HAMR or Heat-Assisted Magnetic Recording is a hard drive technology that enables data recording on ultra-stable materials. A laser heats a microscopic spot for a fraction of a nanosecond, reducing its magnetic resistance so the head can magnetize it, and instant cooling reliably fixes the bit.
HAMR hard drives are used in data centers and hyperscaler storage systems where huge capacity over 30 TB per drive is required with reasonable power consumption. This is the foundation of future cold and cloud storage, helping reduce the cost per gigabyte for archives, Big Data, and IoT, slowing the advance of solid-state drives in the mass capacity segment.
Typical problems relate to laser diode reliability at high temperatures, the risk of disk lubricant breakdown due to heating up to 400 degrees Celsius, and the difficulty of heat removal from the nanospot. Moreover, parasitic radiation is inevitably generated, requiring protection of adjacent tracks from accidental erasure, which complicates the optical path design and raises production costs.
How HAMR works
The operating principle of HAMR is radically different from perpendicular magnetic recording or PMR, which uses weaker magnets. The key barrier is the magnetic recording trilemma: to increase density, media grains must be made smaller, but then thermal energy at room temperature spontaneously flips their magnetization, the superparamagnetic effect. In PMR this is circumvented by creating structures with high coercivity, but the write head physically cannot produce a field strong enough to switch them. HAMR solves the trilemma by separating the processes in time: the FePt alloy media has enormous stability at room temperature, but under local laser heating through an optical transducer, a near-field antenna, to the Curie point its coercivity drops almost to zero. In this brief moment the ordinary magnetic field easily changes the grain magnetization, after which the region instantly cools, freezing the recorded bit with incredible reliability unattainable for PMR drives without risk of data loss.
HAMR functionality
- Principle of thermomagnetic recording. Local heating of the media by laser to the Curie temperature reduces the magnetic layer coercivity by several orders of magnitude, allowing a standard write field to switch highly stable material. Without heating, the write field is insufficient to change the bit states.
- Magnetic layer material. FePt alloys in the L10 phase with high magnetocrystalline anisotropy are used, providing thermal stability for grains under 6 nanometers in diameter. The structure with an ordered [001] axis is strictly perpendicular to the platter plane and has a coercivity exceeding 15 kOe at room temperature.
- Plasmonic antenna transducer. The Near-Field Transducer or NFT optical element is shaped as a metallic nanostructure of the bowtie or lollipop type made of gold or silver. It converts the volume electromagnetic wave of the laser into near-field radiation, forming a heating spot of subwavelength size from 20 to 30 nm.
- Laser diode integration. A semiconductor laser with power up to 30 mW is mounted directly on the magnetic head slider, emitting light at a wavelength of around 830 nm. The electrically modulated diode pump current is synchronized with the write coil switching without parasitic delay.
- Track thermal profile. The system creates a temperature gradient with a peak above 700 K in the write region and a sharp drop to room temperature at the spot periphery. The high gradient, reaching 10 K per nanometer, ensures sharp bit transitions and suppresses adjacent track erasure.
- Optical path and energy delivery. The laser emission passes through a waveguide made of a high refractive index dielectric doped with Ta₂O₅ or Si₃N₄. A focusing grating forms a beam directed into the NFT at a precisely adjusted angle for maximum efficiency of the decaying plasmon resonance.
- Heater gap control. An embedded resistive microheater alters the slider geometry, forcing the NFT pole to approach the disk surface at a distance under 1 nm. Active gap adjustment by thermal protrusion ensures stable optical contact with the magnetic layer.
- Switching mechanism in a gradient. The heated spot is offset relative to the trailing edge of the write pole along the media motion direction. The material cools in the presence of a strong write field, fixing the grain magnetization according to the applied field direction after crossing the blocking temperature.
- Dual pulse synchronization. The laser pulse precedes the write current edge by units of picoseconds, creating the necessary thermal window by the time the magnetic field edge arrives. This timing diagram eliminates parasitic field impact on cold grains with high coercivity.
- NFT wear problem. The plasmonic antenna operating temperature exceeds 500 K, provoking thermally activated diffusion of gold atoms and tip deformation under thermomechanical stress. NFT geometry degradation leads to irreversible heating power loss and a drop in signal-to-noise ratio.
- Thermally induced jitter broadening. Fluctuations in absorbed laser energy cause random changes in the heating spot diameter. Variations in the width of the reduced coercivity zone modulate the switching front position, increasing timing jitter and reducing the read window margin.
- Heat dissipation into the substrate. HAMR disks employ a multilayer heat-sinking absorber based on high thermal conductivity metals directly beneath the recording layer. Heat removal deep into the structure prevents temperature accumulation in adjacent tracks during periodic writing and suppresses the cross-heating effect.
- Laminated film technology. To suppress lateral heat spreading, the magnetic layer is divided by insulating interlayers of carbon or oxides under 1 nm thick. Broken vertical heat exchange preserves the gradient, while intralayer lateral diffusion is limited.
- Ultra-high density positioning servo system. The thermal spot serves as the physical center of an implicit track servo mark. Actuator micro-jogging compensates for head deviations, holding the NFT position over the target track with a sigma accuracy better than 0.6 nm, accounting for suspension thermal jitter.
- Background thermal control system. A specialized algorithm continuously measures a sensor resistance on the slider to indirectly estimate the NFT temperature. A closed loop adjusts the laser current in real time, neutralizing the influence of ambient temperature fluctuations and laser diode aging.
- Adaptive track formatting. The head-disk contact state is continuously monitored by light reflection from the NFT. Upon detecting a momentary clearance disturbance, the controller initiates an on-the-fly sector rewrite using a free cache memory segment to compensate for gaps.
