1. Magnetic Data Storage (1) Magnetic recording (a) General (longitudinal recording) (b) Thermal stability (c) Advantage Media Oriented longitudinal media Anti-ferromagnetic coupling media Perpendicular recording Pattern media and nano-particle media High K u medium (HAMR) (2) Magneto-optical recording (3) MRAM (STT RAM) and Flash disc (4) RRAM and PRAM (Random Access Memory) (5) Optical storage and other memory

2. Areal density progress in magnetic recording since its invention (Moser et al. J.Phys D: Phys. 35(2002)R157-167) 1Tbits/in 2 100Gbit in -2

3. Magnetic recording areal density growth along with transistor count per integrated circuit device (McDaniel J. Phys: Condens. Matt. 17(2005)R315).

5. Areal density trends in HDD magnetic recording (Fujitsu Sci. Tech. J., 42(2006)122).

6. Industry first 120 GB 2.5-in Seagate Momentus Ⅱ high capacity mobile drive using TMR reading element (IEEE on Mag. 42(2006)97).

7. Schematic drawing of longitudinal recording system. B is the bit length, W is the track width and t is the medium thickness. d is flying height of the head abov the medium.

8. Schematic representation of longitudinal, digital magnetic recording write process.

9. Transition width α (depends on M r t / H c )

10. (a)magnetization of two transition at x=0 and 200nm. (b) magnetic field detected by read head, solid line is for longitudinal. pw 50 is shown for a read head with zero gap.

11. SNR≈0.31PW 50 BW read / α 2 d(1+σ 2 ) ≈B 2 W read / α 2 d 3 (1+σ 2 ) B is bit length, W read is read width of head, α is transition parameter, d grain diameter, σ normalized grain size distribution width

12. (1) for few particles per bit, the transition becomes less sharp and pickup signal decreases. About 400 isolated particles are required. (2) Noise is due primarily to the formation of zigzag transition between bits and this sawtooth pattern scales roughly as M s 2 /Ku 1/2, (3) the signal is proportional to the number of measured events or particles per bit, N. Hence SNR ~ N 1/2. (4) the heads must approach to the hard disc surface. Recording Media Requirements

13. PtCoCrB films

15. Write head : having a sufficient high M s so that the fringe field exceeds the H c of the medium (500-3000Oe); an adequate magnetic permeability (easy saturated). Read head: low H c, low noise and extremely high permeability in order to respond with a substantial change in flux to the weak fringe field above the medium

16. Schematic M-H loop for ideal magnetic recording medium and head material. For write head: µ >>1, M s large and B r =0; For read head: µ >>1, H c = 0

17. Thin film recording head. Left, layout of pole pieces and windings; right, enlarged, cross-sectional view of magnetic pole pieces Film thickness 2-3 micrometer; Gap 200 nm. Thin film recording head

18. h=1-2 µm, w=2-4 µm t=10-20 nm Δρ/ρ =2.0% Ni 81 Fe 19 Magnetoresistive read head (1980-90 from 10 -100 Mbit in -2 )

19. Spin-Valve Read head h=2-6 µ m and w=10 µ m

20. Summary (1) SNR≈0.31PW 50 BW read /α 2 d(1+σ 2 ) ≈B 2 W read / α 2 d 3 (1+σ 2 ), (2) Transition width α(depends on M r t / H c ) (3) Signal: small M r t, large Hc, small distance between head and disc, large GMR or TMR (4) Areal density: decreasing the dimensions: B, W Read, diameter of grain.

21. Before 1985: γFe 2 O 3 medium, Ferrite ring head ( ～ 10Mbin -2 ) 1980: 1st thin film read head, continuous magnetic thin film with high H c, small α(25% CGR); 1990: 1st MR read head, decreasing thickness and, in turn, the transition distance (80% CGR); 1997: 1st GMR read head (100% CGR); 2000: 1st AFM medium, increasing the effective volume. 2006: 1st TMR head for 80-100 Gbit in -2 longitudinal recording The develop of the magnetic recording

22. Thermal Stability In the physics of magnetic recording there are two key factors in achieving very high areal density: (1)The superparamagnetic effect (thermal stability); (2)The finite sensitivity of the readback head. In both cases, the limitations arise because the signal energy becomes so small as to be comparable with the ambient thermal energy.

