Magnetic Microscopy
in Rock Magnetism

2026 IRM summer school

A bulk moment hides the grains that carry it

Bulk magnetometry measures:

\[\mathbf m = \int_V \mathbf M(\mathbf r)\,dV\]

  • It typically operate at the cm scale in rock magnetism studies.

  • Magnetic moment is typically indirectly measured.

Magnetic microscopy measures:

  • Stray magnetic field, flux, or field gradients outside magnetic grains

  • Local magnetization direction or magnetic-domain structure within, across, or at the surface of individual grains

  • The instrument measures field-, force-, phase-, or polarization-based contrast; magnetization (moment) is usually inferred through calibration or inversion.

  • The scale range from sub-mm to nm.

Every magnetic measurement begins with a forward problem

The sample contains magnetization, but most microscopes measure a field, force, phase, resonance, or polarization after geometry has filtered it.

\[\mathbf M(\mathbf r) \longrightarrow \mathbf B(\mathbf r,z) \longrightarrow s(\mathbf r) + \epsilon\]

Direct-ish contrast

XMCD, SEMPA, and Kerr microscopy show which way the magnetization points, but only for the part they can see—usually the sample surface or an average through its thickness.

Stray-field contrast

SQUID, Hall, MFM, NV, and Bitter imaging respond to the external field, its projection, or its gradients. Standoff suppresses short wavelengths.

A susceptibility bridge measures the response to a small AC field

Drive and balance

An AC current creates a small oscillating field. The sample responds with \(M(t)=\chi H(t)\) and changes the flux in one side of a balanced pickup bridge.

ReadoutLock-in detection separates the in-phase and out-of-phase voltage at the drive frequency.
  • Measures reversible, low-field response.
  • Frequency dependence can reveal relaxation time and thus grain size.

A VSM turns controlled motion into an induced voltage

Bulk moment along one axis

The specimen vibrates between pickup coils while an applied field sets its magnetic state.

ReadoutThe changing magnetic flux over the pick up coils produces a voltage, \(V=-d\Phi/dt\), proportional to the sample moment after calibration.
  • Measure hysteresis, backfield curves, and FORCs.
  • The result is the (calibrated) moment of the bulk specimen.

The JR-6 turns rotation into \(d\Phi/dt\)

Rotation encodes direction as phase

A defined-size specimen spins at constant \(\omega\) inside paired pickup coils, making the coil flux oscillate.

ReadoutFor a centered specimen, \(\Phi_{\rm link}(t)\!\approx\!K m_\perp\cos(\omega t-\alpha)\), so \(V=-d\Phi_{\rm link}/dt\). Fourier amplitude and phase give two components transverse to the spin axis.
  • Reorient the specimen to recover the third component.
  • Irregular shape or offset changes the coil coupling and waveform.
Measuring principle, nominal specimen geometry, and accuracy caveats: AGICO JR-6A / JR-6 User Manual v3.2 (2022). Animation is a reciprocity-based teaching model, not instrument CAD.

A SQUID converts flux into superconducting phase

\[I = I_c\sin\delta, \qquad \Phi_0 = \frac{h}{2e}\]

Quantum tunneling

Cooper pairs tunnel through Josephson junction barriers. Current depends on the phase difference of two superconducting condensates.

Flux-locked readout

Flux in the pickup loop changes SQUID voltage. Feedback cancels the change, and the feedback signal becomes a linear flux measurement.

Josephson phase relation: NIST Quantum Locking Ranges.

Flux shifts phase; bias produces voltage

The weak links make flux observable

The two paths enclose flux, so identical junctions acquire different gauge-invariant phase drops.

\[ \begin{aligned} \delta_1-\delta_2 &= 2\pi\frac{\Phi}{\Phi_0} \\[3pt] I_{c,\mathrm{SQUID}} &= 2I_0\left|\cos\!\left(\pi\frac{\Phi}{\Phi_0}\right)\right| \\[3pt] V &= \frac{\hbar}{2e}\frac{d\delta}{dt} \end{aligned} \]
Two current componentsThe bias splits through both arms: \(I_1=I_B/2-I_{\rm circ}\) and \(I_2=I_B/2+I_{\rm circ}\). The bias term points the same way; the circulating term points oppositely. Above \(I_c\), the phase runs and voltage appears.

