Skip to content

Estimating Earth’s past field strength from individual sources in archaeological ceramics using quantum diamond microscopy

FieldValue
JournalCommunications Earth & Environment
Published2026-06-09
AuthorsUalisson Donardelli Bellon, Gelson F. Souza-Junior, Wyn Williams, Adrian R. Muxworthy, Leonardo Uieda
DOI10.1038/s43247-026-03725-3

Quantum diamond microscopy recovered an estimated palaeointensity of about 44 µT without heating the ceramic. The result agreed closely with an independent classical heating Thellier-Thellier estimate from a sister sample. The protocol tracked individual magnetic anomalies through alternating-field demagnetisation and anhysteretic remanent magnetisation. The authors describe this as the first palaeointensity recovery based on many individually resolved anomalies.

Specimen RSLG1-03 came from the Jesuit Missions in Brazil and was interpreted to acquire thermoremanent magnetisation during kiln firing. Its unoriented thin section retained pristine natural remanent magnetisation. It dates from 1657 to 1687 AD.

Bulk measurements used a 2G RAPID superconducting rock magnetometer in a magnetically shielded room at Harvard University. Alternating-field demagnetisation covered 0 to 100 mT, and anhysteretic remanent magnetisations were then acquired at the same field steps along positive Y with a 50 µT laboratory bias.

The quantum diamond microscope, hereafter QDM, covered two regions selected for pronounced dipolar anomalies identified in an earlier proof-of-concept study. Each region measured 1410 µm × 2256 µm with 2.35 µm grid spacing. The sensor stood 1 to 5 µm above the sample. A 0.9 mT acquisition bias was periodically reversed, leaving an effective net bias of 400 nT.

After ARM acquisition, pulse magnetisation produced a 2.70 T isothermal remanent magnetisation along positive Y and a 0.15 T IRM along negative Y. These states provided reference points for tracking the same anomalies across treatments. Affine registration corrected manual repositioning using LED images, ORB feature matching and RANSAC outlier rejection.

Automatic detection found 1,178 anomalies across the two maps. The MAPI workflow estimated source depth with modified Euler deconvolution, then inverted each region of interest for three magnetic moment components. The 2.70 T IRM map defined the source regions, which were propagated to the other treatments.

Only inversions with R² > 0.9 in both IRM states were retained. Sources changing direction by less than 90° between the two IRMs were also discarded because the method inferred coercivities beyond the AF treatment window. The results describe these as high-coercivity sources with Bcc above 150 mT.

Inverted moments ranged from 3.1 × 10⁻¹⁵ to 1.8 × 10⁻¹² A m². The median was 2.8 × 10⁻¹⁴ A m², and the upper quartile was 7.1 × 10⁻¹⁴ A m². The median estimated zc, measured from sensor to particle, was -8.51 µm. Seventy-five percent lay within 12.36 µm of the surface.

Under a magnetite saturation assumption, the moments imply equivalent spherical diameters of about 0.2 to 2 µm, with a median near 0.5 µm. Accounting for reduced magnetisation in vortex grains shifts the interpretation toward 0.5 to 1 µm and mostly single-vortex or multi-vortex grains. These remain model estimates because the inversion recovers dipoles rather than particle geometry.

Direction checks reveal agreement and anisotropy

Section titled “Direction checks reveal agreement and anisotropy”

The filtered QDM directions closely matched the thin section’s characteristic remanent magnetisation from whole-sample measurements. Fisher statistics and cumulative addition both placed the inverted mean near the bulk ChRM.

The bulk NRM moved from steep inclinations toward the horizontal during AF demagnetisation. The authors link this to a viscous remanent magnetisation component carried by large multidomain grains. QDM directions were clustered near the ChRM from the NRM stage onward. The authors propose that strong non-dipolar anomalies from these grains fail the inversion filters. Deep multidomain grains can still look dipolar, so filtering cannot guarantee stability.

ARM directions moved toward the applied field but did not align fully at 100 mT. The QDM result showed the same trend, which the authors link to remanence anisotropy.

Pseudo-Thellier recovery depends on inversion quality

Section titled “Pseudo-Thellier recovery depends on inversion quality”

The pseudo-Arai relation compared normalised NRM remaining with ARM gained at the same AF step. The nominal fit covered a 30 to 70 mT interval containing at least seven points. Monte Carlo perturbations combined uncertainty from the fitted remanence curves with directional dispersion. Orthogonal-distance regression refitted every synthetic segment.

