L10 FePtX–Y media for heat-assisted magnetic recording


Corresponding author: e-mail dieter.weller@hgst.com; Phone: (+1) 408-717-7062, Fax: (+1) 408-717-9068


Highly chemically ordered L10 FePtX–Y nano-granular films with high perpendicular magnetic anisotropy are key media approaches for future heat-assisted magnetic recording (HAMR). They are sputtered at elevated temperature on glass disks coated with adhesion, heat sink, and texturing layers. Adding X = Ag reduces the required deposition temperature and X = Cu lowers the Curie temperature. Current seed layers are NiTa for adhesion and heat sink and well-oriented MgO (002) layers for highly textured FePtX(002) grains surrounded by Y = carbon and/or other segregants. Magnetic anisotropies larger than 4.5 × 107 erg cm−3 and coercivities beyond 5 Tesla have been achieved. The combination of thermal conductivity and Curie temperature determines the required laser power during recording. Key goals are to optimize media, heads, head-disk-spacing, and read-back channels to extend the areal density to 1.5–5 Tb in−2. pssa201329106-gra-0001

Head and media in heat-assisted magnetic recording1. LD, laser diode; TFC, thermal fluctuation control; NFT, near field transducer. 1Lidu Huang et al., “HAMR Thermal Modeling Including Media Hot Spot”, APMRC 2012.

1 Introduction

Heat-assisted magnetic recording (HAMR) is a new technology, which will increase the achievable bit or areal density (AD) in magnetic hard disk drives (HDD). High magnetic anisotropy, chemically ordered FePt will be used for new thermally stable recording media compared with CoCrPt, which is currently used in perpendicular magnetic recording (PMR). In order to write bits into HAMR media, a laser locally heats beyond the Curie temperature TC and rapid cooling in the presence of the head field produces transitions in the magnetic orientation. AD is the amount of data that can be stored in a given amount of space on a disk inside the hard disk drive. It is the number of bits per square inch and is determined by magnetic flux changes in the down track direction (fci = flux changes per inch) multiplied by the cross track density (tpi = tracks per inch). In today's PMR products the AD is ∼1700 kfci × 375 ktpi ∼ 640 Gb in−2 (Giga bits/square inch) and the bit aspect ratio BAR is kfci/ktpi ∼4.5. In a recent HAMR demonstration by Seagate and coworkers [1, 2] the AD was extended to 1 Tb in−2 and the bit aspect ratio was just a little lower, BAR ∼ 4. Higher AD means that data can be packed more tightly onto the surface of a disk, resulting in overall greater storage capacity. The measured bit error rate needs to be BER ∼ 10−2 or better. It depends on the signal-to-noise-ratio (SNR), where the signal is determined by the remanent magnetization and head-disk-spacing (HDS). The media noise is dominated by transition position fluctuations or jitter, which depends on the number of grains per bit and thermal gradient during HAMR [3, 4]. Media noise is also a function of grain cluster size and distribution, randomness of positions, texture and orientation, and magnetic uniformity of the grains that comprise the media. Highly anisotropic materials promise a significant reduction in thermally stable grain size from D = 7–9 nm in today's PMR CoCrPt media down to D = 3–5 nm in future HAMR FePt media. An extension of PMR from currently AD ∼600–700 Gb in−2 to shingled magnetic recording (SMR) up to AD ∼1.5 Tb in−2 is planned. HAMR is the main approach beyond PMR/SMR with aggressive goals of 1.5–5 Tb in−2. Alternatives to HAMR are microwave assisted magnetic recording (MAMR) [5-7], bit patterned magnetic recording (BPMR) [8-10], and two-dimensional magnetic recording (TDMR) [11, 12]. Extensions beyond 3–5 Tb in−2 will likely require the combination of HAMR and BPMR.

The HAMR media material focus as is discussed in detail in Sections 2–5 is on fully chemically ordered, high anisotropy FePt, which is well-known since1973 [13]. Possible alternatives are NdFeB [14] (1984) and/or SmCo5 [15] (1967), which were investigated early on as hard magnets but recently are also studied as thin films: NdFeB [16, 17] and Sm–Co [18, 19]. In addition, Co/Pt or Co/Pd type multilayers have fairly recently been studied for HAMR media applications [20]. The promise there is that the Curie temperatures are easily adjustable and growth at room temperature as opposed to FePt, FeNdB, or Sm–Co is appropriate. Critical HAMR FePt media requirements were discussed by Rottmayer et al. [21]. They include small grains with narrow size distribution and high temperature dependence of the switching field, which scales with the magnetic anisotropy. Reduced average grain-to-grain spacing or grain pitch of 〈Dp〉 = 3–5 nm and small grain size distributions below σD/〈D〉 = 15% are challenging because of elevated temperature growth requirements. Another challenge is the required recording (heating) temperature in HAMR, which is above the Curie temperature [22]. The fact that TC ∼85 K of disordered fcc A1 FePt is much lower than TC ∼50 K of chemically ordered fct L10 FePt [23] causes magnetic and TC distribution problems due to grain shape and size dependent chemical ordering [24-27]. The reduction of the Curie temperature of small grains is consistent with finite-size scaling theory and is smallest for grains elongated along the axis of magnetic anisotropy with an aspect ratio of δ/D = 2 [28] (δ = film thickness).

Key findings so far are that sputtered, highly chemically ordered granular L10 FePtCuAg–C media, deposited on suitable glass substrates at temperatures up to 600–650 °C are promising candidates for future HAMR products [29-31]. X-ray diffraction (XRD) and transmission electron microscopy (TEM) reveal high L10 ordering based on an A(001)/A(002) ∼ 2 ratio, an average grain size 〈D〉 = 7.2 nm and a size distribution σD/〈D〉 = 16% [29]. Magnetic characterization with a 9 T physical property measurement system (PPMS) VSM magnetometer shows perpendicular coercivity up to HC = 5.2 T, HK > 9 T, and Ku > 4.5 × 107 erg cm−3 (HK = anisotropy field; Ku = uniaxial perpendicular magnetic anisotropy) [32]. Additional measurements with a 14T PPMS VSM tool show HK ∼10 T [33]. Seagate has published spin stand data resulting in AD = 1 + Tb in−2 and a bit-aspect-ratio BAR ∼ 4 [1, 2]. TDK has more recently announced AD = 1.5 Tb in−2 with a significantly lower BAR = 1.2 [34]. Advanced Storage Technology Consortium (ASTC) HAMR roadmaps discuss BAR = 5–3 and increased linear densities to move AD from 1.5 toward ∼5 Tb in−2 in about 5 years, assuming a cumulative annual growth rate of 20–40% [35].

