Verification vs Validation
Verification asks whether the equations are solved correctly; validation asks whether the right equations were chosen for reality.
Two Different Questions
Scientific computing draws a sharp line between two activities that casual usage often conflates. Verification asks whether a code solves its governing equations correctly. Validation asks whether those equations describe the physical world well enough for the intended use. A code can be perfectly verified and still invalid, because it faithfully solves the wrong model.
A compact slogan credited to Roache captures it: verification is solving the equations right; validation is solving the right equations. Verification is a mathematics and software problem, settled by comparison to exact solutions and by convergence tests. Validation is a physics problem, settled by comparison to experiment.
Why the Order Matters
Verification comes first. Comparing an unverified code to experiment mixes two error sources: numerical error from the discretization and model-form error from the physics. If the two happen to cancel, agreement with data is meaningless and will not survive a change of mesh or regime. Only after numerical error is bounded can a discrepancy with experiment be attributed to the model.
What Each Produces
- Verification produces evidence about numerical accuracy: observed order of accuracy, discretization error estimates, iterative-convergence residuals.
- Validation produces evidence about physical fidelity: comparison metrics between simulation and measurement, with uncertainty bars on both sides.
In Practice at Kronos
The Hyperion breeder and the burner designs are simulation and design work, not built hardware, so their published claims rest on verified solvers first. Physical validation against operating fusion devices and dedicated experiments is a separate, ongoing obligation that no amount of code verification can substitute for. The design record keeps the two ledgers distinct so a reader can see which claims are mathematical and which are physical.
The practical discipline: never report a validation comparison without first reporting the verification evidence that bounds the numerical error contaminating it.