Saved in:
| Hovedforfatter: | |
|---|---|
| Format: | Recurso digital |
| Sprog: | engelsk |
| Udgivet: |
Zenodo
2026
|
| Fag: | |
| Online adgang: | https://doi.org/10.5281/zenodo.20042320 |
| Tags: |
Tilføj Tag
Ingen Tags, Vær først til at tagge denne postø!
|
Indholdsfortegnelse:
- <p>Experimental data for atomic masses (AME2020), ionization energies (NIST), covalent bond energies, semiconductor band gaps, DNA base pairs, protein interactions, fundamental biochemical reactions, as well as the training of a ternary neural network, admit a compact arithmetic representation of the form:<br> $$E = \Lambda \cdot 4^{m} \cdot \sum_{k=0}^{14} \varepsilon_k \beta_k$$<br> where $\Lambda$ is a scale constant (7.726 MeV for masses, 5.950 eV for chemistry and biology), $m$ an integer, $\varepsilon_k \in \{-1,0,1\}$, and $\beta_k$ are 15 universal constants of empirical origin. For 5811 ionization energies, the mean relative error is 0.00065% and 78.5% of the points are predicted with an error better than 0.001%. The mean relative error for 295 nuclear masses is 0.0287%. The $\beta_k$ numerically coincide with 15 of the 48 non‑degenerate roots of the exceptional lattice $\Lambda_{72}$, suggesting a deep geometric link that remains to be elucidated. </p>