23. The signal to media noise is approximately by the number of magnetic grains (or switching units) per bit: SNR media ~ Wbt / V Where, wbt (bit volume, read-width x bit-length x thickness) v (the grain volume) In order to avoid thermal instability, a minimal stability ratio of stored magnetic energy, K u V, to the thermal energy, K B T, K u V/K B T ≌ 50 - 70

24. A favorable lattice matching between CoPtCrB (1120)[0001] Is parallel to CrX (002)[110]. Toney et al., IEEE Trans. On Mag 99(2006)033907. Oriented longitudinal media (K u )

25. A perfect orientation (large Ku) carries out: (1) a low media noise (2) a high signal level (3) a smaller transition parameter (4) a narrower switching field distribution OR = M r / M r per >2.5 for current L media mechanically texturing metal disk substrate anisotropic etching of the substrate directional deposition of the media

26. (Parkin PRL 64 (1990)2304). Oscillation Exchange Coupling in Co/Ru/Co MLs

27. Interlayer antiferromagnetic coupling media Schematic illustration of (a) a two layered AFC media, (b) LAC media with high J and (c) advanced three layers LAC media for much lower M r δ. Longitudinal

28. M r t = M r t 1 – M r t 2 K u V 1 <K u V eff < (K u V 1 +K u V 2) K u V/K B T ≌ 50 - 70 In the case of two layers AFC media single layer media (a) and (b) an AFC media, J ex =0.06 erg/cm 2, H ex ~800 Oe.

29. (b) thermal decay at -500 reversal field. Fitted by Eq.(1) APL 77(2000)3806

30. Comparison of amplitude loss as aresult of thermal degration of single layer media and AFC media

31. System parameters for estimating transition width and PW 50 Design (AFC media) 60Gb in -2 200 Gb in -2 Bit length, B (nm) 38 27 Track width, W (nm) 280 115 Magnetic coecivity, Hc (Oe) 4000 5000 M r t (memu/cm 2 ) 0.32 0.2 Grain size (nm) 8.1 6 Head to media spacing (nm) 30 15 Shield to shield spacing, g(nm) 700 500 Transition parameter, a (nm) 12.8 6.2 Pulse width PW 50 (nm) 99 54 User bit density (pw 50 /B) 2.6 2.11 IEEE on Mag. 39(2003)651 (Komag Inc.)

32. Magnetic Recording (1) Traditional longitudinal recording is approaching to its limit (100 Gbit in -2 is achieved ). (2) perpendicular recording offers about 421Gbit/in2 (Seagate demo) and 178.8Gbit/in2 (market). (3) the next big challenge is 1 Tbit in -2 for recording industry. The possible models : pattern media; high K u media (HAMR); STT (Spin torque transfer) – RAM.

33. Schematic drawing of a perpendicular recording system with SUL and a single pole head. perpendicular recording

36. Advantages of PA recording: a. high orientation ratio b. lower media noise (α smaller) c. increase of signal and thermal stability d. writing field large

37. *Toshiba extends 2.5-inch mobile HDD Family with 200GB market-leading capacity (178.8 Gb/in2) (May 2007 market), *Fujitsu intros 250GB perpendicular drive (second quarter of 2007), *In the first half of 2007, Hitachi has brought hard drive areal density halfway to the 345 Gbits/sq. in. market with the 1 TB, 3.5-inch (Deskstar 7K1000).Deskstar 7K1000 * Seagate 500GB for 2.5-inch (notebook), 2.5TB for 3.5 inch desktop (41650 hours music, 800,000 photo, 4100 hours digital video) to emerge in 2009. (Hitachi demo) Perpendicular Recording

38. Perpendicular recording hard disc drive

39. PR recording using AF coupling media

40. Magnetic recording on a CoPd/Pd/CoPd dot array: (a) GMR readback signals after dc magnetizing the sample (00) state and after applying a write pulse of 30 and 50 mA, creating, respectively, states (01) and (11);(b) SMRM image 1Tbit in -2 for 40nm period

41. (a) topography image of the patterned area. P 2 is write pole. (b) Magnetic force microscopy image of a square wave pattern Patterned media made by a focused ion beam Thermal stable, even if K u is small; transition parameter