- Radius-dependent power modulation. When moving to the outer zones of the platter, the linear velocity increases, shortening the exposure time. The controller dynamically boosts the laser power proportionally to the radius, maintaining constant absorbed energy per pulse and equalizing the thermal profile.
- Error-correction coding. Due to the high probability of individual erasure events caused by NFT instability, iterative low-density parity-check codes with a block length over 4 KB are applied. The decoder exchanges soft decisions with the channel detector, recovering data with a minimal number of iterations.
- Microwave recording mode without heating. An alternative MAMR approach uses a spin torque oscillator to generate a microwave field in the gap, reducing dynamic coercivity without raising the temperature. Hybrid implementations combine mild heating and an oscillator, shifting part of the heating demand from the NFT to the spin torque.
- Recording density prospects. Modern prototypes demonstrate a bit cell area of less than half a squared light wavelength, achieving densities of 4 Tbit per square inch and above. Further scaling faces limits in thermal spot management and the search for materials with a sharper Curie transition.
Comparisons
- HAMR vs PMR or Perpendicular Magnetic Recording. In PMR, recording is performed solely by the head magnetic field, which limits density by the superparamagnetic limit. HAMR overcomes this barrier by using laser surface heating to the Curie point, temporarily lowering the material coercivity and allowing magnetization of ultra-stable grains inaccessible to conventional perpendicular recording.
- HAMR vs MAMR or Microwave-Assisted Magnetic Recording. Both technologies aim to overcome the recording trilemma but use different physics. MAMR employs a spin torque oscillator to generate a microwave field that lowers the switching energy. HAMR uses a purely thermal effect with a subwavelength laser spot, currently providing a higher write field gradient and a fundamentally greater density margin.
- HAMR vs SMR or Shingled Magnetic Recording. SMR does not alter the media recording physics but manages track geometry by overlapping them like shingles to increase density, creating complications with random rewriting. HAMR is implemented independently of SMR and can be combined with it, providing density growth through new FePt media while SMR merely optimizes the layout of existing domains.
- SMR (Overlapping track recording for increased density)
- HAMR vs EAMR or Energy-Assisted Recording without heat. Unlike heat-assisted HAMR, some EAMR concepts attempt to use electric field or piezo effect to lower coercivity. The key difference is that HAMR is already realized in commercial products, as laser heating through a plasmonic antenna proved to be a technologically more mature and controllable process than managing electric fields at the nanoscale.
- EAMR (Local heating for stable recording)
- HAMR vs BPM or Bit-Patterned Media. BPM assumes lithographic creation of physically isolated magnetic islands on the platter. This radically changes media production. HAMR, in contrast, preserves standard granular media but with a new alloy chemical composition, requiring only the addition of an optical path in the head, making it an evolutionary step unlike the revolutionary but expensive BPM lithography.
OS and driver support
HAMR implementation at the operating system level is realized through standard block device drivers, as the laser heating and recording are managed exclusively by the drive controller microcode, making the disk fully transparent to SATA or NVMe commands. The operating system interacts with the HAMR drive through typical AHCI or NVMe drivers without any kernel modification, while all operational specifics, near-infrared laser diode calibration, spot temperature control, and heating pulse synchronization with the magnetic field are handled by the embedded SoC with firmware that automatically adjusts the laser power and recording parameters depending on media wear and thermal expansion.
Security
HAMR hardware security is built on a combination of self-encrypting drive technology per the TCG Opal 2.0 standard with AES-256 hardware acceleration executed on the same controller that manages the laser module, while the encryption keys never leave the cryptographic coprocessor boundaries. To prevent attacks via firmware substitution, a chain of trust with digital signature verification of the firmware at boot stage via an embedded ROM bootloader is implemented, and temperature sensors with the laser monitoring system can hardware-lock writing upon detection of abnormal overheating, excluding physical platter damage and data leakage due to thermal degradation.
Logging
The logging subsystem of HAMR drives is implemented based on the SMART standard, supplemented with unique attributes for monitoring the laser-optical block lifecycle: the drive firmware continuously records the laser diode operating hours, peak write temperatures, the number of corrective retries, and the total absorbed energy. At the host system level, this data is aggregated via the standard smartctl utility, which accesses the controller internal log through ATA READ LOG EXT commands, and the engineering firmware automatically logs critical events into protected non-volatile NAND memory, forming a detailed log for predictive failure analysis of the laser optics and degradation of the FePt thermomagnetic layer.
Limitations
The key physical limitation of HAMR lies in the complexity of managing the thermomagnetic spot at ultra-high densities: laser heating of the media to the Curie temperature of around 700 K for the FePt alloy in a time on the order of one picosecond requires precision control, while parasitic thermal expansion and heat dissipation to adjacent tracks impose a fundamental limit on track pitch. From a software and hardware perspective, the limitations manifest in increased write latency due to the preheating cycle, which is partially compensated by DRAM and non-volatile buffer caching, as well as the impossibility of using standard data recovery methods from FePt media due to the unique material coercivity, making recovery outside the manufacturer’s facility economically unfeasible without a specialized laser bench.
History and development
The concept of thermomagnetic recording was theoretically substantiated in the late 1990s in Seagate and IBM laboratories, but practical implementation became possible only after the development of planar laser diodes and stable FePt media with high coercivity requiring heating to reduce magnetic rigidity. The technology development traveled a path from first prototypes with external lasers to a fully integrated system where the laser, near-field optical transducer or NFT, and magnetic head are combined in a single slider, and by 2025 commercial samples reached capacities of 32 TB and higher, with further evolution focusing on the introduction of bit-patterned media and the use of multi-beam heating systems for parallel recording on several surfaces simultaneously.