Magnetic dot arrays with multiple storage layers

Magnetic dot arrays with multiple storage layers

Magnetic dot arrays with multiple storage layers

M. Albrechta! Department of Physics, University of Konstanz, D-78457 Konstanz, Germany

G. Hu and A. Moser Hitachi San Jose Research Center, 650 Harry Road, San Jose, California 95120

O. Hellwig Berliner Elektronenspeicherring-Gesellschaft für Synchrotronstrahlung (BESSY), Albert-Einstein-Strasse 15, D-12489 Berlin, Germany and Hitachi San Jose Research Center, 650 Harry Road, San Jose, California 95120

B. D. Terris Hitachi San Jose Research Center, 650 Harry Road, San Jose, California 95120

An approach to increasing the data storage density of magnetic recording was investigated wherein dot arrays are combined with multiple magnetic storage layers. The latter consists of two magnetically decoupled perpendicular Co/Pd multilayer stacks. As a result of the difference in the coercivity between the two stacks, the orientation of the remanent magnetization of each stack can be set independent of the orientation of the other layer. Therefore, each dot allows storing four different magnetization states, which give rise to four different readback signal levels. Thus, the investigated media structure allows doubling the storage density in magnetic recording applications. It was found that even for thick spacer layers a weak magnetostatic coupling of the storage layers is present preferring ferromagnetic alignment of the storage layers. © 2005 American Institute of Physics. fDOI: 10.1063/1.1904705g

In conventional magnetic recording systems the recording medium consists of weakly coupled magnetic grains, where the required signal-to-noise ratio sSNRd needed for high-density recording is achieved by statistically averaging over about 100 grains/bit. In order to increase the areal density, the traditional engineering approach requires that all design parameters of the head and the recording media be scaled to smaller dimensions. These scaling laws involve a reduction of the recording layer magnetic thickness and the grain diameter to maintain SNR. As the volume of the grains is reduced in the scaling process, the magnetization of the grains may become unstable due to thermal fluctuations, and data loss may occur. This phenomenon, also referred to as the “superparamagnetic effect,” has become increasingly important in recent years, as new magnetic hard disk drive products are designed for higher areal densities.1

Patterned magnetic media, where a single magnetic bit is recorded on a predefined, single-domain magnetic dot in an array, is one of the proposed approaches for extending magnetic storage densities beyond the limit set by thermal decay.2,3 In this recording schema the thermal stability is given by the increased volume of an individual dot, which is much larger than the grain volume in a conventional recording medium. However, in order to achieve 1 Tbit/in.2,a nanostructure array with a period of 25 nm over a fullpatterned disk stypically 1–2.5 in. in diameterd is required. Despite the fact that extensive studies have been performed on the fabrication methods4–6 supporting bit densities of up to 200 Gbit/in.2, large area ultrahigh-density magnetic patterns with low defect rates and high uniformity are still not available today.

Here, we report on a different approach to increasing the storage density by combining patterned media with multiple independent magnetic storage layers. The latter can be achieved using several magnetically decoupled layers, where the magnetization of each layer can be switched independently.7 Therefore, the magnetization integrated over all storage layers stotal number nd gives rise to a possible 2n different signal amplitudes for each dot, which can be read back by a conventional giant magnetoresistive sGMRd sensor used for magnetic recording applications. Thus the recording density is increased by a factor of 2sn−1d, e.g., 1 Tbit/in.2 could be achieved using two storage layers per dot with a more feasible period of 40 nm instead of 25 nm using a single layer.

To demonstrate the feasibility of the multilayer storage technique, we studied a bilayer magnetic dot array with a period of 300 nm. A SiO2/Si substrate was topographically patterned by e-beam lithography followed by reactive ion etching.8 The resulting SiO2 pillars were 150 nm in diameter and 80 nm in height. The magnetic films were deposited onto the pillars where the walls of the pillars act to isolate the pillar tops from each other leading to exchange-isolated single-domain magnetic dots.

In order to achieve four independent magnetic states, there are two requirements on the film properties. First, the adAuthor to whom correspondence should be addressed; electronic mail:

two magnetic layers have to be magnetically decoupled, which requires a thick nonmagnetic spacer layer. Second, the switching fields of the two layers have to be well separated so that by following a prescribed sequence of field pulses each layer can be addressed. To meet these criteria, two Co/Pd multilayer stacks, separated by a 5-nm Pd spacer layer, were sputtered onto the patterned SiO2 substrate at room temperature. The composition of the film sfrom the surface to the substrated is as follows:

Cs40 Åd/Pds10 Åd/fCos2Å d/Pds5.5 Ådg6/Pds50 Åd/ where a 40-Å carbon overcoat is used as a protective layer against corrosion and film abrasion caused by scanning the sample with a recording head. By changing the Co and Pd thickness the strength of the magnetic anisotropy and coercivity of the film can be easily altered and optimized.9,10

I. RESULTS AND DISCUSSION A. Magnetic properties of continuous films

Magneto-optical Kerr effect sMOKEd hysteresis measurements on the continuous full film reveal the distinct switching of each Co/Pd multilayer stack at different applied fields as shown in Fig. 1. In order to confirm that the magnetic interaction through magnetostatic coupling between the two layer stacks is negligible, minor loops were measured. The coupling field is determined by the offset of the minor loop at zero applied field and only a rather small offset of about 10 Oe was found. Therefore, the shape of the hysteresis loop can be simply understood as a superposition of the hysteresis of two almost independent magnetic layers. While the fCos3.3 Åd/Pds10 Ådg5 multilayer stack slower layerd has a coercivity of 350 Oe, a value of 700 Oe was found for the fCos2Å d/Pds5.5 Ådg6 multilayer stack supper layerd.

