This experimental study asks how inclined three-dimensional scaffolds affect exchange bias in sputtered CoFe/IrMn bilayers. Two-photon lithography produces 20 µm-long microwire scaffolds at 0°, 30° and 45°, and dark-field magneto-optical Kerr measurements resolve their magnetic reversal. Smooth wires show increasing exchange bias and coercivity with tilt, consistent with geometrically reduced coating thickness; rougher inclined wires show weaker bias enhancement. The demonstrated scope is three inclinations and micron-sized wires, with quantitative AFM roughness measurements restricted to flat

TWO-PHOTON LITHOGRAPHY / 3D MAGNETISM

A printed shape can tune a magnetic interface. Discover how tilting a microscopic wire changes its coating—and the way its magnetization reverses.

THE IDEA

A magnetic preference, built into an interface.

CoFe is a ferromagnet : its magnetic moments can align together. IrMn is an antiferromagnet , with opposing moments. Coupling at their interface gives the CoFe layer a preferred magnetic direction. This is exchange bias .

Because the interaction comes from the interface, changing the thickness of the magnetic layer changes its influence. Here, geometry supplies the thickness change.

01 / VIRTUAL FABRICATION

Put the geometry to work.

Same deposition. Different angle. Can you get the CoFe coating below 4 nm?

tG = t0 cos θ

Thickness measured normal to the tilted surface

At 0°, CoFe is 5.50 nm thick and ideal relative exchange bias is 1.00 times.

Your goal: make the CoFe coating thinner than 4 nm.

The continuous-angle model assumes a smooth interface and sufficiently thick IrMn, giving ideal relative bias 1/cos θ. The paper measures separate wires at 0°, 30° and 45°. At 45°, the modeled coating is 29% thinner; actual magnetic fields are explored below.

FROM LIGHT TO A FUNCTIONAL INTERFACE

Print the scaffold. Then add the magnetism.

Write with a focus

A Microlight3D system uses a focused 532 nm pulsed laser. Two-photon absorption polymerizes resist within the focal volume, writing the three-dimensional scaffold.

Approximate voxel: 200 nm laterally, 500 nm axially.

Coat the top facet

A wider top strip on a narrower supporting wall forms a T-shaped cross-section. It separates the magnetic wire on top from the coating on the sidewalls.

Each wire is 20 µm long and 1 µm wide.

Set the preference

The CoFe/IrMn layers are sputtered in a 50 mT in-plane magnetic field, establishing the exchange-bias direction without high-temperature annealing.

Ta 10 / Pt 2 / CoFe 5.5 / IrMn 14 / Pt 5, nominal nm. Ta/Pt seeds growth; Pt caps the stack.

02 / THE MEASURED RESPONSE

Read the magnet’s memory.

Change the wire. Replay the field sweep. See the loop shift—and widen.

CoFe5.5 / IrMn14

Replay follows the measured points; each plotted loop averages 120 reversals.

Layer names show nominal thickness in nm.

The preferred direction shifts the loop along the field axis.

The half-width measures the resistance to magnetic reversal.

0° wire: exchange bias 25.0 mT; coercivity 3.1 mT.

On smooth wires, exchange bias increases with inclination.

Redrawn from the authors’ open TU Wien dataset ↗ (CC BY 4.0). The plot preserves the measured field and normalized MOKE signal; displayed parameters come from the dataset’s extracted values.

WHAT THE COMPARISON REVEALS

Tilt strengthens the bias

For smooth CoFe5.5/IrMn14 wires, the measured exchange bias rises from 0° to 45°. Coercivity also increases, from 3.1 to 4.5 mT.

The interfaces remain effective

Fits of exchange bias against inverse CoFe thickness give nearly equal slopes for smooth wires and planar films (mT·nm). The fits include an intercept; the absolute field is not simply the slope divided by thickness.

IrMn needs enough thickness

The 3 nm planar IrMn sample has essentially no exchange bias, while a wire with about 3.5 nm geometric IrMn thickness retains it. The critical thickness lies between these values.

03 / LOOK CLOSER

The surface matters, too.

Explore the paper’s microscopy and supplementary evidence.

Figure 3. Laser printing and sputter deposition, with electron microscopy of the fabricated wires. Scale bars: 5 µm.

Singh et al., npj Spintronics 4, 30 (2026). Original figures and supplements, displayed at reduced size; CC BY 4.0.

A smooth starting point

Supplementary Table S1 separates the scaffold’s roughness from the magnetic deposition. Optimized printing preserves a surface close to the glass substrate over the 2–50 nm spatial-wavelength window.

RMS roughness, before and after coating. The coated-glass sample uses CoFe8.5/IrMn14; the coated wire uses CoFe5.5/IrMn14.

Roughness meets direction

The rougher inclined wires show weaker bias enhancement. The authors attribute this to directional deposition on rough surfaces: long-wavelength features can shadow incoming material, creating local thickness variations.

Green-A resist was printed under optimized conditions. OrmoBio used the same settings and produced rougher scaffolds. Quantitative AFM roughness measurements were made on flat wires; inclined-wire surface features were examined by electron microscopy.

THE BIGGER PICTURE

Geometry becomes a fabrication parameter.

This study demonstrates exchange-biased magnetic interfaces on three-dimensional printed scaffolds. It connects the shape you fabricate to the layer thickness you grow and the magnetic response you obtain.

That connection helps guide multilayer fabrication for future three-dimensional spintronic architectures. The paper establishes the materials and measurement route; its device applications remain directions for further work.

Research you can explore.

Balram Singh, Aman Singh, Stefan Mikulik, Jakub Jurczyk, Volker Neu and Amalio Fernández-Pacheco. npj Spintronics 4, 30 (2026). doi:10.1038/s44306-026-00148-w ↗

Supplementary information ↗Source dataset ↗CC BY 4.0 ↗

Article and supplementary figures © The Authors 2026, reproduced at display scale. Measured data are redrawn from the CC BY 4.0 dataset by Balram Singh, Amalio Fernandez-Pacheco and Aman Singh; numerical formatting and display precision are adapted. The graphical abstract is an AI-generated conceptual illustration based on source Figures 2–4, with adapted colors, composition and schematic magnetic loops. Interactive geometry and spin depictions are teaching models.