42. The islands are lithographically patterned into regular array in the recording medium; For 1 Tbit in -2, the island array periodicity is 25 nm and the lithographic linewidth is ~12.5 nm for equal island and trench width. The transitions must be precisely written between two islands 1 Tbit/in 2 for patterned media

43. TEM image of a 3D assembly of FePt nanoparticles. Image size is 130nm x 130nm and particle diameter is 4nm. Nanopartical media are made in a chemical process, then annealed to obtain a hard magnetic phase. Nanoparticle media (self arrangements)

44. AD = pδK / hNk B T δ= 10nm, K=7x10 7 erg cm -3 h = KV/k B T= 60, T=330K, p=0.56 L bit =2.64 nm(10-12 atoms) cross-section of the bit, 60-80 atoms volume 8 x 8x 50=3200 to 9 x 9x 50=4050 atoms Given AD≈92 Tbin -2 HAMR McDaniel Seagate Ultimate limit to thermally assisted magnetic recording J Phys:C 17(2005)R315-332

45. hybrid recording (Solid immersion lens) ZnS:SiO 2 NA ~1.1 Media: Co 69.48-x Tb 30.52 Ag x, x=0-25.68

46. SmCo has a K u value about three times high FePtX, and this might push AD estimate into 250-300 Tbin -2. The entire printed contents of the United State Library of Congress ( ~10Tb) could be stored on a 30 mm diameter disk (50Tb/in2). This is about the size of US fifty-cent coin. Fujitsu paves way for 5TB hard drives (1Tbits/in2, 04/12/2006 demo a spot size 88nm x60nm using HSRM)

47. Magneto-optical Effect θ k is defined as the main polarization plans is tilted over a small angle; ε k = arctan(b/a).

48. (a) Assembly of apparatus (b) Rotation of polarization of reflecting light.

49. Magneto-optical Recording Principle of thermomagnetic recording (Curie point writing): (a) before, (b) during and (c) after the writing.

51. From Oppeneer Magneto-optical Kerr spectra in Handerbook of magnetic Materials, Edited by Buschow (Vol.13) Experimental pola Kerr ritation an undoped MnBi sample (Di et al. 1992) and Al-doped MnBi (Shang et al., 1997) sample at room temperature.

52. High-density MRAM (Magnetic random access memories) Schematically representation of MRAM structure and M-H, ΔR/R characteristics of the PSV.

54. Schematic of the read and write processes in a PSV random access memory.

55. Table: composition and dimensions of the principle layers in a current representative MARM device.

56. Advantage of MRAM (1) It combines the speed of SRAM with the non-volatility of flash (2) It also offers a low-power memory solution which eventually may match DRAM’s capacity and density (3) No limit for write-read cycles (flash 100,000) (4) Radiation-resistant 7/2 2006 Toshiba and NEC, 16 megabit density, read write speed 200 Mbytes/sec, operation at 1.8V, a chip 78.7 mm2; 9/6 basic technology for 256 Mbit 4Mbits MRAM enter in market (2007 March meeting). As embedded memory in Automobiles, $20 for a half megabyte. Disadvantage

57. Fig. 1.R–H loop in MgO-based MTJ. Fig. 2. Magnetoresistance versus variation current in MgO-based MTJ structures using a current pulse width of 200 s. Fig. 3. Read/write cycle test. Inset: Resistance change over 110 cycles with a current pulse width of 200 s. The MgO-based magnetic tunnel junction film Ta/CuN/Ta/PtMn/CoFe/Ru/CoFeB/MgO/ CoFeB/Ta. (Lee etal., IEEE Trans on Mag 43(2007)917). STT RAM

58. (a) a 20 µm wide channel with three pairs of Hall probes. (b) a domain wall was prepared at the boundary of regions 1 and 2, and its position after application of a current pulse was monitored by R Hall =V Hall /l, H c1 >H c3 >H c2 Yamanouchi et al Natural 428(2004)539 Single-crystal multilayer 25nm(Ga 0.95 Mn 0.05 )As/500nm(In 0.15 Ga 0.85 )As/ 100nmGaAs on (001) semi-insulating GaAs Substrate (Tc=90K).

59. At t=0, the domain wall is at the boundary of regions 1 and 2; when a negative current pulse is applied, M direction in region 2 is reversed. minusPlus pulse83K

60. initial state After I=-300 µA After I=+ 300 µA MOKE images of sample A using 546 nm light at 80K.