Absolute intensity used the 50 µT ARM bias and a calibration factor c = 2.2 ± 0.59 for fired ceramics and source clays. The calibration factor contributes substantial uncertainty to every absolute estimate.

Inversion criterionMonte Carlo median slopePalaeointensityPaper comparison
R² > 0.8-1.67about 38 µTWhole thin section over a similar range: -1.62, about 37 µT
R² > 0.9-1.95about 44 µTHeating result for sister sample: about 44 µT before anisotropy correction

The R² > 0.9 nominal slope was -2.53, but the Monte Carlo median was -1.95. ARM variability at 60, 70 and 80 mT increased uncertainty and made palaeointensity sensitive to the quality threshold. For R² below 0.7, Monte Carlo estimates ranged from 25 to 35 µT.

The Figure 4 error bars show the c contribution and remain identical across thresholds. Their size therefore does not display the separate effect of source selection.

Nanotomography explains the excluded signal

Section titled “Nanotomography explains the excluded signal”

Focused ion beam-scanning electron microscopy examined a strongly non-dipolar region excluded by QDM inversion. Reconstruction used 947 serial images at 10 × 10 nm pixels and 15 nm slice spacing. Segmentation isolated 1,880 dense particles, interpreted as magnetic iron oxides.

Detected sizes extended from the 0.015 µm resolution limit to 3 µm. Mean maximum elongation was about 0.170 µm, and most particles fell between 0.10 and 1 µm. Scanning electron microscopy with energy-dispersive X-ray spectroscopy found anhedral iron oxides around 5 µm, with some as small as 1 µm, and suggested titanomagnetite-series minerals.

Finite-element models covered 584 particles within 5 µm of the surface, sized 50 nm to 1 µm. Each was initialised in 50 random magnetic states and minimised to local energy minima using magnetite parameters at 20 °C.

Larger particles produced vertical fields up to two orders of magnitude stronger than smaller modelled particles. They accounted for nearly 90% of the total response relative to smaller grains. The authors interpret this as dominance by large, less stable grains. They also note that filtering loses information from stable sources.

R² measures dipole fit, not magnetic stability or recorder quality. Nanoscale tomography would be needed to verify that retained anomalies contain only stable carriers. The dipole model cannot resolve strongly non-dipolar anomalies, and buried multidomain grains can appear dipolar.

Only two maps limited the source population. Manual height changes could alter moment magnitude, while uneven source retention affected reconstructed curves. The authors cite a 20-map workflow that recovered NRM from ceramic and basalt. Each map covered 1410 µm × 2256 µm at 2.35 µm spacing.

With sufficient coverage, cumulative inverted moments could replace Fisher unit vectors and reduce bias from strong sources. Tracking each anomaly across AF and ARM stages can also flag unstable behaviour before intensity fitting.

Anisotropy remains a limitation. The ARM had not aligned with the applied field at 100 mT, and tomography showed vertical clustering of particle axes. The authors state that anisotropy corrections are likely required, but they did not apply the related ceramic selection criteria.

The calibration factor remains empirical. Micromagnetic modelling of particle size and shape could support a correction factor for vortex-state magnetite, but no analytical formulation exists. For microscope design, the results imply that standoff and map coverage control apparent dipolar shape and source retention.

Quantum diamond microscopy requires a diamond sensing element, and this study placed the sensor 1 to 5 µm above a ceramic thin section. 6CCVD offers MPCVD-grown single-crystal and polycrystalline diamond plates and wafers up to 125 mm, with single-crystal material polished to Ra below 1 nm. The diamond materials catalog lists the formats, while engineering support can address sensor integration and mounting requirements.

View Original Abstract

Abstract When archaeological ceramics are fired in a kiln and cool, microscopic magnetic minerals can acquire a magnetisation aligned with Earth’s magnetic field, preserving information about its past strength. Conventional palaeomagnetic methods recover this signal from whole samples, but bulk measurements can mix reliable and unreliable recorders, while heating-based experiments may alter the minerals being measured. Here we present a non-heating protocol using quantum diamond microscope measurements of individual magnetic sources in a thin section of an archaeological ceramic. We use magnetic field maps to isolate near-surface magnetic anomalies, recover their magnetic moments, and track their behaviour during controlled laboratory experiments. Selectively recovered directions closely match whole-sample measurements from a cryogenic rock magnetometer. For well-separated, dipole-like sources that pass quality and behaviour filters, estimates of Earth’s past magnetic field strength agree with independent measurements from related samples. These results demonstrate that magnetic microscopy provides a promising route to high-precision micropalaeomagnetic analysis.