Technology card
ShippingHeat-Assisted Magnetic Recording
HAMR
Why smaller grains need write assistance, how heating affects recording precision, and what HAMR experiments and models actually establish.
Glossary entry on hesela.dev · JSON record · Magnetic anisotropy · Magnetic coercivity · Recording trilemma · Near-field transducer · Thermal writability · Thermal stability
Related
On this page6 sections
In one minute
HAMR uses localized optical heating to let a magnetic write head record high-anisotropy media, which regain thermal stability as they cool.
Seagate reported 20 TB HAMR shipments in calendar 2020 in its January 2021 results. On 17 January 2024, it announced the later Mozaic 3+ platform and a 30 TB-plus hyperscale volume ramp, with some customer qualification still pending. Neither supplier statement establishes unrestricted availability, and 2024 is not the first HAMR shipment milestone.
At unchanged material anisotropy, reducing magnetic grain volume reduces the energy barrier against thermal reversal. Smaller does not automatically mean more stable. Maintaining retention with smaller grains requires stronger anisotropy, but this makes writing harder: the stability, signal quality and writeability requirements compete. The 2014 NFT review explains this trade-off; it is not a fixed capacity ceiling for every disk.
Coercivity answers a different question from thermal stability: how large a reverse field brings a specified magnetic quantity to zero after saturation. IEC 60404-7 distinguishes flux-density, polarization and magnetization conventions; report the quantity and conditions, not just a number. NIST's reference problem also separates coercivity from remanence, the state at zero applied field. TDK Headway describes HAMR heating as temporarily lowering media coercivity while the magnetic pole writes. This field-strength measure, in A/m, is not the anisotropy energy density or a direct drive-lifetime prediction.
A laser supplies light through a waveguide to a near-field transducer (NFT), a plasmonic structure that concentrates energy below ordinary far-field focusing scales. The medium is heated locally so the magnetic pole can establish the recorded direction. The laser supplies assistance, not a replacement for the magnetic write field.
The 2018 optical-transfer study examined fabricated prototype heads with cathodoluminescence and scattering near-field microscopy, supported by electromagnetic modeling. A particularly useful caveat: removing the recording medium shifted the antenna resonance. A head measurement outside its operating environment is therefore not automatically the same optical response seen in a drive. These are component measurements, not drive-lifetime tests.
Heating also introduces a second question: will the grain end up in the intended state? Evans and colleagues call this thermal writability. Their equilibrium model separates a sufficiently low switching barrier from a sufficiently strong statistical preference for the written direction. Its numerical density estimates assume idealized bit-patterned recording and a specified write field; they are not universal limits for commercial granular HAMR.
Curie-temperature distribution is a spread of material transition temperatures across grains, not the temperature profile of a laser spot. Binh and colleagues modeled finite-size effects in FePt using atomistic Monte Carlo calculations and a mean-field comparison. Surface disorder increasingly affected the grain interior near a 3.5 nm lateral size in their geometry. That is a model-dependent crossover, not a universal minimum grain size or a commercial design specification.
A spatial thermal gradient, in K/nm, describes how sharply temperature changes with position. It is not a cooling rate in K/s. Hosen and Victora studied how material-property distributions affect recording at different gradients; their institutional abstract supports that dependency, not a universal conversion to drive error rates. Faster thermal changes also leave less time to switch: Meo and colleagues compared atomistic and granular models with 100 repetitions per condition. Their isolated-grain calculation varied pulse duration, peak temperature and field; it did not calculate a complete-medium BER.
Transition jitter describes variation in recorded boundary positions. Its estimator matters: Jongjaihan and Kaewrawang used five simulated tracks, each with 31 bits and 30 boundaries, and averaged per-track jitter rather than pooling every boundary into one standard deviation. Their density sweep changed both track width and hotspot size. A lower reported jitter therefore cannot be attributed to density alone. These are numerical recording results, not observed drive failures.
Another less obvious issue is smear: contamination accumulating near the NFT. Tom and colleagues modeled optical forces, drag and thermophoresis. Their initial comparison used an 8 nm head-disk gap and different silica-particle shapes. Optical trapping was insufficient for the modeled small sphere but could matter for the flatter ellipsoid. Particle geometry and interface conditions are essential to the inference.
That 2023 work does not measure a fleet failure rate. Its head design and underlying electric and temperature fields were proprietary, and van der Waals forces were outside the quantitative comparison. The result identifies a possible contamination mechanism under modeled conditions; it cannot establish that all HAMR drives develop the same failure mode.
For a dated commercial reference, Seagate announced on 3 March 2026 that Mozaic 4+ was qualified and in production with two hyperscale providers, supporting capacities up to 44 TB. This supports the shipping badge as a vendor report. It is neither an invention date nor a claim that every capacity, interface and channel is generally available.
