MARS15: Radiation Transport Simulation Code

• MARS Code is a general-purpose Monte Carlo simulation framework that models hadronic and electromagnetic cascades in complex 3D accelerator and shielding geometries.
• It integrates specialized modules like LAQGSM, EGS5, and ALICE2011 for accurate particle interactions over a wide energy range from keV to TeV.
• The system features advanced ROOT-based geometry, mesh scoring, and residual-dose analysis, supporting detailed predictions for accelerator and radiation applications.

MARS, most commonly encountered in contemporary accelerator literature as MARS15, is a general-purpose radiation-transport Monte Carlo code developed since 1974 for detailed simulation of hadronic and electromagnetic cascades in arbitrary three-dimensional geometries representing shielding, accelerator, detector, and spacecraft components. In the MARS15(2012) description, its energy coverage extends from a fraction of an electronvolt to 100 TeV; a later status report specifies $1\,\text{keV}$ to $100\,\text{TeV}$ for hadrons, muons, heavy ions, and electromagnetic showers, and $10^{-5}\,\text{eV}$ to $100\,\text{TeV}$ for neutrons. Across these regimes, the code is used for particle transport, radiation-environment modeling, energy deposition, background production, activation, radiation damage, nuclide production, decay, transmutation, and residual activity analyses in complex engineering systems (Mokhov et al., 2014, Rakhno et al., 3 Oct 2025).

1. Scope, design objectives, and operating modes

MARS15 is designed for high-accuracy radiation and particle-transport studies associated with accelerator and shielding problems, with particular emphasis on the demands imposed by Intensity Frontier facilities using Megawatt beams, including ESS, FAIR, and Project X. Those facilities require consistent treatment across a wide energy span, from sub-eV neutron behavior through GeV-scale hadronic production and MeV-scale electromagnetic cascades, together with activation and radiation-damage calculations in realistic beam lines, targets, shielding assemblies, and detector environments (Mokhov et al., 2014).

A defining characteristic of the code is its ability to treat many physical processes exclusively, inclusively, or in mixed mode. This is operationally important because shielding design, background analysis, source-term generation, and residual-activity studies do not impose identical fidelity requirements. Exclusive event generation is relevant when detailed secondaries or topology matter; inclusive treatment is often sufficient for transport-driven source descriptions; mixed mode permits hybrid workflows within a single code base. In the broader MARS family, MARS15 functions as the central transport and activation engine, and later specialized systems such as FermiCORD are implemented as add-ons around that core rather than as replacements (Grebe et al., 2016).

2. Transport physics and model development

A major development described for MARS15(2012) is improved modeling of particle production in the 0.7 to 12 GeV range, a regime of direct importance for medium-energy accelerator targets and shielding. The code incorporates inclusive and exclusive event generators, including a modified LAQGSM module for this interval, with improved elementary total, elastic, and inelastic cross sections for NN and IN interactions, explicit reaction channels such as $N+N \to N+N+m\,m$, $T+N \to N+m\,m$ with $m<5$, $B+B \to B+Y+K$, $T+B \to Y+K$, $\bar{K}+B \to Y+T$, and $100\,\text{TeV}$0, $100\,\text{TeV}$1 pair production, as well as phase-space/isobar combinations and extensions down to GDR and below. The same development track also extends treatment to arbitrary light nuclear projectiles and nuclear targets. For charged pion production in the few-GeV region, the paper notes that a two-source model is useful as an inclusive generator until low-energy pion production in LAQGSM is improved further (Mokhov et al., 2014).

At lower energies, MARS15 adds improved modeling of proton inelastic interactions below 20 MeV through ALICE2011, converted into an event generator for nucleon, photon, and heavy-ion reactions and matched to CEM and LAQGSM above 20–30 MeV. On the electromagnetic side, MARS integrates EGS5 for shower simulation from 1 keV to 20 MeV; MARS manages EGS5 initialization, shared geometry, magnetic fields, and scoring, while users control the electromagnetic package through PEGS input files, which MARS can generate with default values. This extension is specifically identified as important for transition effects in fine accelerator and detector structures, background studies, and medical applications (Mokhov et al., 2014).