61. Fig. 1. STT-RAM addresses each bit individually by flowing current directly through the bit. Unintended writing errors are completely eliminated. The future of scalable STT-RAM as a universal embedded memory By Farhad Tabrizi, Grandis, Inc. (02/21/2007)

62. Fig. 1. STT-RAM addresses each bit individually by flowing current directly through the bit. Unintended writing errors are completely eliminated.

63. Fig. 2a " Conventional MRAM cell. A magnetic field, generated by the bit line, cladding, and write word line, is used to switch between the "0" and "1" states.

64. Fig. 2b " STT-RAM cell. By eliminating the write word line, bypass line and cladding, a STT-RAM cell is considerably smaller than a conventional MRAM cell.

65. Fig. 2c " Total required current in STT-RAM continues to scale lower with increasingly smaller geometries. Conversely, conventional MRAM switching current increases with smaller geometries.

67. spin polarization along the axis parallel to the vector M L of local ferromagnetic polarization in will be present in the electrons impinging on M R. S 1,2 = (I e g/c)S 1,2 x (S 1 x S 2 ) g = [ -4+(1+p) 3 (3+S 1 · S 2 )/4p 3/2 ] (J.C.Slonczwski 3M 159(1996)L1) (C.Heide et al., PRB 63(2001) 064424).

68. Fig. 1. Plan-view and side-view scanning micrographs of a and b arrays of Co circular rings, c NiFe/FeMn elliptical rings, d NiFe/Cu/Co elliptical rings, and e and f NiFe/Cu/Co pseudo-spin-valve elliptical ring device with six nonmagnetic contact wires. NANORING for MRAM

69. Nanoring for MRAM We have achieved nearly 100% vortex reversal in the asymmetric nanorings, while the symmetric nanorings can accommodate only 40%. PRL 96(2006)027205

70. (a) AAO (porous anodic aluminium oxide membranes), (b) 100 nm Fe film by RF, (c) and (e) Fe nanoring in AAO pores. Nanoring fabrication

71. SEM micrograph of the top view of the as –prepared AAO template. The inset shows the oblique view of AAO showing the aligned nano- channls.

72. Samsung offers flash disk as laptop upgrade Report by Robinson in COMPuting, 22 May 2007 ------------------------------------------------------------------------------------------------------------- From the end of Jun, the 32GB SSD will be sold through memory specialist just Rarm as part of its integral brand For $350, 64GB model is due ship in the coming months

73. Competition Between Hard Disc and Flash Disc ------------------------------------------------------------- Flash can be a potential replacement for hard disks because of its high performance, noise free running, light weight, fast data access and less power consumption. iSuppli believes that by the fourth quarter of 2009, 24 million Notebooks will be sold with some form of flash data storage, compared to 143,600 in the first quarter of 2007. That's nearly 60 per cent of anticipated laptop sales. But the price factor is still the most active barrier.

76. Magnetic Hard Disc Freecom Mobile HDD Drive 160GB USB-2, Release Date: Wednesday 2nd May 2007,Our Price: £78.95 ; LaCie 500GB USB 2.0 Hard Drive, Release Date: Wednesday 20th September 2006, Our Price: £114.95 Seagate 500GB for 2.5-inch (notebook), 2.5TB for 3.5 inch desktop (41650 hours music, 800,000 photo, 4100 hours digital video) to emerge in 2009. (Hitachi demo)

77. Lui et al., APL 76(2000)2749 RRAM

78. 材料电阻会因施加的电压脉冲而发生巨变（大约在 1 万～ 10 万 倍之间）的现象。停止施加电压脉冲后，可维持变化后的电阻值。 利用这种现象而形成的非挥发性内存就是 RRAM 。 十仓领导的研究小组认为，这种大范围的电阻变化可能缘于一种 名为 “ 强相关（ strongly correlated ） ” 的现象。强相关是指， 某种材料不经过半导体而在绝缘体与金属之间进行迁移。激发这种 现象的就是电压脉冲。 due to de-localization of localized valence electrons by high electric fields ( IEEE 2005 ).

79. Sharp Develops Basic Technology for RRAM, Next-Generation Nonvolatile Memory -------------------------------------------------------------------------- Dec 11,2006 A memory capable of programming data at rates about 100 times faster than flash memory. These results are the first step toward the practical use of this memory technology. It will continue in the future aimed at bringing a commercially viable product to market.

