{
  "slug": "thermomechanical-probe-storage",
  "term": "Thermomechanical probe storage",
  "aliases": [
    "Millipede",
    "Millipede memory"
  ],
  "definition": "Thermomechanical probe storage records data as nanoscale indentations in a polymer film, using heated scanning probes to write and sense them.",
  "note": "Mechanism: A heated tip and applied force form an indentation. Readout uses heat loss across the gap to the medium: entering a pit changes the heater temperature and electrical resistance. Each probe addresses its own field through mechanical scanning. [c1] [c3]\n\nHistorical results: The 2000 demonstration separated single-probe densities of 400-500 Gb/in2 from 100-200 Gb/in2 with a 32 x 32 array. The 2002 paper reported single-probe density up to 1 Tb/in2. These are different experimental scopes, not the capacity of a shipped drive. [c1] [c2]\n\nIn 2008, a prototype recovered three 1024-byte sectors stored across four data fields at a reported effective density of 840 Gb/in2. The detected symbol error rate was approximately 10^-4 before ECC; the demonstrated sector output was error-free after ECC. Successful decoding does not mean the physical channel made no errors. [c4]\n\nParallelism: Fabricated probe count, simultaneously addressed probes and user throughput are distinct. The 2003 design example used 256 active probes, a 5-microsecond symbol interval and constrained coding to obtain 34.1 Mbit/s; this was a design calculation. [c3]",
  "status": "research",
  "advantages": [],
  "limitations": [
    {
      "text": "Tip and medium wear, long-term indentation stability and the power cost of parallel operation require separate engineering evaluation.",
      "citationIds": [
        "c3"
      ]
    },
    {
      "text": "Writing a nearby indentation can partly erase an existing one, so smaller spacing changes the recording channel as well as density.",
      "citationIds": [
        "c4"
      ]
    }
  ],
  "related": [
    "areal-density"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "citations": [
    {
      "id": "c1",
      "title": "Vettiger et al.: The Millipede - More than one thousand tips for future AFM data storage, IBM J. Res. Dev. 44 (2000), pp. 323-340",
      "url": "https://www.eecs.northwestern.edu/~jianwei/projects/MEMS/vettiger.pdf",
      "accessed": "2026-11-07"
    },
    {
      "id": "c2",
      "title": "Vettiger et al.: The millipede - nanotechnology entering data storage, IEEE Transactions on Nanotechnology 1 (2002), pp. 39-55",
      "url": "https://research.ibm.com/publications/the-millipede-nanotechnology-entering-data-storage",
      "accessed": "2026-11-07"
    },
    {
      "id": "c3",
      "title": "Eleftheriou et al.: A Nanotechnology-based Approach to Data Storage, VLDB 2003, sections 2-4 and Figure 6",
      "url": "https://www.vldb.org/conf/2003/papers/S01P01.pdf",
      "accessed": "2026-11-07"
    },
    {
      "id": "c4",
      "title": "Pantazi et al.: Probe-based ultrahigh-density storage technology, IBM J. Res. Dev. 52 (2008), pp. 493-511",
      "url": "https://doi.org/10.1147/rd.524.0493",
      "accessed": "2026-11-07"
    }
  ],
  "updated": "2026-11-07",
  "provenance": [
    "Historical research architecture, not a current product-availability claim. No universal invention date or commercial launch is assigned.",
    "c1 sections 2-3 and 8 distinguish probe operation from array results; c3 sections 2-4 distinguish fabricated components, measured signals and design projections.",
    "c2 was checked through the IBM Research abstract and institutional bibliography; its full paper was not used for detailed mechanisms.",
    "c4 was read in the author-uploaded full text, especially pp. 499 and 505-507. The detailed sector demonstration is narrower than the prototype's supported parallelism.",
    "The 2008 reported pre-ECC rate is approximate; no count-based confidence interval, fleet failure rate or lifetime extrapolation is inferred."
  ]
}