Appearance
Mapping elements with XRF means scanning a manuscript point by point with a focused X-ray beam and recording the characteristic fluorescence each spot emits, producing one greyscale image per element. The best-practice core is simple: lock your acquisition geometry, bracket every session with a reference standard, and store the full instrument state as paradata so any map can be reproduced and compared later. Everything below is the checklist that makes those three things hold across a whole collection rather than a single lucky scan.
What does macro-XRF mapping give you that imaging does not?
Multispectral imaging tells you where reflectance changes; XRF tells you what element sits there. That distinction matters when two inks look identical under visible and near-infrared light but were made from different vitriol sources. A macro-XRF (MA-XRF) scanner — a Bruker M6 Jetstream, an XGLab ELIO raster, or a lab-built motorised stage with a handheld head — moves a polycapillary-focused beam (typically 25-100 microns) over the page and bins each point's spectrum into element channels.
The output is a data cube: x by y by energy. You almost never look at the raw cube; you extract element maps (Fe, Cu, Pb, Hg, Ca, K, Zn) and reason about ratios.
How do you choose dwell time, step size and tube settings?
These three parameters trade scan time against signal and resolution. A defensible survey recipe for parchment:
| Parameter | Survey value | High-res ROI |
|---|---|---|
| Step size | 200-500 microns | 25-50 microns |
| Dwell per point | 100-200 ms | 300-500 ms |
| Tube voltage | 40-50 kV | 50 kV |
| Tube current | 200-600 microA | tuned to ~30% deadtime |
| Beam spot | 100 microns | 25 microns |
Total time scales with the inverse square of step size, so halving the step roughly quadruples the scan. A 30 by 40 cm folio at 300 microns and 150 ms is already a multi-hour acquisition — plan sessions accordingly.
What belongs in your XRF acquisition checklist?
Run this before every page, not just every session:
text
[ ] Tube: voltage, current, filter recorded
[ ] Geometry: working distance, beam spot, detector angle
[ ] Reference: standard glass (e.g. NIST 1412) run pre & post
[ ] Focus: surface in focal plane across full travel (check corners)
[ ] Deadtime under ~40% (reduce current if higher)
[ ] Map extent and step size logged
[ ] Dose check: no point dwelled beyond institutional limit
[ ] Paradata file written alongside the cubeThe reference-standard run is the single most-skipped step and the one that most often makes cross-session comparison impossible.
How do you process the cube into trustworthy maps?
Use PyMca or its batch engine for bulk fitting rather than simple region-of-interest sums, because ROI integration leaks counts between overlapping peaks (Pb-L and As-K, for example).
python
from PyMca5.PyMcaPhysics.xrf import FastXRFLinearFit
fit = FastXRFLinearFit.FastXRFLinearFit()
result = fit.fitMultipleSpectra(
y=cube, # shape (rows, cols, channels)
configuration=cfg, # .cfg with elements, geometry, tube
concentrations=False,
)
fe_map = result["parameters"][result["names"].index("Fe K")]Always batch-fit the same .cfg across the collection. Changing the energy calibration or the element list mid-project silently makes earlier maps non-comparable.
How do you compare inks across folios?
Single-element maps mislead because beam current and surface distance drift. Normalise. Plot ratios such as Cu/Fe and Zn/Fe per ink region and the batches separate:
python
import numpy as np
ratio = np.where(fe_map > threshold, cu_map / fe_map, np.nan)A scribe who switched ink supply mid-quire usually shows up as a step change in Cu/Fe between gatherings — invisible to the eye, obvious in the ratio map.
What are the most common quality failures?
Sum-peak and escape-peak artefacts masquerading as trace elements; defocus at the page edges flattening real signal; and undocumented current changes that turn an intensity comparison into nonsense. Treat any "new element" that appears only at high count rates as a pile-up artefact until proven otherwise.
Key Takeaways
- XRF maps element distribution, not visible ink — use it to answer "what is this made of," not "where is it."
- Lock tube voltage, current, geometry and filters, and record them as paradata for every scan.
- Bracket every session with a certified reference standard run before and after.
- Survey at 200-500 microns and 100-200 ms; reserve fine resolution for small ROIs because time scales as step squared.
- Batch-fit with one PyMca
.cfgso maps stay comparable across the collection. - Reason in element ratios (Cu/Fe, Zn/Fe), not single channels, to separate ink batches.
- Watch for sum-peaks, escape-peaks and edge defocus before claiming a trace element.
Frequently Asked Questions
What does an XRF map actually show on a manuscript?
It shows the spatial distribution of chemical elements — typically iron, copper, lead, mercury, calcium and potassium — measured from characteristic X-ray fluorescence. Each pixel is an estimate of how much of an element is present at that point, not an image of the visible ink.
What is a sensible dwell time and step size for parchment?
For survey maps, 100-500 microns step size with 100-300 ms dwell per point is a good starting compromise. Drop to 25-50 microns only over small regions of interest, because total scan time scales with the square of the resolution.
How do I keep XRF maps comparable across a whole collection?
Fix the geometry (tube voltage, current, working distance, filters), record it as paradata, and run a certified reference material at the start and end of every session. Comparable settings plus a reference standard are what make element ratios defensible later.
Can XRF mapping damage a manuscript?
Macro-XRF at heritage settings is non-contact and generally considered non-destructive, but prolonged dwell on one spot delivers a real radiation dose. Keep dwell times modest, never park the beam, and follow your institution's dose limits.
How do I tell two inks apart in an XRF map?
Compare element ratios rather than single channels. Two iron-gall inks may both show iron but differ in copper or zinc content; plotting Cu/Fe or Zn/Fe per region usually separates batches that look identical to the eye.
Do I need vacuum or helium purge for light elements?
If you care about elements lighter than potassium (e.g. silicon, aluminium, sulphur), yes — air absorbs their low-energy X-rays. For the iron-to-lead range that dominates manuscript inks and pigments, ambient air is usually fine.