2 Future media requirements

HAMR media progress and achieving areal densities beyond 1.5 Tb in−2 is very challenging. Based on Wang et al. [2], current HAMR recording is limited by switching field distribution (SFD) and thermal spot size. Additional SFD sources exist in HAMR compared to PMR, making reduction critical for AD improvement. Improving both near field transducer (NFT) recording head dimensions and media thermal properties are critical to make progress. Reducing the Curie temperature distribution math formula/TC toward 2% is critical and requires improved sputter conditions, optimized seed layers, lower growth temperature, and perhaps continuous granular composite (CGC) [36, 37] and/or exchange coupled composite (ECC) type structures [38, 39]. New characterization methods including the distribution of the anisotropy field math formula [40] and correlating math formula with the grain size distribution σD are needed. Intergranular exchange, a thick media layer, canted field, low head velocity, and 850 K peak temperatures are helpful for reducing jitter, as, e.g., discussed by Victora et al. [41]. Micromagnetic modeling discusses transition broadening due to large fields at 45° angle [42]. Thermal agitation during recording dominates medium noise for small grain size media. To enable grain-pitch limited SNR, high thermal gradients >15 K nm−1 are ultimately needed [42]. Recently, Jimmy Zhu has also discussed improvements by reducing the field angle to 5°, which may be achieved by adding an SUL [42]. Finally, HAMR has additional levels of complexity beyond today's hard drives, as recently discussed by Rausch et al. [43].

2.1 Current and future media overview

Magnetic media materials discussed in data storage from current PMR/SMR technology (hexagonal CoCrPt) to future HAMR/MAMR and/or BPMR technology (L10 FePt or alternatives) are summarized in Table 1, which is an update to Refs. [44-46]. The last column of calculated thermally stable grain diameters is based on a thickness to diameter ratio δ/D = 2, which is about the highest aspect ratio to maintain perpendicular spin orientation and avoid domain formation [47-49]. The calculation is based on KuV/kT = 60 and T = 350 K (see details in the Table 1 description). The current goal in PMR/SMR media remains to reduce the grain size and grain pitch (spacing) and increase HK beyond ∼22 kOe to maintain thermal stability. The anisotropy requirement is Ku ∼ 1 × 107 erg cm−3, which is achievable in low Cr and high Pt content CoCrPt media, close to ordered hexagonal Co3Pt. At shorter than nanosecond write time the coercivity H0, is about a factor of two lower than HK based on today's recording field angle close to 45° and an orientation variation of the easy axis (c-axis) of σ ∼ 3°. In today's PMR products at AD ∼ 600–700 Gb in−2 the magnetic domain or cluster size is reduced to 10–20 nm, close to the grain pitch (spacing). It strongly depends on the inter-granular exchange coupling. Controlling the magnetic cluster size and pushing the cluster size toward the smaller grain region is critical to obtain the small grain media benefit for higher AD, as recently discussed by Ikeda et al. [50]. Thinner Ru-type seed and interlayers between the soft underlayers (SUL) and magnetic layers allow recording heads with smaller dimensions to produce sufficiently high write fields and sharp enough gradients to proceed to AD = 1–1.5 Tb in−2.

Table 1. Magnetic media material possibilities for extremely high density recording as previously discussed in 1995 [44], 1999 [45], and 2007 [46] and updated by a thickness to diameter ratio δ/D = 2
     T = 350 Kpssa201329106-gra-0002pssa201329106-gra-0003pssa201329106-gra-0004pssa201329106-gra-0005
 alloy systemmaterialKu (107 erg/cm3)MS (emu/cm3)HK (kOe)TC (K)Dp (a) (nm)Dp (b) (nm)Dp (c) (nm)Dp (d) (nm)
  1. Dp is the average thermally stable grain diameter assuming KV/kBT = 60 and T = 350 K, kB = 1.3807 × 10−16 erg K−1 and volumes (a) V = π/4 × D2 × 10 nm (cylinders), (b) V = D3 (cubes), (c) V = 4/3 × π × (D/2)3 (spheres) and (d) V = π/4 × D2 × δ (cylinders with δ/D = 2). The thickness δ is 10 nm or larger in today's media but will drop for smaller diameters going forward.
  2. aTC in today's alloy media depends on the Cr and Pt content and has increased.
  3. bTC in multilayers strongly depends on the Co thickness.
pssa201329106-gra-0007CoX/Pt(Pd) multilayersCo3/Pt101.245053.3∼700b5.
pssa201329106-gra-0008ordered Ll0/Ll1 phasesFePd1.8110032.77604.
pssa201329106-gra-0009rare-earth transition metalsFe14Nd2B4.6127072.45852.