A U-channel rock magnetometer couples specimen flux into a SQUID

Full vector moment

The sample moves through axial and transverse superconducting pickup coils. Their response functions separate \(M_x\), \(M_y\), and \(M_z\).

ReadoutPickup current couples flux into a DC SQUID. A flux-locked loop supplies the linear feedback voltage.
  • Extremely sensitive remanence measurement.
  • Translation gives a finite sensor-response curve.
  • The output is a bulk vector moment.
Vector pickup geometry: Jackson et al. (2010). Modern U-channel implementation: MARUM 2G Enterprises system.

An MPMS rejects background flux with a reverse-wound gradiometer

One axial moment component

The outer single turns are \(-1\) and \(-1\); the two central turns are \(+1\) and \(+1\). Their symmetric signed sum rejects uniform fields and linear gradients, but not a nearby dipole.

Two motion modesDC: fit SQUID signal versus a long translation. VSM: vibrate at the center and lock in the periodic signal.
  • The U-channel instrument instead combines transverse saddle pairs with circular axial coils for vector remanence.
Pickup geometry and sample-shape context: Jackson, Sølheid & Marvin, IRM Quarterly 11(4), 2001–2002. DC fitting and VSM lock-in operation: Quantum Design MPMS 3 User’s Manual.

Scanning SQUID microscopy measures flux with exceptional sensitivity

How it works

A lithographic pickup loop rasters above the sample. Coupled Josephson junctions convert loop flux to a voltage held linear by a flux-locked loop.

CollectFlux on a rectangular grid plus height, background, and calibration scans; convert to \(B_z\) only with the loop response.
  • Excellent for weak remanence.
  • Loop diameter and standoff set resolution.
  • Cryogenic sensor, not necessarily cryogenic sample.
Geological thin-section application: Mervine et al. (2017). Synthetic data.

Cooling through the Verwey transition creates monoclinic twins

A first-order structural transition

Near 120–125 K, magnetite changes from cubic \(Fd\bar{3}m\) to monoclinic \(Cc\). Charge/orbital order, strain, and magnetic anisotropy change together.

Cooling sequenceMonoclinic nuclei grow; differently oriented variants meet and form ferroelastic twin walls.
  • Variants inherit parent cubic \(\langle100\rangle\) axes.
  • Twin walls can redirect or pin magnetic domains.
Transition symmetry and twinning: Kasama et al. (2010) and Lindquist et al. (2019). Low-temperature charge order: Senn, Wright & Attfield (2012). Distortion is exaggerated.

Measurement workflow

Source
Magnetic stateremanent or induced moment
Spatial structuredomains and domain walls
Magnetic carriergrains, phases, and inclusions
Coupling
Fieldfield, flux, or field gradient
Mechanicalforce, torque, or deflection
Wave or spinphase, Zeeman shift, dichroism
Readout
Electricalvoltage or current
Opticalintensity or polarization
Mechanicalcantilever motion
Electronintensity or phase
Inference
Direct mapcontrast, field, or gradient
Calibration / inversionphysical units or source magnetization
Confidenceresolution, noise, and uncertainty

Instrument families differ in what they actually measure

Family and examples Raw signal Best suited to
Electron microscopyLorentz TEM · holography · Differential phase contrast Electron intensity, phase, or diffraction shift Nanometre-scale domains and walls
X-ray and spin electronXMCD · ptychography · SEMPA Dichroism or electron-spin polarization Element-specific or vector contrast
Flux and transportSQUID · Hall Pickup-loop flux or Hall voltage Quantitative stray-field maps
Scanning probeMFM · single NV Force gradient or spin-resonance frequency Nanoscale maps near a surface
Wide-field opticalQDM · Kerr · indicator film · Bitter Fluorescence, polarization, or particle density Fast context and in-situ changes

Microscopy resolution length scale

Resolution, sensitivity, field of view, speed, perturbation, and sample preparation trade against one another.

atomic–nm
electrons / X-rays
10–100 nm
MFM / single NV
0.5–10 µm
QDM / Hall / Kerr / Bitter
10–200 µm
scanning SQUID
mm–cm
bulk moment

The Lorentz force

\[\mathbf F = q\,(\mathbf E + \mathbf v \times \mathbf B)\]

Electron optics

A transverse magnetic induction changes electron momentum. Lorentz TEM converts that deflection into intensity; DPC reads the diffraction-pattern shift.

