to achieve a density in which the probability of finding two atoms in the same mode is in an experimentally detectable range.We report in this Letter the first observation [13] of the second order correlation of a laser-cooled atomic beam, which is the atomic analogy of the Hambury Brown and Twiss experiment on an optical source [14]. An ultracold metastable 20 Ne atomic beam in the 1s 3 ͓3s : 3 P 0 ͔ state was generated by releasing atoms from a Ne trap in the 1s 5 ͓3s : 3 P 2 ͔ state by optical pumping [15]. We restricted the area of the detector to cover only the diffraction limited portion of the atomic beam source and measured the time-interval distribution between two atoms that successively hit the detector. The temporal correlation showed a peak around the origin that corresponded to the energy distribution of the atomic beam.The joint probability P͑r 1 , t; r 2 , t 1 t͒ of finding an atom at t and r 1 and then an another atom at a later time t 1 t and r 2 is P͑r 1 , t; r 2 , t 1 t͒ ͗Cjd y ͑r 1 , t͒d y ͑r 2 , t 1 t͒ 3 d͑r 2 , t 1 t͒d͑r 1 , t͒jC͘ , where d y ͑r, t͒ and d͑r, t͒ are the operators to create and annihilate the atom at r and t, respectively. The probability P͑t͒ of detecting two atoms separated in time by t is obtained by averaging the above expression on r 1 and r 2 over the detector surface and on t. In our experimental setup the evaluation of P͑t͒ is not difficult, if the atomic beam...