The media requirements for HAMR exceed those of PMR/SMR. Besides grain size, shape, and distributions, thermal properties of magnetic and seed layers need to be optimized to achieve sharper transitions and sufficient SNR ratios for higher ADs. L10 FePt is currently the key HAMR media candidate in the industry [1, 2, 29-35]. Based on Table 1, a grain diameter down to 3.0 nm for δ/D = 2 is possible. L11 CoPt alloys have also high magnetic anisotropy Ku = 4.9 × 107 erg cm−3 and in principal allow small grains down to D = 3.4 nm. They have recently been discussed [51, 52]. The Curie temperature, which is 90 K higher than in L10 FePt was reduced to TC = 583 K by adding 35 at% Ni, which reduces the anisotropy in Co15Ni35Pt50 down to Ku = 1.4 × 107 erg cm−3 and increases the thermally stable grain size up to D ∼ 5.1 nm. The other high Ku alternatives are ordered Fe14Nd2B [14, 16, 17, 53-55] and/or SmCo5 [18, 19, 56-58], which have not been focused on for granular media so far, mostly because of possible corrosion issues. A recent announcement of Helium filled drives, however, may relax this constraint and possibly make these viable alternatives to FePt [59]. Fe14Nd2B, in particular, is of interest because of a low TC = 585 K, a high saturation magnetization MS = 1270 emu cm−3 and a very high magnetic anisotropy Ku = 4.6 × 107 erg cm−3 (see Table 1). SmCo5 is of less interest because of a much higher Curie temperature and lower MS. Both Fe14Nd2B and SmCo5, however, have a much more complex microstructure compared L10 FePt and are therefore primarily used for hard magnets.

2.2 CoCrPt perpendicular magnetic alloys

Figure 1 shows research progress that was made in CoCrPt based media in longitudinal and PMR. The TEM images highlight the average grain size and distribution from 〈D〉 = 12 nm, σarea ∼ 0.9 in 1999 [60] to 〈D〉 = 8.5 nm, σarea ∼ 0.6 in 2001 [61], and 〈D〉 = 8.5 nm, σarea ∼ 0.3 in 2008 [62].

Figure 1.

TEM images of progress in CoCrPtX granular media in longitudinal and perpendicular magnetic recording (Refs. [60-62]).

The 10 Gb in−2 media with 〈D〉 = 12 nm show a somewhat amorphous phase or voids between grains [60]. The 35 Gb in−2 media are based on CoCrPtB and use oxide segregants [61], achieving reduced grain size and exchange coupling. Today's media have a similar, unchanged grain size but reduced grain size distribution. The goal remains to further reduce the grain size, minimize distributions, grain boundary spacing, and exchange coupling [62, 63].

2.3 FePt self-ordered magnetic array (SOMA) potential

Simultaneous nucleation and growth in physical vapor deposition (PVD) leads to log-normal distributions, which is a fundamental problem. The challenge remains to move to novel, mass production compatible, deposition techniques. In that regard, monodisperse FePt nanoparticles and ferromagnetic nanocrystal superlattices, i.e., self-organized magnetic arrays (SOMA) were discussed early on in 2000 [63-71]. Figure 2 shows the chemical synthesis technique, which is based on high temperature solution phase decomposition of Fe(CO)5 and reduction of Pt(acac)2 in the presence of stabilizers. Oleic acid and oleyl amine are used to produce FePt nanoparticles.

Figure 2.

Schematic illustration of FePt nanoparticle formation from decomposition of Fe(CO)5 and reduction of Pt(acac)2 (adapted from Ref. [25]).

Figure 3a and b shows FePt TEM images of 6 nm [64] and 4 nm [66] nanoparticle arrays. They are monodisperse with low σarea ∼ 5%.

Figure 3.

TEM images of monodisperse L10 Fe55Pt45 nanoparticle assemblies with (a) D = 6 nm (left) [64] and (b) D = 4 nm (right) [66] diameter. Note that σarea ∼ 5% in both cases (adapted from Refs. [64, 66]).

The FePt composition is tuned by adjusting the molar ratio of Fe(CO)5 to Pt(acac)2. However, as synthesized FePt nanoparticles possess disordered fcc structure and are superparamagnetic at room temperature. Thermal annealing at 585 °C induces high level fct L10 ordering and a coercivity up to ∼9 kOe [64], but it also leads to grain growth and sintering or coalescence [71]. The composition was optimized to Fe55Pt45 [64-66]. Interestingly, segregation of Pt to the surface of small, ∼3 nm FePt nanoparticles, changes the core composition to Fe52Pt48 [72, 73]. Progress to move to highly textured, cylindrical, and spatially separated grains remains very challenging but is critical to use FePt SOMA for future HAMR and/or BPMR media.

3 Sputtered HAMR media

Today's focus is magnetron sputtering to generate fiber-textured FePt media with small grains and high magneto-crystalline anisotropy. The crystalline orientation is controlled by epitaxy between seed and magnetic layers. Correct choice of segregants and sputter conditions optimizes grain boundaries and grain shape. The primary difference between fct FePt and hexagonal CoCrPt is elevated temperature growth (∼600 °C) and/or rapid thermal annealing (RTA) at temperatures up to 800 °C, which are needed to achieve L10 chemically ordered FePt grains with high magnetic anisotropy [74, 75]. Even though the L10 FePt phase has a higher activation enthalpy compared to the disordered A1 phase [76], films deposited at room temperature form a low magnetic anisotropy disordered A1 phase alloy; the gain in energy of the ordered L10 phase is not sufficient to allow atoms to occupy the correct L10 lattice sites due to a lack of Fe and Pt atom mobility at room temperature.

Perpendicular magneto-crystalline anisotropy up to Ku ∼ 5 × 107 erg cm−3 has been achieved in granular L10 FePt films [22] allowing thermal stability down to D ∼ 3.3 nm diameter grains close to D = 3 nm for Ku = 7 × 107 erg cm−3 in Table 1. Elevated deposition temperature growth effects, such as grain growth and coarsening are significantly enhanced at higher temperatures. The growth regime with fixed numbers of nucleation sites, common for current low temperature deposition processes, can be significantly modified by high atom mobility at elevated temperature, thus increasing the grain size. Apart from the complex microstructure effects, chemical reactivity of the FePt is enhanced with increased temperature. Note that the desired high degree of L10 chemical ordering can be quickly degraded if Fe is participating in other than the L10 phase, e.g., due to Fe oxidation. Another process is chemical interaction with the segregant material, if it is partially incorporated into the core of FePt grains. The right combination of segregants, deposition parameters and deposition temperature to optimize the properties of granular FePt films is the primary challenge in HAMR media fabrication. Progress, including media design and recording modeling are key ongoing activities [18, 77], which are discussed next.

