What Is Heat-Assisted Magnetic Recording?

Heat-assisted magnetic recording (HAMR) is a magnetic data storage technology that uses a localized laser pulse to momentarily heat the recording medium during the write process, enabling data to be written at higher areal densities than conventional perpendicular magnetic recording (PMR) permits. By transiently raising the temperature of a tiny spot on the disk surface, HAMR reduces the coercivity of the recording material long enough for the write head's magnetic field to flip the bit's polarity, then allows the spot to cool and the written state to lock in magnetically. The approach was conceived as early as 1954 in an RCA patent and became commercially available in production hard disk drives beginning in 2022, as documented by the Computer History Museum's storage technology timeline on HAMR.

The fundamental motivation for HAMR is the superparamagnetic effect that limits conventional magnetic recording. As engineers reduce bit size to pack more data on a disk platter, the magnetic grains that store each bit must also shrink. Below a critical grain volume, thermal fluctuations at room temperature are enough to randomly flip the bit, destroying stored data. Conventional PMR reached practical areal density limits near 1 Tbpsi (terabits per square inch) because the coercivity needed to keep small grains stable exceeded the field a PMR write head could generate. HAMR resolves this by using a high-anisotropy, thermally stable medium that is only writeable above its Curie point.

Near-Field Optical Heating and the Write Process

Each HAMR write head integrates a semiconductor laser diode and a near-field transducer (NFT) that focuses optical energy to a spot far smaller than the optical diffraction limit, typically 50 nanometers or less in diameter. The NFT converts laser light into a strongly localized plasmonic near-field that heats the target grain to approximately 400 to 500 degrees Celsius in under a nanosecond. At that temperature, the grain's coercivity drops to a level the write head's electromagnet can overcome. Once the laser pulse ends, the heated spot cools at rates exceeding 10^10 degrees Celsius per second, locking the written magnetization in place.

The integration of optics with magnetic recording heads at nanometer precision represents one of the most demanding manufacturing challenges in consumer storage hardware. Write head alignment, NFT geometry, thermal management, and laser reliability all require tight control to sustain the billions of write cycles expected over a drive's operational life.

Recording Media and Areal Density

HAMR drives use iron-platinum (FePt) alloy recording media in the L1₀ crystal phase, which has a magnetocrystalline anisotropy energy about two to three orders of magnitude higher than the cobalt-chromium-platinum alloys used in PMR. This high anisotropy is what confers long-term bit stability at grain sizes small enough to support areal densities above 4 Tbpsi. FePt layers are deposited on the disk substrate with careful control of grain size distribution and orientation.

Research on advanced FePt recording systems, including Nature Scientific Reports modeling of HAMR recording media performance, demonstrates that grain pitch control below 7 nanometers is achievable and that media noise can be managed to support bit error rates compatible with error-correction systems in production drives. Seagate's Mozaic 3+ platform, the first commercial HAMR product family, achieves over 3 TB per disk platter, compared to roughly 1.5 TB per platter in current PMR designs, and the company's HAMR technology overview projects 50 TB capacity drives by 2026.

Applications

Heat-assisted magnetic recording has applications in a range of data-intensive storage contexts, including:

• Hyperscale data center storage infrastructure requiring high capacity per rack unit
• Cloud object storage and archival systems where cost per terabyte is the primary constraint
• Video surveillance and continuous-recording systems with large write volumes
• High-performance computing storage nodes for scientific simulation datasets