domain wall
/ doh-MAYN wawl /
Inside a magnet, neighboring regions called domains point in different directions. They cannot meet at a hard, abrupt cliff — that would cost too much energy. Instead the moments swing around gradually, twisting through a thin transition layer from one domain's direction to the next's. That gentle twisting layer is the domain wall.
Why a gradual turn rather than a sudden one? The exchange interaction hates having neighboring moments point far apart, so a sharp flip would be terribly expensive. The wall compromises: it spreads the reorientation over many atoms, each turning only slightly from the one before, so no two neighbors disagree by much. The wall is typically tens to hundreds of atoms thick — a real, locatable, movable object inside the crystal, not just a line on a diagram.
Domain walls matter because moving them, rather than flipping atoms one by one, is how a magnet actually magnetizes and demagnetizes. They can be nudged by applied fields or by electric currents, and the dream of 'racetrack memory' is to store bits as a procession of domain walls pushed along a nanowire. The subtle catch: walls get pinned on defects and impurities, so they jerk rather than glide — those jerks are the audible clicks of the Barkhausen effect and a major source of a magnet's stubbornness.
In proposed 'racetrack memory,' data lives as a row of domain walls strung along a magnetic nanowire. A pulse of current shoves the whole train of walls past a fixed reader, like beads sliding along a thread, reading out one stored bit after another.
Racetrack memory stores bits as domain walls and slides them along a nanowire with current pulses.
A domain is a region; a domain wall is the boundary between two regions. The wall has real thickness and structure, and how wide it is depends on a tug-of-war between the exchange interaction, which wants it broad, and magnetic anisotropy, which wants it thin.