3.1 FePt media design

A typical FePt media design is shown in Fig. 4. High temperature glass substrates available from different companies (e.g., Hoya, Asahi, Ohara) allow growth and annealing up to 600–650 °C. The choice, combination, and thickness of seed and heatsink layers is critical to optimize FePt growth and thermal conductivity. Various options have been tried since the early 1990s to develop and optimize granular L10 FePt media [78-80].

Figure 4.

Basic HAMR media design including granular FePt separated by segregants. Seed layers establish FePt(002) texture. Heatsinks in combination with seed layers establish thermal gradients and write power requirement (see below).

Many seed layers promoting L10 FePt (001) perpendicular orientation with (200) textured underlayers have been studied, including Pt/Cr [80], CrRu [22, 81, 82], RuAl [83, 84], TiN [84], Cr [84], CrMo [85], Ti [86], TiC [87], Ag [88-90], and MgO [91-93]. The current focus is 5–15 nm thick fcc MgO (200) seed layers with <001> orientation and 9% mismatch with FePt.

Heatsink layers are critical to control the thermal conductivity during HAMR recording: Ag [88, 89, 95, 96], Au [95], Cr [87, 88], and NiTa [29, 96] are candidates. NiTa is also used as a proper adhesion layer on glass.

FeXPt alloy additions to optimize/reduce the growth temperature and Curie temperature are Au [95], Ag [29-33, 95, 97], Ni [98, 99, 103], and Cu [94, 100-103]. Five to 10 at% Ag alloy additions are typically used to reduce the deposition/annealing temperature and primarily Ni and Cu are used to reduce TC. In 2002, Thiele et al. [99] discussed (Fe1−xNix)Pt alloy compositions and showed that x ∼ 15 at% Ni reduces the Curie temperature down to ∼600 K. The anisotropy drops 43% from Ku,max ∼4.6 × 107 erg cm−3 to Ku ∼2.6 × 107 erg cm−3 [99]. Modeling in 2004 suggested that adding Cu instead of Ni up to 20 at% only marginally reduces Ku [100]. In 2011, Wang and Barmack [103] discussed the experimental impact of ternary additions of Ni and Cu on the A1 to L10 transformation in FePt films. They showed that only 5–8 at% Cu compared to 8–12 at% Ni are necessary to achieve TC ∼ 600–650 K, indicating that Ku may be kept at ∼4 × 107 erg cm−3 with a relatively small amount of either Cu or Ni [99, 103, 104]. The focus today is to add ∼5–8 at% Cu to FePt to reduce TC down to 600–650 K.

FePt-Y segregants are needed to decrease the grain size and exchange coupling. Many segregants have been tried including SiO2 [96, 97, 105, 106], TiO2 [96, 107], AlOx [88, 108], Al2O3 [109], AlN [110], TaN [111], Ag [93, 94], MgO [80, 82], Ta2O5 [112], W [113], Ti [114], B2O3 [115], B/Ni [116, 117], C/BN [116], SiNx [118], SiNx/C [119], and C [29-33]. Today's focus is on 30–50 at% C [29-32] and similar levels of SiO2 or TiO2 [96, 97].

Recently FePt–C films as templates for FePt–Y (Y = SiO2 or TiO2) on double seed layers (e.g., MgO/MgO–TaO) have been examined [106]. SUL, as indicated by Wang et al. [2] will help sharpen transitions during recording. ECC granular/continuous media are well known for thermally stable PMR [120-123]. Regarding HAMR, there are many ECC activities aiming to optimize media. In 2003, Thiele et al. [124] discussed FePt/FeRh exchange spring structures, which were modeled by Guslienko et al. [125]. In 2005, Suess discussed exchange spring recording media for areal densities (AD) up to 10 Tbit in−2 [126]. Many other options were discussed in the literature [127, 128, 133], including FePt/[Co/Pt] × N [129, 130] and FePt/Fe [131-134].

In general, it is important to understand complex, inhomogeneous ferromagnets [135], micromagnetism, and microstructure [136], and pinning of domain walls in composite particles [137] as modeled by Kronmüller [135].

3.2 FePt sample preparation

FePtX–Y (X = Ag, Cu; Y = C, …) samples were prepared by sputter deposition in a high-throughput multi-chamber industrial tool based on Intevac Lean 200. In order to obtain FePtX–Y films with (001) texture having a magnetic easy axis perpendicular to film plane, a ∼10–15 nm (002) textured MgO seed layer was deposited on a NiTa adhesion layer at room temperature. Chemically ordered L10 FePt is a tetragonally distorted lattice with c/a = 0.96, where a = 0.386 nm is measured in the plane of alternating Fe and Pt layers and c = 0.370 nm is orthogonal to these planes. The magnetic easy axis is oriented along the (001) direction. In order to realize films with perpendicular magnetic anisotropy, the grains of L10 FePt must be oriented such that the c-axis is perpendicular to the film plane. The MgO seed lattice parameter (0.412 nm) is larger than the FePt lattice parameters. The 9.7% lattice mismatch at the MgO/FePt interface creates tetragonally distorted FePt nucleation sites, which stimulate the formation of the L10 FePt phase with desired (001) texture and leads to a reduction in the ordering temperature, down to ∼600–650 °C. The high melting point of MgO (2852 °C) is advantageous since it allows for deposition at elevated temperatures without intermixing at the MgO/FePt interface. The media was fabricated on glass substrates compatible with the high temperatures used during growth. The composite film was co-sputtered in a triatron source from FePt, Ag, Cu, and Carbon + Y segregant targets.

