4.2. MAVRIC-Shift: Radiation Shielding Analysis Sequence with Shift

4.2.1. Introduction

MAVRIC-Shift integrates the Shift fixed-source Monte Carlo solver into the MAVRIC shielding sequence. MAVRIC continues to prepare problem-dependent cross sections and, for CADIS and FW-CADIS, uses Denovo to construct an importance map and biased source. Shift performs the final Monte Carlo transport calculation in place of Monaco. Main enhancement with MAVRIC-Shift is support for parallel processing which is enabled for both Deonovo and Shift.

This section describes the MAVRIC-Shift implementation in SCALE 6.3.

4.2.2. Input requirements

Input requirements for MAVRIC-Shift are designed to match MAVRIC with a few exceptions that are mentioned below. Users should be able to change the sequence name from =mavric to =mavric-shift and tend to any unsupported features and/or parameters to convert. their model to MAVRIC-Shift as seen in the example below.

Select the Monte Carlo engine with the sequence name:

Example 4.2.1 MAVRIC sequence selection

=mavric
My shielding calculation with Monaco

=mavric-shift
My shielding calculation with Shift

MAVRIC-Shift retains the MAVRIC block layout and most input syntax. Changing the sequence name is sufficient only when every source, tally, response, and importance map option used by the model is supported by the Shift interface. An unsupported filter or control can change the physical problem; do not remove one only to make the input execute. It is highly recommended to use Fulcrum to validate MAVRIC-Shift models when converted from MAVRIC.

The shared definitions, sources, tallies, parameters, and biasing blocks are defined in Sect. 8.2. MAVRIC-specific importance map input is defined in Table 4.1.4 and Table 4.1.5.

4.2.3. Sequence capabilities

MAVRIC-Shift supports the principal MAVRIC calculation modes shown in the table below.

Table 4.2.1 Supported MAVRIC-Shift capabilities

Capability	MAVRIC input	MAVRIC-Shift behavior
Analog forward transport	Omit the importanceMap block	Shift performs the final Monte Carlo calculation without a Denovo importance map calculation.
CADIS	importanceMap and adjointSource	Denovo calculates the adjoint flux; Shift uses the biased source and weight windows.
FW-CADIS	respWeighting or fluxWeighting	Denovo calculates the forward and adjoint fluxes; Shift uses the resulting biased source and weight windows.
Multigroup and continuous-energy transport	Input library name, library, or ceLibrary	Both final-transport energy modes are supported.
Neutron, photon, and coupled transport	neutron and photon	Neutron-only, photon-only, and coupled neutron–photon transport are supported.
Region tallies	regionTally	Flux, energy bins, supported responses, uncertainty, and figure of merit (FOM) are reported.
Mesh tallies	meshTally, gridGeometryID, and cylGeometryID	Flux, energy bins, supported responses, uncertainty, FOM, and *.3dmap postprocessing are supported.
Response definitions	response and doseData	User-defined responses and supported dose and reaction responses are supported.
Source material filter	mixture in a source definition	The source can be restricted to a material mixture.
Adjoint source	adjointSource, boundingBox, and locationID with responseID	The adjoint source can be defined over a volume for CADIS or FW-CADIS, but point adjoint sources are not supported.
Parallel final transport	Execution option; no MAVRIC input keyword	Shift can use MPI when the installed SCALE build includes MPI.

For CADIS and FW-CADIS, Denovo performs the deterministic calculation and Shift consumes the resulting source biasing and weight window data. Supported tally summaries include the mean, standard deviation, relative uncertainty, and figure of merit (FOM).

4.2.4. Limitations relative to MAVRIC-Monaco

MAVRIC-Shift does not support every Monaco input. Use MAVRIC-Monaco, or revise the model, when it requires an entry listed in the table below. It is highly recommended to use Fulcrum for validation of models when converting a MAVRIC-Monaco model to MAVRIC-Shift model to identify unsupported features.

