The time evolution of protons and 3 He fragments from Au+Au/Pb+Pb reactions at 0.25, 2, and 20 GeV/nucleon is investigated with the potential version of the Ultrarelativistic Quantum Molecular Dynamics (UrQMD) model combined with the traditional coalescence afterburner. In the coalescence process, the relative distance R 0 and relative momentum P 0 are surveyed in the range of 3-4 fm and 0.25-0.35 GeV/c, respectively. For both clusters, a strong reversed correlation between R 0 and P 0 is seen and it is time-dependent as well. For protons, the accepted (R 0 , P 0 ) bands lie in the time interval 30-60 fm/c, while for 3 He, a longer time evolution (at about 60-90 fm/c) is needed. Otherwise, much smaller R 0 and P 0 values should be chosen. If we further look at the rapidity distributions from both central and semi-central collisions, it is found that the accepted [t cut , (R 0 , P 0 )] assemble can provide consistent results for proton yield and collective flows especially at mid-rapdities, while for 3 He, the consistency is destroyed at both middle and projectile-target rapidities. PACS numbers: 24.10.Lx, 25.75.Dw, 25.75.-q, 24.10.-i a E-mail address: liqf@hutc.zj.cn 1 I. MOTIVIATIONThe production mechanism of particles and nuclei is a fundamental and essential problem for the whole evolution of the universe. With the help of macroscopic and/or microscopic transport models for heavy ion collisions (HICs) within a large range of beam energies from the GSI Schwerionen Synchrotron (SIS) up to the BNL Relativistic Heavy Ion Collider (RHIC) and the CERN Large Hadron Collider (LHC) energies, the process of a complete compression and decompression from initial two colliding nuclei can be described dynamically, with the possible occurrence of (phase) transitions from nuclear liquid to gas (LG) or from quark-gluon plasma (QGP) to hadron gas (HG). Such a model is, e.g. the Ultrarelativistic Quantum Molecular Dynamics (UrQMD)[1, 2] [? ] . After that, an afterburner is usually chosen for the freezeout of various particles (free baryons, mesons, etc.) or fragments (deuterons, tritons, Helium isotopes, etc.) which can then be used for comparison with corresponding experimental data. It is unavoidable that the afterburner should be paid much more attention when a serious comparison between calculations and experiments is needed for extracting, although, mainly the information of the compression phase at the early stage. Therefore, a large amount of such models are available but active in different beam energy regions (with their own problems when describing data). At low bombarding energies from several tens to several hundreds MeV/nucleon, the statistical multifragmentation model (SMM) [3, 4], the statistical evaporation model (HIVAP) [5, 6] and the statistical model (GEMINI) [7] are frequently used; at energies from several hundreds MeV/nucleon to several GeV/nucleon, the conventional phase-space coalescence model "Minimum Spanning Tree" (MST) [8] as well as the Simulated Annealing Clusterisation Algorithm (SACA)[12] are t...