3.3 FePtX–Y texture and L10 ordering (XRD)

The presence and degree of L10 ordering is established by XRD on the FePtX–Y based media, as shown in Fig. 5. The FePt(002) diffraction peak is well-defined indicating a high degree of texture. The superlattice FePt(001) peak, indicating L10 chemical ordering is present at ∼23.7°, while the wide background peak between 17° and 35° corresponds to the amorphous glass substrate. The high degree of chemical ordering of L10 FePt manifests itself as a large integrated peak intensity ratio A(001)/A(002) ∼ 2 [139]. For single crystal L10 FePt ordered films, the intensity ratio corresponding to complete chemical ordering can be calculated from diffraction equations and is ∼2. However, sputtered films usually are not perfectly textured, i.e., there is a c-axis spread around the direction perpendicular to the film. Thus, for samples shown in Fig. 5 the full width half maximum (FWHM) of the rocking curve is ∼6.5°. Taking into account finite rocking curve widths and measurements of instrument parameters as suggested by Yang et al. [138] yields an ordering parameter of S ∼ 0.45 × [A(001)/A(002)]1/2 = 0.9 for these films [139]. Apart from mis-orientation of the easy axis, randomly textured grains can be present in the film as well, for example in case of the grain formation away from the texture defining MgO interface or when MgO/FePt lattice mismatch does not provide enough strain to enforce proper orientation. Presence of these mis-oriented grains can be suppressed in part by optimizing deposition parameters and segregant amounts. In Fig. 5 an improvement in the amount of textured material for the same nominal thickness of the FePtX–Y layer manifests itself as increase of the total intensity of the textured FePt(001) and FePt(002) peaks.

Figure 5.

Out of plane XRD spectra for three samples with same FePtX–Y thickness δ ∼ 7 nm. Variation of the deposition parameters allowed to suppress randomly textured grains of the FePt films, which is manifested by increased XRD intensity of the textured FePt(001) and FePt(002) peaks.

Another structural defect that is common in textured L10 FePt alloys is orientation of the easy magnetic c-axis in the in-plane direction of the film. The 9.7% lattice mismatch at the MgO/FePtX + Y interface allows to partially suppress this, so that the in-plane variant is not visible in Fig. 5. However, in-plane XRD scans shown in Fig. 6 highlight in-plane orientations and indicate some misorientation of ordered L10 FePt, which results in undesired magnetic behavior and noise during recording (see Section 4). To obtain better textured films additional work on underlayers, which could enforce better c-axis orientation, is ongoing.

Figure 6.

Typical in-plane XRD spectra measured for δ ∼ 7 nm thick L10 FePtX–Y alloys. Note that apart from FePt(110) and FePt(200) peaks expected for well-textured films with c-axis perpendicular to the film, the presence of in-plane FePt(001) and FePt(002) intensities explain why variants of the hard magnetic axis are observed (see Section 4).

3.4 FePtX–Y microstructure (TEM)

In Fig. 7 the microstructure of media with carbon segregant is shown [29]. Due to de-wetting, the FePtAg–Y film is broken into smaller and smaller grains with increasing segregant volume fraction in deposited films, forming small isolated grains, as shown in plan-view transmission electron microscopy (PV TEM), Fig. 7a (top). FePt average core grain sizes of 〈D〉 = 7.2 nm with a narrow grain size distribution of σD/D = 16% was obtained. Segregant volume fractions exceeding 35% are needed to achieve these small grains and good grain segregation. However, the disadvantage of large amounts of segregant material is the presence of very small grains with grain diameters below 3 nm (see histogram in Fig. 7), which are not included in the fit for grain size distribution. As seen in the inset to Fig. 7, small grains are numerous; however, some of the small grains counted could arise as a result of the image processing technique that can erroneously detect uneven contrast in the background of the PV TEM image. Formation of very small grains can also be governed by the inherent properties of the segregant/FePt interface, which also affects the grain shape. The presence of spherical FePtAg–Y grains can clearly be seen in Fig. 7c, which shows a cross-sectional TEM image. This particular grain structure limits the thickness of the magnetic layer, since a second layer of disoriented grains can form when the FePtAg–C layer reaches a critical thickness of ∼7 nm. Limiting the thickness of the magnetic layer can reduce the possibility of forming a second layer of grains, but often larger lateral grain size will be needed to support thermally stable grains.

Figure 7.

(a) Plan-view TEM image showing granular δ ∼ 7 nm thick FePtAg–C media. (b) The inset shows a histogram of the grain size distribution. The red line is a log-normal fit resulting in 〈D〉 = 7.2 nm σD/D = 16%. (c) The cross-sectional TEM image shows spherical grain shapes [29].

To avoid this problem one needs to find segregants that are capable to maintain small grains at elevated deposition temperatures and, in contrast to just carbon, will not force spherical grain shapes. In Fig. 8 granular FePtX–Y media with mixed C + Y segregants are used. As one can see, the grain shape is significantly different from pure C and one can model a grain structure with a Voronoi-type of microstructure. Voronoi means that the grains fill the space more fully than in cases where they are more round [140]. This property of C + Y segregants, in principle, allows to maintain a constant grain boundary width and to suppress very small grains. However, as can be seen in Fig. 8a, a microstructure with enhanced numbers of small grains was obtained. The histogram in Fig. 8 shows that in the case of C + Y segregants, the grain size is significantly modified and leads to a bimodal distribution with an undesired population of grains centered at ∼2.9 nm and a very broad distribution of grain sizes. This population of grains will have a negative impact on magnetic properties, since thermal stability criteria are not met for <3 nm grains. For good read/write performance one needs to obtain a grain microstructure in FePt similar to what is used in today's PMR media, with small grain size distribution and thin uniform grain boundaries, see Fig. 1 (PMR section).

Figure 8.

About 7 nm thick FePtX–Y granular media with carbon–Y segregants. Note that (a) the addition of Y allows to reduce the grain size, (b) however, leads to a bimodal distribution of grains with a population of small grains.

Therefore, variations of the deposition process parameters and C to Y ratio as segregants were investigated. In Fig. 9a, a sample with optimized deposition parameters and C + Y segregant is shown. In contrast to Fig. 8, the grain structure is more uniform as confirmed by grain size analysis. Even though distribution of grain diameter is somewhat wider than for pure C, grain pitch (average distance between two neighboring grains) is smaller and has a narrower distribution due to improved grain boundary thickness uniformity.

Figure 9.

