Attenuation Calculator: polychromatic beams and quasi-monochromatic sources for X-ray CT energy selection

A single-file, browser-based X-ray attenuation calculator for micro-CT, industrial CT, and X-ray imaging design.
Computes monochromatic (2–500 keV) and polychromatic (20–500 kVp) attenuation, optical depth, and transmission for 450+ species, minerals, battery materials, elements, polymers, biological tissues, ceramics, metals, and alloys. Polychromatic mode uses SpekPy v2 (spekcalc) tungsten-anode tube spectra with up to four user-defined filters (Be, Al, Cu, Mo, Sn) and reports effective mu, half-value layer, effective energy, and beam-hardening factor. Custom energy grid for regular-spaced data export. NIST-verified mass attenuation data via xraydb. Runs entirely in the browser, no server or install required.

### Fixed
- **Cobalt was parsed as carbon + oxygen when regenerating sub-10 keV mass-attenuation data.** The formula normaliser in `scripts/extend_species_low_energy.py` contained a typo-correction rule that uppercased a lowercase `o` appearing between a letter and a digit. The rule was intended for malformed oxide formulas (`Fe2o3` → `Fe2O3`) but also matched the second character of the element symbol **Co**, rewriting `Co0.15` to `CO0.15`. Cobalt was therefore replaced by carbon plus oxygen wherever it appeared in a fractional-stoichiometry formula.
- **Affected species (5 of 450):** NMC111, NMC622, NMC811, NMC955 and NCA — the layered-oxide lithium-ion cathodes. All five had their 2–10 keV μ/ρ values computed with carbon standing in for cobalt, and the **Co K-edge at 7.7089 keV was absent from their tables entirely**. Deviations from correct values ranged from ~10% to ~23% (median, per species) and up to 53% in the 7.71–8.33 keV window between the missing Co edge and the Ni edge.
- **Not affected:** all other 445 species; every value at 10 keV and above (those come from the original tabulated data and were never regenerated by this script); all filter tables; all spectral data; the custom-compound builder; and every computational routine. Monochromatic and polychromatic engines are unchanged.
- The same rule would have corrupted Ho, Mo, No and Po in fractional formulas. No species in the current table triggered those cases, but the fix covers all five symbols.

### Changed
- `fix_formula()` in `extend_species_low_energy.py` now skips the uppercase-O substitution when the lowercase `o` completes a genuine two-letter element symbol (Co, Ho, Mo, No, Po). Genuine typos such as `Fe2o3` are still corrected.
- The 2–10 keV tables for the five affected species were regenerated with the corrected parser. Both K-edges now present: Co at 7.7089 keV and Ni at 8.3328 keV, each with edge-adjacent sample points at E ± 1 eV.

### Verified
- All 3,954 sub-10 keV comparisons across every parseable species now agree with xraydb to within 0.010% (mean 0.000%).
- External accuracy unchanged: xraydb agrees with the NIST X-ray mass attenuation tables across 2–500 keV for liquid water, polystyrene and elemental iron — mean absolute deviation 0.02%, maximum 0.05%, at the four-significant-figure resolution limit of the NIST tabulation.
- Corrected NCA edge structure: 1.78× jump at the Co K-edge (15 at.% Co), 3.64× at the Ni K-edge (80 at.% Ni).
- Byte-level diff confirms exactly 5 of 455 species records changed; no other content altered.
- JS syntax-check clean (`node --check`).

### Impact on published results
- Any figure, table or exported dataset using NMC or NCA **below 10 keV** should be regenerated. Results at 10 keV and above are unaffected, as is all polychromatic output at every tube voltage (the spectra carry negligible fluence below 10 keV after filtration).