Transport probes

In a Hall cross, the same sideways force creates a transverse voltage proportional to the local field component.

Lorentz microscopy review: Phatak, Petford-Long & De Graef (2016). Hall microscope: Shaw et al. (2016).

Lorentz TEM turns deflection into domain contrast

How it works

Electrons crossing in-plane induction are deflected. In Fresnel mode, a defocused image makes domain walls bright or dark; in Foucault mode an aperture selects a deflected beam.

CollectWide-field images at controlled defocus, often through field or temperature sequences.
  • Excellent for dynamics and thin foils.
  • Contrast is not a direct magnetization map.
  • Rock preparation may alter stress or oxidation.
Mechanism: JEOL Lorentz electron microscopy guide; quantitative developments: Phatak et al. (2016). Interactive rays are schematic.

Lorentz TEM instrument overview

How it works

Electrons crossing in-plane induction are deflected. In Fresnel mode, a defocused image makes domain walls bright or dark; in Foucault mode an aperture selects a deflected beam.

CollectWide-field images at controlled defocus, often through field or temperature sequences.
  • Excellent for dynamics and thin foils.
  • Contrast is not a direct magnetization map.
  • Rock preparation may alter stress or oxidation.
Mechanism and quantitative developments: Phatak et al. (2016). Interactive data are synthetic.

DPC and 4D-STEM retain the full diffraction pattern

How it works

A focused electron probe rasters over the foil. A segmented or pixelated detector measures the beam’s center-of-mass shift at every probe position.

CollectA diffraction pattern at every \((x,y)\) position—hence “4D”—then calibrate shifts into projected electric and magnetic fields.
  • High data volume, flexible reconstruction.
  • Structural diffraction can mimic field shifts.
  • Scan distortion matters.
Lorentz/DPC context: Phatak et al. (2016). Interactive data are synthetic.

Hall effect: magnetic deflection becomes voltage

From force to voltage

A perpendicular field pushes moving carriers sideways. Edge charge builds a transverse electric field until the electric and magnetic forces balance.

Key readoutThe voltage between the two side contacts changes sign when either the field direction or dominant carrier sign reverses.
  • Current supplies the moving charge carriers.
  • \(B_\perp\) controls the transverse Lorentz force.
  • Carrier density, charge, and sensor thickness set the response.
Teaching schematic uses the displayed current, field, and voltage sign convention.

A Hall cross gives a robust local field voltage

How it works

Bias current flows through a small cross. The Lorentz force separates charge, producing \(V_H \propto I B_\perp/(nqt)\).

CollectRaster \(V_H(x,y)\) while recording topography or height; subtract offset and calibrate field responsivity.
  • Wide field range and simple readout.
  • Works from ambient to cryogenic conditions.
  • The active cross averages the local field.
Instrument design: Shaw et al. (2016). Synthetic data.

Phase stores information that intensity can miss

\[\Delta\phi = C_E\!\int V\,dz - \frac{e}{\hbar}\int \mathbf A\cdot d\mathbf l\]

Electron holography

An electron wave that crossed the specimen interferes with a reference wave. The reconstructed phase contains electrostatic and magnetic contributions.

Ptychography

Overlapping coherent diffraction patterns constrain a complex object. Magnetic sensitivity must still come from Lorentz phase or dichroic contrast.

Electron holography: Tonomura (1987). Ptychography primer: Rodenburg et al. (2025).

Electron holography reconstructs magnetic phase

How it works

An electrostatic biprism overlaps an object wave with a vacuum reference wave. Fourier processing recovers phase; its spatial gradient gives the projected in-plane induction after electrostatic phase is removed.

CollectAn interferogram, vacuum reference, and usually reversed or flipped states to separate magnetic from mean-inner-potential phase.
Foundational review: Tonomura (1987). Interactive data are synthetic.

