A simulation explains something when a specified model connects its inputs, operations and outputs to the question being investigated. A plausible animation is insufficient. The calculation must implement the model accurately, and the model must be appropriate to the physical comparison.
Two different checks
Verification asks whether the implementation and numerical solution behave as intended. Validation asks how adequately the model represents the target system for a stated use. NASA's modelling standard distinguishes these activities and ties credibility to the intended application.
A verified simulation can still represent the wrong physics. A physically promising model can also be implemented badly. Separating the two questions helps locate an error instead of treating every disagreement as a failure of the underlying idea.
The following demonstration isolates a numerical question. Its reference is the continuous-time solution of the same nine-cell model. No experimental data are used, so agreement establishes a property of the calculation rather than validation of a material system.
A nine-cell diffusion model
Place nine cells on a periodic ring and initially put one unit of material in the central cell. Each cell exchanges material with its two neighbours according to duⱼ/dτ = uⱼ₋₁ − 2uⱼ + uⱼ₊₁. The time τ is dimensionless. Indices wrap around the ring.
The exact continuous-time dynamics preserve the total amount and keep every amount nonnegative. An explicit Euler approximation advances the amounts with
uⱼ(new) = λuⱼ₋₁ + (1 − 2λ)uⱼ + λuⱼ₊₁, λ = τ/N.
All new values are calculated from the same old array. N is the number of equal updates used to reach the target time. For 0 ≤ λ ≤ 1/2, the three coefficients are nonnegative and sum to one. Each new value is therefore a weighted average of old values.
Check a numerical diffusion calculation
Keep nine periodic cells fixed. Compare N Euler updates with the continuous-time solution of this same cell model. The graph's vertical axis rescales automatically.
Solid green: continuous-time referenceDashed purple: Euler approximation
Read the nine cell values
Conservation can hide a failure
Set τ = 4 and N = 4, giving λ = 1. The four Euler updates produce amounts 1, −4, 10, −16, 19, −16, 10, −4, 1 around the displayed ring. Their sum is still one. Several cells nevertheless contain negative material, which is impossible for the quantity being modelled.
Total conservation is only one check. The scheme conserves the sum because every outgoing contribution appears elsewhere in the periodic array, even when its coefficients no longer form a nonnegative average. A conserved total cannot certify the physical admissibility or numerical accuracy of every component.
The nonnegative-weight condition is sufficient for positivity in this model. It is not asserted to be the exact spectral stability boundary of every finite ring. Larger steps can produce negative values and growing numerical modes; the particular grid and update rule determine the detailed stability limit.
Compare with the same equation
The reference curve is calculated by diagonalizing the nine-cell periodic operator. Its mode k decays as exp[−4τ sin²(πk/9)]. Adding the nine modes reconstructs the amount in every cell. This reference keeps the spatial cells fixed while removing the Euler time approximation.
Increasing N at fixed τ tests time-step refinement. It does not test whether nine cells are enough to represent a continuous material. Spatial refinement would change the model, its length calibration and the relation between the coefficients and a physical diffusion constant.
The graph rescales its vertical axis to keep all selected values visible. Read the numerical error and axis labels when comparing settings: a visually similar curve on a different scale can represent a very different discrepancy. The reference and approximate values remain available as text.
Explanation needs a physical dictionary
To use this calculation for a real diffusion experiment, specify the cell length, time scale, boundary conditions, initial preparation and detector response. Compare predictions with observations that were not used to adjust every parameter. A good fit at one time need not establish the mechanism.
The book's physical dictionary and auditable-model programme apply this discipline to SMM/DTT. A finite generative model becomes informative when its update, completion and readout determine several outputs under the same calibration.
The ring demonstration is an independent teaching calculation. It does not derive diffusion from inner time. Its contribution is methodological: make the equation explicit, expose the approximation, check more than one invariant and separate numerical agreement from physical evidence.
Sources and further reading
- NASA, Standard for models and simulations, NASA-STD-7009B, for verification, validation and intended use.
- The periodic diffusion equation, Euler comparison and discrete-mode solution are fully specified teaching calculations in this essay.
- Mohamed Haj Yousef, One physical dictionary and A model that can be audited.

