Independent checks, built in from the start.

The habit

Six of the seven papers he published between 2013 and 2016 are comparisons: the same physical problem solved by two or more independent codes, or by theory and a code, with the differences explained. A reactor model is only as good as the code, the nuclear data and the geometry behind it, and the cheapest way to catch an error in any of the three is to build the model twice and see whether the answers agree.

What he has checked

Coupled-reactor theory against simulation. His reactor design depends on the theory of two neutronically coupled cores. He tested that theory on the Deuterium Critical Assembly, a two-region experimental assembly, by computing multiplication factors and coupling coefficients as the water levels in each core changed: first with MCNP5, then with both MCNP5 and Serpent. The results agreed, which established Serpent as valid for multipoint coupled-core calculations.

MCNP against Serpent on a power reactor. On a full CANDU 6 core he compared MCNP6.1 and Serpent for the multiplication factor, radial and axial flux, and power distribution. The codes agreed closely, and Serpent was the faster of the two.

Both against the industry code. He then validated whole-core burnup in MCNP6 and Serpent against WIMS-AECL, the lattice code used by the CANDU industry. K-effective against burnup agreed across the core’s life. The benchmark is still cited: a 2025 paper on using SMR spent fuel in CANDU reactors cites it.

A fast depletion code against the references. With the University of West Bohemia team he benchmarked UWB1, a depletion code that takes about ten seconds per depletion step, against WIMS, Serpent and MCNP6 on the 37-element CANDU bundle. That result lets burnable-absorber designs be screened quickly, then confirmed with the slower reference codes.

Why a second code matters

Verification and validation is how a model earns trust with a regulator, a design team or a reviewer. His doctoral thesis follows the same order: verify the theory, validate the codes on a known reactor, and only then design something new. In June 2026 he completed the OECD Nuclear Energy Agency’s Introduction to OpenMC, the open-source code now common in small-modular-reactor work.