Ptychographic tilt series can recover a 3D vector field

How it works

Overlapping probes recover phase at each tilt. Harrison et al. used pre-edge phase XMCD to measure a magnetization projection for each circular polarization.

CollectTwo perpendicular tilt series—73 and 71 projections—constrain \(M_x\), \(M_y\), and \(M_z\) inside the particle support.
  • Curved vortex core and Bloch point.
  • About 50 nm maximum spatial resolution.
  • Limited angles create missing-wedge distortion.
Giant-magnetofossil vector tomography: Harrison et al. (2025). Simplified reconstruction geometry; no paper figure is copied.

Zeeman splitting turns a spin into a field meter

\[ \begin{aligned} \frac{H_{\mathrm{NV}}}{h} &=D S_z^{2}+\gamma_e\,\mathbf B\!\cdot\!\mathbf S,\\[3pt] f_{\pm}&\approx D\pm\gamma_e B_{\parallel}, \qquad \Delta f=2\gamma_e B_{\parallel}. \end{aligned} \]

Prepare

Green light preferentially returns the NV spin to the bright \(m_s=0\) state.

Measure

Microwaves sweep through spin resonances. Their field-dependent splitting gives the projection of \(\mathbf B\) along an NV axis.

NV magnetometry review: Barry et al. (2020). Geological QDM: Glenn et al. (2017).

A nitrogen–vacancy defect traps a spin inside diamond

From crystal defect to sensor

One nitrogen atom replaces carbon beside an empty lattice site. In the negative charge state, six defect electrons form an \(S=1\) ground state.

Zero-field splittingElectron–electron interactions place \(m_s=\pm 1\) about \(D=2.87\) GHz above \(m_s=0\), even before an external field is applied.
Defect geometry and measurement sequence: NIST NV Center Magnetometry. Energy-level physics: Barry et al. (2020). Original schematic; electron-density lobes are qualitative.

ODMR reveals the field-split spin transitions

Read an ODMR spectrum

Green light prepares the bright \(m_s=0\) state. A microwave sweep transfers population when \(hf\) matches an allowed spin transition, reducing red fluorescence.

Move the controlsA 0.9–1.4 mT bias field sets the baseline separation; the sample field shifts it locally. \(^{15}\)N gives a doublet and \(^{14}\)N a triplet.
ODMR model adapted from course Notebook 01; physical interpretation follows Barry et al. (2020). Synthetic spectrum.

A single NV trades speed for nanometre proximity

How it works

One NV center sits in an AFM tip. ODMR at each pixel gives the field projection along the known NV axis while the tip maintains close standoff.

CollectTopography plus one or more ODMR frequencies at each point, giving the quantitative projection \(B_{NV}=\mathbf B\!\cdot\!\hat{n}_{NV}\) rather than MFM’s tip-dependent force-gradient contrast.
  • Quantitative, non-contact field projection.
  • Ambient and nanoscale.
  • Point spectroscopy makes large maps slow.
Scanning-NV implementation: Tetienne et al. (2015). Synthetic data.

A wide-field QDM reads every pixel in parallel

How it works

A shallow NV ensemble lies beneath the diamond surface. A camera records fluorescence while microwaves sweep through ODMR resonances.

CollectAn image stack over microwave frequency, fit an ODMR spectrum per pixel, then transform four NV-axis projections into field components when justified.
  • Micrometre maps over millimetres.
  • Optical and magnetic images co-register.
  • Resolution–sensitivity–NV-layer tradeoff.
Geological QDM performance and calibration: Glenn et al. (2017). Synthetic data.

Three more couplings complete the toolkit

Force and force gradient

MFM reads \(\partial F_z/\partial z\); induced moments in Bitter fluid drift approximately along \(\nabla |B|^2\).

Polarization rotation

Kerr and Faraday effects convert a magnetization-dependent dielectric response into optical intensity.

Helicity-dependent absorption

XMCD compares left- and right-circular X-ray absorption near an element edge.

\[C_{\rm XMCD}=\frac{I^+-I^-}{I^++I^-} \propto \hat{\mathbf k}\cdot\mathbf M\]
MOKE: Schäfer (2007). XMCD-PEEM review: Ghidini et al. (2022).