Stopping-power modeling was also revised. The conventional Bragg’s rule can disagree with measurements by as much as 20% at low energies and for low-$100\,\text{TeV}$2 materials, so MARS15 implements the cores-and-bonds method (CAB) for compounds in the 1 keV to 3 MeV interval, incorporating chemical-bond effects fitted to experiment; at higher energies it uses the Sternheimer and Peierls density correction algorithm. A later status report extends the low-energy nuclear-data infrastructure further through updated TENDL support for $100\,\text{TeV}$3, $100\,\text{TeV}$4, $100\,\text{TeV}$5, $100\,\text{TeV}$6, $100\,\text{TeV}$7, $100\,\text{TeV}$8, and $100\,\text{TeV}$9 projectiles. In that report, TENDL-2023 is described as extending the covered range down to 250 keV and up to 600 MeV, with energy distributions, angular distributions, and residual-nuclei data for stable and unstable targets; the exclusive option is already extended, while the inclusive option $10^{-5}\,\text{eV}$0 is still being tested (Rakhno et al., 3 Oct 2025).

3. Activation, radiation damage, and residual-dose analysis

MARS15 includes a substantial activation and damage-analysis stack. A new neutron displacement-per-atom (DPA) module covers neutrons from $10^{-5}\,\text{eV}$1 to 20–150 MeV using the NJOY99+ENDF-VII database for 393 nuclides. The implementation calculates displacements with the NRT model and then corrects the result using the experimentally measured defect production efficiency $10^{-5}\,\text{eV}$2, defined as the ratio of the number of single interstitial atom vacancy pairs (Frenkel pairs) produced in a material to the number of defects calculated using the NRT model. The same development cycle also improved displacements from elastic Coulomb scattering of charged particles by modifying the screening parameter, with particular relevance for low-energy heavy particles (Mokhov et al., 2014).

For nuclide inventory and residual-activity studies, MARS15 provides a user-friendly 3-step decay and transmutation workflow based on DeTra. The sequence is: run MARS to calculate nuclide production and stopping rates with interface files such as NUCLIDES; call built-in DeTra to solve the Bateman equations using transmutation trajectory analysis; and process the output to obtain specific activities and production rates. The resulting output can be used directly for residual dose-rate estimation. In later status reporting, residual activity can be handled through either built-in DeTra or the SandiaDecay library (Mokhov et al., 2014, Rakhno et al., 3 Oct 2025).

Residual-dose analysis was subsequently extended in two distinct directions. First, the 2025 status report describes a method for calculating the 3D spatial distribution of residual dose in a single computer run without an intermediate source, replacing an earlier emphasis on 2D contact residual-dose distributions. That method was validated against SINBAD data and FLUKA calculations, with agreement reported as excellent except for a copper sample at cooling times shorter than about two days (Rakhno et al., 3 Oct 2025). Second, the FermiCORD system was developed as a MARS15-based residual-dose code for arbitrary facility geometries. FermiCORD is a C++ add-on to Fortran-based MARS15 and implements a two-stage algorithm: MARS15 first computes radionuclide inventories and approximate residual dose on contact; then decay gammas are sampled from activated structures, propagated again in MARS15, and used to score residual dose at arbitrary observation points. To reduce spatial smearing in large structures, FermiCORD subdivides regions, often with Delaunay triangulation, and assumes that the probability density for radionuclide production is proportional to the residual dose on contact at a point. Benchmarking against DORIAN and CERF data showed reasonable agreement, with disagreement against CERF data never worse than about a factor of 2 (Grebe et al., 2016).

4. Geometry, beam-line modeling, and scoring infrastructure

Geometry handling is a major part of the MARS design. The 2012 development paper introduced a new ROOT-based geometry and visualization system permitting model creation from ROOT primitives, composite shapes, assemblies, three-dimensional visualization, and interoperability with other Monte Carlo systems, including Geant4. The paper explicitly states that ROOT models can be exchanged between MARS15 and Geant4, and that the MARS beam line builder was redesigned around the ROOT geometry package so that beam-line three-dimensional models can be generated automatically from optics files and element descriptions (Mokhov et al., 2014).

The later status report broadens this into a multi-modal geometry stack. MARS15 supports non-standard geometry with highest priority, implemented either as a Fortran file, a C++ file using ROOT geometry, or a GDML file. For simpler systems, the built-in Extended geometry constructive-solid-geometry option is intended for relatively simple or moderately complex arrangements, including beam lines without bending magnets and models with up to a few hundred objects. For more complicated systems, especially beam lines with horizontal and vertical bends, ROOT geometry is described as preferable. The same report adds a beam line builder based on MADX input, using the beam-line axis as the driver so that beam pipe, magnets, tunnels, penetrations, and related components follow the actual trajectory. It also introduces a tracker/stepper based on the PTC tracker, coupling precise analytical transport in vacuum inside the aperture with Monte Carlo transport in matter outside the aperture (Rakhno et al., 3 Oct 2025).

