Repository navigation
Al K-beta line weight: exSpy’s 0.0132 against a measured 0.028–0.030 #66
Description
Finding
exSpy 0.3.2 ties Al Kβ at 0.0132 of Kα, and the kit copies that weight for parity. NIST EPQ gives 0.0315.
On the public Leeds pure-Al wedge (Snelson, Collins, MacArthur; DOI 10.5518/1826; CC BY 4.0; Titan at 200 kV, Super-X), fitted with the broadened continuum edge from #65:
- A free Kβ fits to 0.028, or 0.030 at 1.5575 keV.
- The data excess at 1.56–1.65 keV drops from +718 counts to −121.
- The 1.65–2.2 keV misfit falls to +5 %, and χ²/n to 2.35.
- A free Gaussian scanned over 1.54–1.63 keV finds its χ² minimum at 1.56–1.58 keV, which is the Kβ energy.
- Incomplete charge collection was tested and refuted. An extra continuum split fixes the real data but makes synthetic Si worse.
Why nothing shipped
The synthetic generators plant exSpy's 0.0132, so the synthetic truth is planted at the exSpy weight, not the physical one. Tying the fit at the EPQ value therefore breaks the synthetic tests. A free Kβ trades with the continuum on synthetic data: it reads 0.011 against 0.0132 planted.
Wanted
- The synthetic generators plant the physical Kβ/Kα weights from an EPQ table.
- An opt-in line-weight table option (
.epq), keeping the exSpy table as the default parity path. - Re-measure Si on the Leeds wedge and on the synthetic films, and decide the default from both.
- Check Mg and Si Kβ the same way.
Activity
Paullo9 commentedon Oct 31, 2026
Measured with the broadened continuum edge (#65). Real data: the public Leeds standards, Snelson, Collins, MacArthur, DOI 10.5518/1826, CC BY 4.0 (Titan 200 kV, Super-X; whole-map pools). Synthetic truth is planted by the generators, not measured; the synthetic fits are noise-free.
Free K-beta (best chi2 over a weight scan):
A free K-beta reads low on synthetic data because it trades with the continuum: 0.025 against 0.0315 planted. On that scale the Leeds 0.0255 matches EPQ for Al, and the 0.0185 favours EPQ for Mg. For Si, EPQ's 0.057 is not supported by MgO; Si K-beta sits at the Si K edge of the detector.
Decision: the default stays
.exspy..epqis opt-in throughFitPolicy.lineWeights. The generators plant EPQ by default (EDX_SIM_KBETA), and the committed fixtures stay at eXSpy. A mixed table (EPQ for Al and Mg) is the open question.