MFM senses a field derivative, not the field itself

How it works

A magnetized tip oscillates above the surface. Magnetic force gradients shift cantilever phase or resonance frequency.

CollectTopography first, then phase or \(\Delta f\) at a defined lift height. The greyscale animation shows tip-dependent force-gradient contrast over a patterned magnetic film—not a calibrated \(B_z\) map.
  • Fast access to domain-scale texture.
  • Topographic and electrostatic cross-talk.
  • The tip can move soft domain walls.
Rock-magnetic application and 40 nm lift-mode protocol: de Groot et al. (2014). Synthetic data.

Bitter fluid makes surface domain walls visible

How it works

A thin colloid of superparamagnetic magnetite particles is placed on a polished magnetic surface. Particles concentrate where the stray-field gradient is strongest, commonly where domain walls meet the surface.

CollectAn optical micrograph after the particles settle; a saturated-state reference can help separate scratches and polishing relief from magnetic decoration.
  • Simple, inexpensive, and wide-field.
  • Shows wall locations—not magnetization direction.
  • Surface preparation and fluid condition control contrast.
Method and preparation: Liu et al. (2017). Field-gradient interpretation: Williams et al. (1992). Synthetic data.

Kerr microscopy images domain motion in real time

How it works

Reflection from a magnetized surface slightly rotates polarization or changes ellipticity. An analyzer turns that small change into intensity contrast.

CollectReference-subtracted camera frames, often synchronized with a swept field to image nucleation and domain-wall motion.
  • Fast, wide-field, and field-compatible.
  • Surface-weighted and optically limited.
  • Polish and reflectivity affect contrast.
Kerr microscopy principles: Schäfer (2007). Synthetic data.

Indicator films convert stray field into Faraday rotation

How it works

A magneto-optic film placed near the specimen rotates transmitted polarization in response to local \(B_z\). A reflective layer returns the light to the camera.

CollectWide-field intensity with background and calibration images; model the film response and physical gap.
  • Large field of view and real-time response.
  • Useful bridge from hand sample to microprobe.
  • The indicator film is itself a spatial filter.
Operation and calibration review: Dorosinskiy & Sievers (2023). Synthetic data.

XMCD microscopy adds element specificity

How it works

Near an absorption edge, left- and right-circular X-rays interact differently with magnetic moments. PEEM images emitted electrons; STXM images transmitted photons.

CollectRegistered \(I^+\) and \(I^-\) images at selected energies and beam directions; compute asymmetry and, with rotations, vector constraints.
  • Element- and orbital-sensitive contrast.
  • Surface-sensitive PEEM; thin-section STXM.
  • Synchrotron access and preparation required.
Rock magnetite application: Carporzen et al. (2015). Synthetic data.

SEMPA maps surface spin polarization

How it works

A focused electron beam releases secondary electrons whose spin polarization follows the surface magnetization. A spin detector resolves two in-plane components.

CollectRaster intensity and spin-asymmetry channels, then rotate or add detector axes for vector reconstruction.
  • Direct vector magnetization contrast.
  • Nanometre-scale surface mapping.
  • Requires ultrahigh vacuum and a very clean surface.
Technique review: Koike (2013). Synthetic data.

Correlative microscopy is stronger than any single map

Locate carriersQDM or SQUID on a thin section
Identify phasesreflected light, SEM/EDS, Raman
Resolve domainsMFM, electron, or XMCD method
Test stabilityfield, temperature, time, or demagnetization

Choose the observable before choosing the instrument

Locate weak remanence
StartQDM or scanning SQUID
Addoptical and chemical registration
Track domain changes
StartMFM or Kerr microscopy
Addheight, tip, and field controls
Resolve nanoscale induction
Startholography or DPC/Lorentz TEM
Addelectrostatic separation and simulation
Identify the magnetic element
StartXMCD or dichroic ptychography
Addspectroscopy and multiple projections
Quantify near-surface stray field
Startsingle NV, Hall, or SQUID
Addsensor response and height series