Scoring and post-processing infrastructure evolved in parallel. Two-dimensional distributions can be stored in HBOOK or ROOT, while three-dimensional scoring uses VTK and is visualized with ParaView. The status report stresses that the geometry model is independent of the mesh, enabling arbitrary scoring grids, including externally generated structured and unstructured meshes. A practically important workflow couples MARS with ANSYS: a ROOT geometry model is converted to STEP, meshed in ANSYS, exported in CGNS, used in MARS for absorbed-energy calculations, and then reused in ANSYS for thermal-mechanical studies. In the MiniBooNE target assembly example, energy deposition on the unstructured mesh produced high-quality results with only minor irregularities, while regular voxel scoring can generate much larger artifacts when multiple materials share a voxel near a boundary (Rakhno et al., 3 Oct 2025).

5. Validation and representative applications

The code’s application domain is strongly accelerator-centered, but its validation record spans both benchmark comparisons and realistic facility studies. In the context of Intensity Frontier machines, the 2012 development paper frames the upgrades as responses to the needs of ESS, FAIR, and Project X, where high-power interactions among targets, shielding, beam lines, and detectors require consistent treatment of transport, activation, and damage across disparate energy scales (Mokhov et al., 2014).

Independent cross-code comparison was reported for detector-background calculations at the 125-GeV Higgs Factory muon collider. After comprehensive tuning of geometry, beam parameters, and scoring definitions, MARS and FLUKA produced rather similar spatial distributions and energy spectra for photons, electrons/positrons, neutrons, and charged hadrons entering the detector. The paper’s summary result is that the average numbers of background particles simulated by the two codes agree within a factor of two. In the improved v7x2s4 shielding configuration, MARS predicts 1.5 to 3 times lower background load than FLUKA, which the authors attribute to a more finely tuned magnet-protection model in MARS at that stage (Mokhov et al., 2018).

Residual-dose applications further illustrate the code family’s role in maintenance and radiation-protection planning. FermiCORD was applied to the Mu2e Production Solenoid Hall, where it predicted higher doses in corners than an earlier approximate method because the full-source Monte Carlo treatment includes distributed activation in walls and other remote structures neglected by simplified source models (Grebe et al., 2016). The newer one-run residual-dose capability described in the 2025 status report is illustrated with the LBNF Hadron Absorber, under 120-GeV protons at 2.4 MW with 1 day cooling and a maintenance configuration involving seven concrete blocks removed, and with the PIP-II Instrumentation Cart at a 177. MeV linac location with a 0.73 W beam-loss rate over a 5-meter beam-pipe segment, 4 weeks of irradiation, and 24 hours of cooling (Rakhno et al., 3 Oct 2025).

6. Development trajectory and technical significance

The development history of MARS is characterized by continuity of physical scope together with repeated modernization of specific subsystems. The 2012 release already constituted a substantially expanded platform for medium-energy hadronic generation, low-energy proton nuclear interactions, keV-to-MeV electromagnetic showers, stopping powers in compounds, neutron DPA, nuclide inventories, and ROOT-based geometry (Mokhov et al., 2014). The 2025 status report shows that the same code line continues to evolve toward broader geometry interoperability, mesh-based scoring, updated low-energy nuclear data, and direct three-dimensional residual-dose computation, while remaining focused primarily on accelerator applications (Rakhno et al., 3 Oct 2025).

This development trajectory also shows a persistent architectural pattern: MARS15 serves as a transport-and-activation kernel around which higher-level workflows are organized. FermiCORD exemplifies that pattern for residual-dose estimation, using MARS15 in both source-generation and decay-gamma transport stages rather than duplicating transport functionality (Grebe et al., 2016). A plausible implication is that the code’s long-term utility derives not only from its physics content, but also from its ability to absorb specialized workflows without losing a unified transport framework.

Current-status reporting emphasizes ongoing software modernization in addition to physics development. Planned or ongoing work includes a more user-friendly GUI, a redesigned main window, better material-file management, removal of obsolete Fortran COMMON blocks, elimination of global static data, and improved modularity (Rakhno et al., 3 Oct 2025). In that sense, MARS remains both a mature Monte Carlo system and an actively maintained engineering code for shielding, activation, residual-dose planning, detector-background analysis, and coupled radiation–thermal–mechanical studies.