About 7 nm thick FePtX–Y granular media with carbon segregant (Y=C). Growth parameter optimization and C/Y ratio allowed suppression of bimodal grain size distribution, while maintaining a small grain size. Note that the intergranular segregant material has a close to constant width.

So far, for all examples fairly thin (∼7 nm) granular FePtX–Y layers were used. In the case of just carbon segregants, spherical grain shapes are a limiting factor, and increasing film thickness results in second layers of non-textured grains. The thickness of the films is an important tuning parameter for grain volume, which affects the thermal stability. Apart from thermal stability, the total magnetic volume of the grains forming a bit defines the read-back signal after recording. Therefore increasing the thickness of the media is beneficial for the SNR ratio, i.e., the recording performance. As can be seen in Fig. 9 for C + Y segregants, the shape of the grains (at least in lateral direction) does not have spherical limitations, therefore an increase in film thickness should be possible.

In Fig. 10 we show FePt–Y granular media with magnetic film thickness increased to ∼10 nm. Plan-view and cross-sectional TEM data show that the grain aspect ratio δ/D is close to 1.5, while the average lateral grain diameter is kept at 〈D〉 ∼6.3 nm. Note that the cross-sectional image in Fig. 10 shows small random grains, which are present in the film as well (blue-dashed circles in Fig. 11). These defects will have to be suppressed for future granular FePtX–C + Y (X = Cu, Ag) media.

Figure 10.

FePt granular media with increased thickness. (a) Plan-view TEM image, (b) histogram of the grain size distribution, and (c) cross-sectional TEM image. Note that grains have an aspect ratio of δ/D ∼ 1.5 with film thickness of ∼10 nm and grain diameter of 〈D〉 ∼6.3 nm. Magnetic media thickness δ ∼ 10 nm.

In Fig. 11 a comparison of cross-sectional images for pure C and C + Y segregants is shown. As one can see, introduction of additional segregants allows suppression of undeveloped grains at the MgO/FePtX + Y interface. This effect is achieved by modifying the surface energy of the segregant. Columnar grains with small lateral grain size allow reduction of the amount of segregants in the film and lead to improved magnetic filling factors of the film. These changes in microstructure are needed to achieve high density HAMR recording.

Figure 11.

(a) and (b) Magnified cross-sectional TEM images from Fig. 7 (C segregant) and 10 (C + Y segregant), respectively. Note that the thickness of the film is increased from δ ∼ 7 nm in (a) to δ ∼ 10 nm in (b). Undeveloped grains at the FePtX + Y/MgO surface, which are marked with red dashed circles, are suppressed.

In summary of this section, we showed that the microstructure of FePt granular media is very different from current PMR media due to a different seedlayer structure and high deposition temperatures. On the other hand it can be steered in the desired direction by choosing proper segregants and optimizing deposition parameters, which is an ongoing activity.

4 Magnetic characterization

Owing to the large Ku of ordered FePt, textured granular thin films of magnetically isolated nanoparticles with sizes well below 10 nm can be fabricated to be thermally stable. In the absence of inter-granular exchange interaction, the room-temperature switching fields can be as large as ∼0.5 HK or more [29], necessitating the use of magnetic characterization tools with superconducting magnets of sufficient strength or carrying out measurements at elevated temperatures [141]. Carrying out magnetometry measurements at temperatures up to the Curie point of FePt (770 K) can be challenging, in view of the vanishing magnetization to be measured, limitations in sample measurement geometry, or length of exposure to high temperature, which can alter the pristine state of the sample even in controlled, inert environments. Nonetheless, such careful measurements can be attempted and have been reported by Bublat and Goll on FePt thin films and nanostructure systems [142]. Here we focus on more typical measurements carried out in a quantum design PPMS equipped with a 9 T superconducting magnet, and overview the typical switching behavior of FePt granular films. We employ minor loop analysis to extract magnetic cluster size and investigate the origin of the room-temperature SFD. Though the magnetic properties near the Curie point are more relevant to HAMR, room-temperature characterization allows a more direct access to sample characteristics that needs improving going forward.

4.1 Hysteresis, SFD and minor loops

Figure 12 shows perpendicular and in-plane magnetization loops of four separate samples highlighting several features that are typically optimized for HAMR media. The left panel in Fig. 12 shows two samples before and after optimization of the deposition conditions that affect the FePt film nucleation and the segregant amount. The representative film in black shows high coercivity of ∼4 T, indicating a large fraction of hard, thermally stable grains textured to have a c-axis out-of-plane. However, it also shows poor remanent magnetization and nucleation fields, which need to be improved to maintain high read-back signal and reduce adjacent track erasure, respectively, during HAMR. In the absence of inter-granular exchange, remanence and nucleation field are affected by dispersions in Ku, volume and c-axis alignment of the grains in the film. The reduction in magnetization before reaching zero field is an indication of having some grains with poor texture, whereas the soft phase near zero field is an indication of the presence of grains with low Ku × V products, some of which may be superparamagnetic.

Figure 12.

(left) Perpendicular magnetization loops of a representative FePtX–Y film (black) and of another film with optimized nucleation layer and segregant amount (red). (right) In-plane magnetization loops of a representative FePtX–Y film (black) and of another film with optimized thickness (red). Note that the right panel shows different samples than the left panel.

The presence of superparamagnetic grains can be inferred by TEM imaging, which shows grains below 3 nm in size (see Section 3). Changing the deposition parameters to increase grain size and improve c-axis alignment shows improvement, shown in the red curve, left panel of Fig. 12. The right panel in Fig. 12 shows two other samples of different thicknesses. If the deposited film thickness is above what the microstructure can support in a single layer of textured grains, then a second layer of ordered, but untextured grains, can grow on the first. Depending on the amount of second-layer grains and their volume, their contribution to the magnetization loops will differ: the more second-layer grains, the more they contribute to the early magnetization loss in the easy axis and to hysteresis in the hard axis.