Table 4.2.2 MAVRIC-Monaco input not supported by MAVRIC-Shift

Capability	Monaco input	MAVRIC-Shift behavior and guidance
Point-detector tallies	pointDetector	Unsupported. Use MAVRIC-Monaco when a point estimate is required.
CAAS KENO-VI fission-source import	meshSourceFile, fissPhotonZaid, fissions, and nu-bar	Unsupported. Use MAVRIC-Monaco or consider using a Watt fission distribution.
Tally time bins	timeBoundsID	Unsupported. Use MAVRIC-Monaco for time-dependent tally bins.
Source unit and region filters	unit and region	Unsupported. The source mixture filter is supported.
Source spatial distributions and scaling	xDistributionID, yDistributionID, zDistributionID, rDistributionID, xScaleDist, yScaleDist, zScaleDist, and rScaleDist	Unsupported. Use only source distributions accepted by MAVRIC-Shift.
Source direction, axis, rotation, and resampling	dDistributionID, dScaleDist, direction, cylinderAxis, rotate, and allowResampling	Unsupported. Use MAVRIC-Monaco if these cards define the physical source.
Explicit source-sampling bias	biasedStrength	Unsupported. Use MAVRIC-Monaco if needed
Mesh-source export	meshSourceSaver	Unsupported. MAVRIC-Shift cannot save the defined source as a Monaco mesh-source file.
Region tally filters	mixture	Unsupported. Use an unfiltered region tally or MAVRIC-Monaco.
Mesh tally filters	unit, region, and mixture	Unsupported. Use an unfiltered Cartesian mesh tally or MAVRIC-Monaco.
Mesh tally detail modes	weightless and saveRespDetails	Unsupported. Use the available Shift mesh tally output or MAVRIC-Monaco.
Adjoint-source location, unit, and region filters	locationID, unit, and region	Unsupported. Define a supported boundingBox adjoint source or use MAVRIC-Monaco.
Adjoint-source material filter	mixture	Unsupported.
General library-derived responses	MT, nuclide, and material	Partially supported. Material-based responses are not supported. Use a supported response, define the response explicitly, or use MAVRIC-Monaco.
Reuse of deterministic flux files	forwardFluxes and adjointFluxes	Unsupported. MAVRIC-Shift must calculate the required Denovo fluxes.
Importance-map source and weight overrides	subCells, mapMultiplier, and saveExtraMaps	Unsupported. Use the MAVRIC-Shift importance map construction.
Legacy Denovo multigroup solver selectors	partUpscatter, krylovType, eigenSolver, multiGSolver, mgSettings, upGroupSolver, maxItersMG, and toleranceMG	Unsupported. Remove the legacy selectors or use MAVRIC-Monaco.
User-supplied Monte Carlo biasing	read biasing	Unsupported. Use analog transport, CADIS, or FW-CADIS.
Run-time, particle-age, and weight-window controls	maxMinutes, nMaxAge, pMaxAge, allowShortImpMap, noCheckAtBirth, and voidAllRegions	Unsupported. Select histories and supported stopping controls instead.
Geometry map generation controls	make3dMaps, make3dMap, makeCylMaps, and makeCylMap	Unsupported. This does not prevent *.3dmap processing of supported Cartesian mesh-tally results.
Creating chart files	makeChart and makeCharts	Unsupported. Use MAVRIC-Monaco if needed.
Response output control	lessOutput	Unsupported.
Distribution fine-tuning and controls	trueCDF, biasedPDF, biasedCDF, weight, importance, runSampleTest, and runSampleTests	Unsupported. Use MAVRIC-Monaco if needed.
Monaco six-check tally suite	Automatic output; noStatChecks disables it in Monaco	Not implemented. MAVRIC-Shift reports tally uncertainty and FOM.

MAVRIC-Shift calculates and reports FOM values. It does not reproduce the six statistical checks printed for Monaco point and region tallies.

4.2.5. Transport engine selection

As of SCALE 6.3.4, we recommend using MAVRIC-Shift when its supported source and tally definitions cover the analysis and parallel Shift transport is useful. Use MAVRIC-Monaco when the model depends on a missing feature, especially point detectors, time dependence, or the SCALE 6.3 CAAS mesh source workflow.

Solver comparisons require equivalent physical models. Use the same nuclear data, geometry, material temperatures, source strength and distributions, responses, tally bounds, variance reduction targets, and stopping criteria. The two Monte Carlo engines do not share random number streams, so tally values should be compared statistically, not for exact equality.

4.2.6. Parallel execution

Monaco particle transport in MAVRIC is serial. Denovo and Shift final transport can run in parallel with an MPI-enabled SCALE installation. The code output reports used executable, process count, histories, CPU time, wall time, uncertainty, and FOM for any performance comparison. A serial SCALE installation exercises MAVRIC-Shift functionality but cannot validate MPI scaling. Users can execute a parallel job by providing -N flag, followed by number of processors desired and optionally provide a machine file via -M.

Note

An analog model without an importanceMap block can be run in parallel as:

scalerte –N 4 –M mach mavric-shift.inp

Important

If a variance-reduction method is enabled via importanceMap then number of processors should be the product of xblocks, yblocks, and numSets when running a MAVRIC-Shift model n parallel.

One SCALE 6.3 development study used a simplified continuous-energy spent-fuel cask model. In the high-fidelity cases, all reported relative standard deviations were below 1 percent. CADIS wall time was 380 minutes with serial Monaco and 18 minutes with 16-process Shift; FW-CADIS wall time was 274 and 24 minutes. These results characterize one model and computing system, not a general speedup.

4.2.7. MAVRIC Utilities

MAVRIC-Shift-generated mesh tallies do not contain geometry and material information stored in *.3dmap files and may not suitable to perform mesh operations requiring such information. As a result, not all utilities will be compatible with MAVRIC-Shift-generated *.3dmap files, please refer to individual utility documentation.

4.2.8. Output and review

MAVRIC-Shift writes the standard SCALE output and message files, final region and mesh tally summaries, detailed supported tally files, and supported Cartesian *.3dmap results. Shift table layout, diagnostic text, timing, and auxiliary output will differ from Monaco output.

For each production calculation, verify:

• normal SCALE and sequence termination and zero reported errors;

• the intended data library, particle modes, histories, and process count;

• source, response, tally, and importance map summaries;

• tally means, standard deviations, relative uncertainties, and FOM; and

• batch stability or other convergence evidence appropriate to the result.

For a parity study, preserve the matched inputs, commands, output and message files, process counts, timings, and the statistical comparison. Two estimates \(x_1 \pm \sigma_1\) and \(x_2 \pm \sigma_2\) are consistent at a chosen multiplier \(k\) when

(4.2.1)\[\left|x_1-x_2\right| \leq k\sqrt{\sigma_1^2+\sigma_2^2}.\]