80. Philips claims 'super Flash' memory breakthrough May/2005 the 'phase-change' material changes phase from one to another using pulses of electric current. A doped Antimony Tellurium (SbTe) compound, unlike other attempts at phase-change memory, well-suited to the standard CMOS process used to make most computer chips. The material's phase change is fast, taking place in under 30ns, in the prototype cell. That is 100 to 200 times faster than today's Flash memory cells, and getting awfully close to DRAM speeds.

82. Phase change vs Magneto-optics Phase change 2001 next generation, 2003-2004 Products DVD-RAM UDO (1) nextGen DVD (2) Capacity/surface, GB 4.7 20 25 No of surface 2 2 2/4 Bit area mm x mm 0.165x0.28 0.33x0.13 0.32x0.185 Transfer rate Mb/s 1.2 4-8 4-6 Magneto-optics Products 3.5 in GIGMO (3) ID photo (4) 3.5 in GIG ID Photo Capacity 2.3 0.73 10.0 3.0 No of surface 1 1 1 1 Bit area 0.67x0.233 0.6x0.235 Transfer rate 8.38 2.5 20+ 110 (1)SONY disc diameter=13 cm, (2) Matsushta prototype, two recording layers (3) disc diameter=9cm, (4) disc diameter=5 cm. Sony: 23GB, 11Mb/s (2003); 50GB, 22Mb/s (2005);100GB,43 MB/s (2007)

83. HARM HARM+PM

85. holographic technology Two coherent beams are necessary.The one is an information beam including user data, the other is a reference beam. During the recording process, they interfere with each other, and the interference pattern is recorded in the media, called a hologram. In the reconstructing process, the information beam can be recon- structed when the reference beam incident on the hologram.

86. SUMMARY (1) Longitudinal Recording AD 120Gbits/in2 SNR ≈ B 2 W read / α 2 d 3 (1+σ 2 ), Thermal stability K u V/K B T ≌ 50 – 70 (2) Perpendicular Recording AD 421 Gbits/in2 (3) Advantage > 1Tbits/in2 Patterned media Heat assisted magnetic recording STT RAM (Nano-ring) RRAM Super Flash (PRAM) FRAM PRAM FLASH (32Gbits)

87. References (1) Magnetic Recording :Advancing into the future A.Moser, K.Takano et al., J. Phys.D:Appl.Phys. 35(2002)R157-167. (2) Longitudinal Magnetic Media Designs for 60-200Gb/in -2 Recording Gerardo A. Bertero et al., IEEE Trans. on Mag. 39(2003)651. (3) The limits to magnetic recording ---media consideration. K.O’Grady, H.Laidler J of Magn. Magn. Matt. 200(1999)616-633. (4) Recording on bit-patterned media at densities of 1Tb/in2 and beyond H.J.Richter et al., IEEE Trans on Mag 42(2006)2255. (5) Heat-assisted magnetic recording R.E.Rottmayer et al., IEEE Trans on Mag 42(2006)2417. (6) Magnetic bistability and controllable reversal of asymmetric ferromag- netic nanorings F.Q.Zhu et al., PRL 96(2006)027205. (7) Current-driven switching of magnetic layers C.Heide et al., PRB 63(2001)064424

88. Thanks !

90. Takuo Tanaka and Satoshi Kawata Three-Dimensional Multilayered Optical Memory Using Two-Photon Induced Reduction of Au3+ Doped in PMMA Akihiro Ohta, Masao Miyamoto, Yoshimasa Kawata, and Masahito Nakabayashi Multilayered Optical Memory for Terabyte Data Storage (IEEE TRANSACTIONS ON MAGNETICS, VOL. 43, NO. 2, FEBRUARY 2007)

91. Ferroelectric Memories, March 2007 The traditional FeRAMs continue to be used in applications such as smart cards and ID cards as well as targeting RFID, automotive and space. Other ferroelectric memory devices such as organic polymer ferroelectric FeFET memories are exciting new interest as well. At least 10 companies continue to work on ferroelectric memories for various applications such as fast, low power memory technology for embedding in SoC. Several novel ferroelectric memory applications have been discussed such as high density probe memory and associative memory. Multi-bit FeRAMs have been shown which could help with density and scaling issues.