Compared to neighbors
PMR describes magnetization orientation; CMR and SMR describe track layout. HAMR describes thermal assistance. These labels are not mutually exclusive generations, and PMR does not mean that tracks must be widened.
Separate the quantities and evidence behind a HAMR claim.
- Coercivity
- Relevant evidenceHysteresis zero crossing with quantity, temperature and field conditions stated (A/m)
- Not establishedAnisotropy energy density, remanence or drive lifetime
- Retention
- Relevant evidenceGrain stability at storage temperature
- Not establishedSuccessful writing at elevated temperature
- Curie-temperature spread
- Relevant evidenceDistribution across grains, measured in K or a stated relative spread
- Not establishedSpatial temperature gradient of the heat spot
- Thermal profile
- Relevant evidenceSpatial gradient (K/nm), pulse duration and cooling rate (K/s)
- Not establishedA single interchangeable measure of recording precision
- Transition placement
- Relevant evidenceA stated jitter estimator in nm for a defined recording model
- Not establishedHost I/O latency, complete-medium BER or fleet AFR
- Optical coupling
- Relevant evidenceNear-field maps for a specified head and environment
- Not establishedService life of a complete drive
- Contamination
- Relevant evidenceForce model with stated particle and gap assumptions
- Not establishedObserved fleet failure probability
- Availability
- Relevant evidenceDated qualification or shipment statement
- Not establishedAvailability of every model or a universal invention date
Uncertainty notes
- No single firstSeen year is assigned: prototype history, publication and volume deployment are different events.
- Hesela reviewed these papers but did not reproduce the head experiments or simulations. No third-party figures are copied.
- The 2014 review supplies physical background; current product status comes from the explicitly dated 2026 vendor announcement.
- Recording-precision studies use different populations and estimators. No simulation is reproduced here; the cited 2023 gradient paper was available as an institutional abstract only.
Sources
- 01IEC 60404-7:2019, sections 3.1-3.2 and 4: coercivity definitions and saturation (official SIST preview) · accessed 2026-11-06
- 02NIST muMAG Standard Problem 2: coercivity, remanence and geometry-dependent reversal · accessed 2026-11-06
- 03TDK Headway: Technologies, HAMR discussion of optical heating and media coercivity · accessed 2026-11-06
- 04Seagate fiscal Q2 2021 results, 21 January 2021: calendar-2020 HAMR shipments · accessed 2026-10-23
- 05Seagate Mozaic 3+ launch, 17 January 2024: volume ramp and qualification scope · accessed 2026-10-23
- 06Zhou et al.: Plasmonic near-field transducer for heat-assisted magnetic recording, Nanophotonics 3(3), 141-155 (2014), doi:10.1515/nanoph-2014-0001; section 1 and equation 1 · accessed 2026-10-16
- 07Visualizing the bidirectional optical transfer function for near-field enhancement in waveguide coupled plasmonic transducers, Scientific Reports (2018), doi:10.1038/s41598-018-24061-3; Figures 2-3 and Methods · accessed 2026-10-16
- 08Evans, Chantrell, Nowak, Lyberatos and Richter: Thermally induced error: density limit for magnetic data storage (2012), doi:10.1063/1.3691196; equations 1-2 and concluding limitations; arXiv v2 · accessed 2026-10-16
- 09Tom, Smith, Ruiz, Dai and Bogy: Optical forces in heat-assisted magnetic recording head-disk interface, Scientific Reports 13, 8451 (2023), doi:10.1038/s41598-023-35126-3; Simulation assumptions and Results · accessed 2026-10-16
- 10Seagate: Mozaic 4+ qualification and production announcement, 3 March 2026 (vendor claim) · accessed 2026-10-16
- 11Binh, Ruta, Hovorka, Evans and Chantrell: The Influence of Finite-size Effects on the Curie Temperature of L10-FePt, arXiv:2207.11831v1 (2022); section III, Tables 1-2 and Figure 7 · accessed 2026-10-16
- 12Meo et al.: Magnetisation dynamics of granular HAMR media by means of a multiscale model, arXiv:2004.13345v1 (2020); sections II.2 and III.2, Figures 2-4 · accessed 2026-10-16
- 13Jongjaihan and Kaewrawang: Micromagnetic Simulation of L10-FePt-Based Transition Jitter of Heat-Assisted Magnetic Recording at Ultrahigh Areal Density, Micromachines 13(10), 1559 (2022), doi:10.3390/mi13101559; section 2, equations 5-6, Table 2 · accessed 2026-10-16
- 14Hosen and Victora: Impact of Curie Temperature Distribution and Anisotropy Field Distribution on the User Density of HAMR Under Different Thermal Gradients, IEEE Transactions on Magnetics 59(11), 3201005 (2023), doi:10.1109/TMAG.2023.3287829; institutional abstract only · accessed 2026-10-16