Figure 13 shows major and minor magnetization loops of a HAMR medium. Although the magnetic hysteresis is not optimal, as explained above, it serves as an example to further investigate the switching behavior. In the mean field approximation [143], the coercive field recoil loop (in red, Fig. 12) is representative of the switching behavior for 50% of the grains (easy switchers) subject to zero mean demagnetizing field from neighboring grains. Comparing this recoil loop with the major loop in the 1st and 4th quadrants (see offset recoil curve in red, Fig. 12) allows to extract the parameters ΔHint and ΔHext, which can be related to the SFD due to intrinsic or extrinsic effects, respectively. Intrinsic contributions arise from distributions in grain size, Ku and orientation, whereas extrinsic contributions arise from the effect of neighboring grains, such as dipolar or exchange interactions. ΔHext can be used to derive the magnetic cluster size [144], which is a measure of the magnetic resolution that the medium can support. In this example, the magnetic cluster size is ∼14 nm, larger than the mean grain pitch of 8.2 nm, indicating some degree of inter-granular exchange. As just discussed, cluster sizes are obtained from ΔHext. The procedure can be seen in Ref. [144]. ΔHint is related to the standard deviation of intrinsic SFD as σint = ΔHint/1.35 = 15.3 kOe (∼30% of HK). Having access to independent characterization of grain size distributions (through TEM analysis) and c-axis orientation distributions (through XRD analysis), we divide the contributions to σint into several components: math formula, σvolume, and σaxis [32]. Typically it is found that the dispersion in grain size is only a small contribution to SFD (σvolume ∼4–5% of HK). Note that σvolume here refers to the contribution of volume to the SFD, not a physical σvolume extracted from TEM. Grain size contributions to the SFD in HAMR media are not expected to be very relevant until the grain sizes reach the edge of thermal stability, and in this case the Ku × V product of the HAMR medium is >100 kBT. A larger contribution to the σint is found to come from the c-axis orientation dispersion, where an XRD rocking curve width of 5–7° can account for a dispersion of σaxis ∼6–12% of HK. Finally, excluding significant additional contributions to the SFD not mentioned here, we can deduce the dispersion in HK by assuming that math formula, typically yielding the biggest contribution to the intrinsic SFD (15–35% of HK).

Figure 13.

Major and minor magnetization loops of a representative FePtX–Y film (black). The recoil loop obtained in the zero mean demagnetizing field (red) is compared with the major loop to obtain ΔHint and (by offsetting it) ΔHext.

4.2 Alloy composition (Fe:Pt, Cu)

In Fig. 14 we show the effect of alloy composition for FexPt1−x–Y (Y = 27 at% C) segregated HAMR media on HC, HK, and saturation magnetization of the alloy (grain core MS). HK is maximized near 50:50 FePt composition where the L10 FePt phase is formed, though we can only provide a lower bound for HK here, due to the lack of sufficient fields to fully saturate the medium in the hard axis (HK > 9 T). As the composition moves away from the optimum, the HK decreases as does the HC, as one would expect when other microstructural parameters (such as grain size and texture quality) are held constant. The grain core MS is found to be linear with the Fe content as expected. As seen in the figure, variations in alloy composition allow for some flexibility in tailoring the magnetic properties of HAMR media. For example, moving to Fe contents above 50% can increase the magnetization of the film to improve read-back signal without a drastic compromise in the HK.

Figure 14.

Dependence of magnetic properties of FexPt1−x–Y segregated HAMR media as function of x. The case for x = 0.5 has a lower bound estimate for HK due to lack of sufficient field in the instrument (i.e., HK > 9 T).

Further optimization of the alloy composition can take place by the addition of other elements, as shown in Fig. 15 for Cu addition in continuous FeCuPt films [100-102]. Here we can see that the addition of Cu is effective at lowering the Curie temperature. The variation of TC with Cu content is not linear, since variations in composition result in differences in the amount of L10 ordering. Also the TC of ordered and disordered FePt phases differs (see discussion below). Depending on the relative balance of Fe and Pt in the ternary alloy a compromise can be made between MS, HK, and TC. Example: tune Fe:Pt balance for HK and MS choice, tune Cu to modify TC.

Figure 15.

Temperature dependence of the magnetization for several continuous films of varying FeCuPt composition, highlighting the ability to alter the Curie temperature.

5 Micromagnetic modeling

Micromagnetic modeling has traditionally been used to characterize magnetic recording media and optimize the magnetic recording process. For HAMR recording, the conventional micromagnetic approach using the Landau–Lifshitz equation is not suitable since this equation does not allow the magnitude of the magnetic moment to vary whereas both experiment and simulations typically agree on the variation with temperature, for example [99, 145]. An extension of Landau–Lifshitz equation for arbitrary temperatures (at and above the Curie temperature) is an active area of research. One popular approach is the so-called Landau–Lifshitz–Bloch (LLB) equation [146, 147]. LLB is derived starting from either classical or quantum spins interacting with a bath described by stochastic Langevin fields and mean field approximation for spin–spin interactions. By requiring a Boltzmann distribution at high temperature, a consistent stochastic LLB can then be identified [148]. Xu and Zhang [149] gave an alternative, though almost equivalent approach starting from a quantum kinetic approach with the instantaneous local equilibrium approximation and concluding with a different stochastic equation. We do not report recording simulations here, but rely on results provided by literature.

In the next section we will outline the key media parameters that affect magnetic recording. We will discuss micromagnetic modeling in order to understand some of the features of the easy axis hysteresis curves (as in Section 4.1) and to predict the expected behavior for varying average grain size for future HAMR media.

5.1 Key HAMR media parameters for magnetic recording

From recording simulations, it is clear that the main media parameters for good HAMR recording performance are the Curie temperature distribution [150], grain density, and narrow grain size distribution [18]. Recording simulations also clearly show that the temperature profile in the recording layer strongly affects the quality of magnetic recording. For a given head design and media properties, there exists an optimal magnetic write field strength and lateral thermal gradient. The magnetic field is required to set the orientation of the field-cooled magnetization while a strong lateral thermal gradient insures the correct position for writing. It will be pointed out that a large lateral thermal gradient can also alleviate recording degradation due to the Curie temperature distribution. However, too large of a thermal gradient can lead to the so-called paramagnetic trap – where the magnetization cools down too fast and may be trapped in the wrong orientation.

