The new 2013 release of Geant4 is the first major release of the simulation toolkit since 2007. Release 10.0 incorporates many changes in almost all areas, including an evolved API for allowing efficient execution of simulation programs on modern parallel architectures, exploiting multi-threading with parallelisation at the event level. The introduction of multi-threading in Geant4 is the result of a long study started in 2009 in collaboration with US colleagues from the North-Eastern University; members from the simulation team in the CERN PH/SFT group have been strongly involved since the first phases of the project, with its conceptualization and the first feasibility studies.

Particular care was taken in designing the system, in order to allow preserving backwards compatibility with user’s code, traditionally running in sequential mode. Considerable effort has been spent in verifying all possible use-cases and guaranteeing reproduction of correct results. Measurements performed on different hardware architectures reveal excellent linearity of throughput speedup of the order of 94% and reduced memory footprint per-thread, with good results also when running in hyper-threading regime.

The new release of Geant4 introduces many new features in different areas, including geometry modelling as well as electromagnetic and hadronic physics. The code in the Geant4 geometry modeller has been completely reviewed to become thread-safe for event-level parallelism. It will include a new set of geometrical primitives (solids) that can be used as alternatives to the traditional primitives as part of an optional component extracted from the AIDA Unified Solids Library. The new library will provide improved implementations for most of the traditional shapes, in particular for those (tessellated solid, polycone, polyhedra) where an innovative spatial optimization technique could be applied.

The Geant4 set of electromagnetic (EM) physics processes and models is a key component of the toolkit, covering a wide spectrum of applications, including simulations of high energy and nuclear physics experiments, beam transport, medical physics, cosmic ray interactions and radiation effects in space. At the LHC, it is important for electromagnetic shower simulation that is essential for the analysis of H->γγ and Z->ee decays and other reaction channels. In Geant4 release 10 all EM sub-packages have been fully adapted to work efficiently in multi-threaded mode while significant effort was spent to improve their overall CPU performance.  A number of EM physics models have been improved; in particular, high-energy gamma conversion, models of fluctuations in thin layers, photo-absorption ionisation model for accurate simulation of ionisation in gases and thin solid layers. Moreover, multiple and single scattering models have been further tuned. The Urban model for multiple scattering has been consolidated and provides very good results for all benchmarks with primary electron beams. In parallel, an alternative “combined” approach of multiple scattering and single Coulomb scattering models has been developed. A number of validations have demonstrated that the combined model provides more accurate simulation results for muon data and it is  applicable for wide variety of target thicknesses, densities, and energies.

The interest of the LHC experiments in the modelling of hadronic interactions in Geant4 is mainly related to the simulation of hadronic showers, which is an essential ingredient of the simulation of jets. In addition they are important for describing hadronic interactions in the trackers. The simulation of hadronic interactions is notoriously complex due to the fact that the fundamental theory of strong interactions, quantum chromo-dynamics (QCD), cannot be applied via a ‘perturbative’ approach. This means that only approximate models can be used to simulate hadronic interactions. Moreover, these models have a limited applicability in terms of type of hadron and energy range, so that a combination of different models is needed to simulate hadron-nucleus collisions from TeV down to thermal energies. The choice of the model (so-called “physics list”) depends on the compromise between accuracy and speed of the simulation; more sophisticated models make the simulation more realistic but slower. The recommended physics list is called "FTFP_BERT" and is made of three main components: a string model, Fritiof (FTF), which handles the high-energy part of the hadron-nucleus interaction, an intra-nuclear cascade model (Bertini) for the intermediate energy range (10 GeV to 100 MeV), and a low-energy model, pre-compound/evaporation, which takes care of the nuclear de-excitation.

The hadronic physics developments in Geant4 included in release 10 are touching many areas. For high-energy collisions, the diffraction dissociation in Fritiof has been considerably revised, and the model has been re-tuned based on an enlarged set of thin-target published data. The Fritiof model has been extended to handle also nucleus-nucleus interactions above a few GeV per nucleon. For intermediate energies, the Bertini-like model has been improved in the two-body angular distributions, and in the phase-space generation of multi-body final states; it has been also extended to treat muon absorption in nuclei. At low energies, the de-excitation model has been revised to be able to handle meta-stable excited nuclei (isomers). Neutron inelastic nuclear cross sections below 20 MeV have been improved, adopting an isotope-based approach based on the more accurate, but slower, neutron high-precision model: neutron energies are grouped in bins and the average values of the neutron cross section are used.

Both for the physics and technical changes made in hadronic physics, a significant effort has been put on testing and validation. The set of thin-target benchmarks, which are used for tuning and validating the hadronic models, is now covering a large combination of projectile particle type, projectile energy, and target materials. A set of simplified calorimeters, which reproduce the main features of the calorimeters used in the LHC experiments, are used to compare the properties of hadronic showers (energy response, energy resolution, longitudinal and lateral shapes) between different versions of Geant4. For the recommended FTFP_BERT physics list, the hadronic shower observables in version 10 remain quite similar to those of version 9.6, with the only exception of tungsten calorimeters, where a more accurate treatment of neutron capture has reduced significantly the energy response and the lateral size of hadronic showers.

In conclusion, Geant4 Release 10.0  represents a big step forward in the evolution of the Geant4 toolkit in terms of its physics and software capabilities. As for any new release of Geant4, a lot of valuable feedback is encouraged and expected from the user community.