The Curie temperature distribution math formula may be the main media parameter for HAMR. Larger distributions math formula have a strong adverse effect on recording performance [150]. math formula arises from non-perfect chemical ordering and finite size grains. If the material is also doped in order to reduce the Curie temperature, the variance on doping will also introduce an additional distribution in Curie temperature. Experimental results [25] show a clear correlation between Curie temperature and chemical ordering. Furthermore, both experiment [25] and “atomistic” LL simulations [28] show as the size of the grains decrease, so does the Curie temperature as a power-law dependence. Hovorka et al. [26] have used these results to relate the grain size distribution to the Curie temperature distribution.

In order to localize heating, thermal management is a new requirement for HAMR media. Ideally, the temperature profile needs to be well-localized in order to write narrow tracks and sharp transitions [21, 27]. This is achieved by the correct combination of recording head and media. Solution of Maxwell's equations will determine where the heat is deposited in HAMR media for a given head design. Given where this power is deposited, solving the heat equation in the HAMR media will then give the resulting temperature profile as a function of space and time. The complexity and coupling of these two problems generally does not allow one to make general statements. Regardless, the heat flow in HAMR media can be controlled by introduction of thermal barriers, anisotropic thermal conductivity and thermal sinks in order to achieve desired temperature profiles in the recording layer. Compared to conventional recording, this puts new constraints on any future HAMR media.

In the next section we will discuss micromagnetic modeling in order to understand some of the features of the easy axis hysteresis curves (as in Section 4.1) and to predicted the expected behavior for varying average grain size for future HAMR media.

5.2 Current modeling results

Figure 16 shows the micromagnetic easy axis hysteresis loop for for FePt with either a log-normal grain size distribution or bi-modal grain size distribution consisting of the same log-normal grain size distribution along with a Gaussian distribution centered at D = 2.5 nm with standard deviation of 0.5 nm. The small grains comprise 3% of the total volume of the hard layer. We see that by inclusion of the small grains, we recover the dip near zero applied field. Clearly these small grains, even if perfectly chemically ordered, are either marginally stable or super-paramagnetic. Since these grains reduce the remnant magnetization, they are undesirable for HAMR recording.

Figure 16.

Predicted easy axis loops for uni- or bi-modal grain size distribution. Small grains (D ∼ 2.5 nm) produce a dip near zero applied field.

In Fig. 17, we model the expected easy axis hysteresis loop for various grain sizes. Here we assume a room temperature HK = 90 kOe (as in Fig. 14) and a constant grain aspect ratio of δ/D = 2. We find once the average grain diameter is less than 5 nm there is a strong reduction of the nucleation and coercive fields. In fact, for the smallest grain (3 nm) the remnant magnetization is unacceptable for magnetic recording. Even for D = 4 nm, the nucleation field is marginal for good recording. Increasing HK to the theoretical maximum (12.2 T) would eliminate these problems; however achieving this maximum in such small grains is very challenging.

Figure 17.

Predicted easy axis loops for varying average grain diameters at constant grain aspect ratio of 2.

So far, we have restricted ourselves to a single recording layer. However, there is no a priori reason for this. One can consider more elaborate media structures such as ECC-type or multi-layered media. By introduction of exchange, both lateral and vertical, in these composite recording layers may help overcome some of the current limitations. One could envision layers with different Curie temperatures that may enhance recording performance. However, even simulations of these structures are difficult simply due to the lack of knowledge of how various parameters change with temperature.

6 Summary

Noise performance and spatial resolution are key parameters in recording media and are ongoing challenges in advancing the AD. The dominant media noise source today is transition jitter. In sputtered media, it reflects the finite size, random positioning, and dispersions in size, orientation, and magnetic properties of the fine grains that comprise the media. Highly anisotropic materials, combined with HAMR, promise significant reductions in the average, thermally stable grain size from currently about 7–9 nm in PMR Co-alloys to about 3–5 nm in chemically ordered L10 FePt-based HAMR media. HAMR limitations and extendibility are studied in light of the recent Seagate 1 + Tb in−2 technology demonstration [1, 2]. Besides magnetic field component dependence on magnetic media properties, thermal spot size and media temperature rise strongly depend on the media thermal conductivity. Media are optimized by growing granular L10 FePt on a relatively thin thermal resistor seed layer on top of a thicker thermal conductor or heatsink layer. In addition SUL are suggested to optimize the magnitude and angle of the magnetic write field. The achievable AD depends on the NFT dimension and the media thermal properties, which require optimization. The main conclusions are that besides grain size, shape, and distribution requirements, thermal properties of both the magnetic and the seed layers need to be improved and optimized to achieve the required SNR ratio for higher AD. HAMR based on L10 FePt media is the key technology that the industry is currently focusing on. MAMR [5-7] and two dimensional magnetic recording (TDMR) [11, 12] are possible alternatives. Extensions beyond 3–5 Tb in−2 will likely require the combination of HAMR and BPMR, which has already been achieved at 1.5 Pb m−2 (0.97 Tb in−2) [151]. Roadmaps and head, media and head-disk-interface requirements and projections are discussed by the ASTC [38, 152].


The authors would like to thank Hans Richter, Olav Hellwig, Barry Stipe, Lidu Huang, and Jim Lyerla for support and suggestions. Many thanks to Gregory Benson, Raymond Kappes, and Chu Tran for improving the Lean 200 sputter tool.


  • Image of creator

    Dieter Weller received a Ph.D. in physics in Germany in 1985 and worked for Siemens, IBM Almaden Research, Seagate R&D and is now at HGST, a Western Digital company. His efforts include the fabrication, characterization, electronic, magnetic, and magneto-optical properties of thin films and multilayers relevant for magnetic recording. He is Fellow of the American Physical Society (APS), published 260 scientific papers and holds more than 60 U.S. patents. His current focus is on heat-assisted magnetic recording.