B. Recording studies on continuous films

On the Co/Pd bilayer film structure magnetic recording experiments were performed using a scanning magnetoresistive microscope sSMRMd.1 Here, the sample was fixed on a x-y stage, controlled by piezoelectric drivers with a resolu- tion of ,2 nm, and scanned at low velocity stypically 5 m/s d while in mechanical contact with the recording head. Conventional longitudinal GMR recording heads were used with a write and read widths of about 150 and 100 nm, respectively. Figure 2 shows SMRM images and extracted GMR readback signals obtained after applying a square wave write pattern using a frequency of 500 flux changes/m sfc/md. Two different write currents of 5 and 10 mA were applied. For a write current of 5 mA both layers are magnetized in the same direction at almost the same vertical position as presented in Fig. 2sad, while applying a write current of 10 mA leads to a 90° phase shift between the two magnetic patterns written in the upper and lower layers generating the four different possible magnetization states fFig. 2sbdg. In order to make the two different situations clearer, schematic drawings of the magnetization orientation in the upper and lower layer are depicted in Fig. 2. The reason for the observed shift in write location in the

FIG. 1. Polar MOKE measurements showing hysteresis loop sopen circlesd and minor loop sblack dotsd of a CoPd/Pd/CoPd continuous film. A maximal reverse field of −600 Oe was applied for the minor loop measurement. A schematic representation of the four possible orientations of the magnetization labeled as s00d, s01d, s11d,a nd s10d is also depicted.

FIG. 2. Magnetic recording on a CoPd/Pd/CoPd continuous film: sad SMRM image stopd and extracted GMR readback signal sbottomd obtained after applying a square wave write pattern using a frequency of 500 fc/m and a write current of 5 mA. A schematic drawing smiddled of the magnetization orientation in the upper and lower layers is depicted as well. sbd Applying a write current of 10 mA leads to a 90° phase shift between the two patterns written in the upper and lower layers generating the four different possible magnetization states.

Downloaded 09 May 2006 to Redistribution subject to AIP license or copyright, see

upper and lower layer is the current dependent spatial magnetic field distribution in combination with the different required switching fields of the two layers.12

C. Magnetic properties of dot arrays

The magnetic properties of the magnetic dot array were measured by remanence hysteresis loops using MOKE. A focused laser beam with a spot diameter of about 20 m was used, which detected signal only from the dot array and the magnetic material in the trench. Interestingly, the Kerr rotation obtained on the dot array changes sign compared to the signal obtained on a continuous film or trench material as shown in Fig. 3. This effect is due to the interplay between the diffraction and magneto-optics occurring in a twodimensional array of magnetic dots.13 The remanent MOKE loop in Fig. 3 shows the switching field distribution sSFDd of the two different layer stacks in the dot array. The lower and upper layers reverse over the same field range of about 3 kOe centered at about 5 and 8 kOe, respectively. Note that the two SFDs are not well separated showing a small range of overlapping switching fields as indicated by the gradual change in the Kerr rotation around 6.5 kOe. In addition, the switching of the continuous magnetic medium deposited in the trench surrounding the dots is observed at low reverse fields and behaves as the full film. This pronounced difference in the SFD and coercivity of the patterned medium in comparison with the full film can be understood in terms of the switching mechanism, which changes from domain wall propagation in the continuous medium to a more Stoner– Wohlfarth-like switching behavior for the single-domain dots.14,15 For the latter, a certain nucleation field needs to be overcome in order to switch an individual layer in the dot. This nucleation field depends strongly on the magnetic and structural properties such as magnetic anisotropy and geo- metrical shape of the dot. For an ideal array with identical magnetic dots the SFD would only be limited by the external demagnetization fields of neighboring dots and the internal demagnetization fields generated by the magnetic bilayer in the dot. Due to the inhomogeneities in the local magnetic properties, the SFD is rather wide and the challenge remains to improve the uniformity.

The identification of the different parts of the hysteresis curve was also confirmed by the magnetic force microscope sMFMd images taken of remanent states after the sample was exposed to different reversal fields as indicated in Fig. 3. After saturation, all dots are in the s11d state, where sXYd refers to the supper, lowerd layers and 1 indicates the positive field-magnetized state, 0 the negative state. When a negative reverse field of about 3.5 kOe is applied, s10d states are initially formed eventually becoming s00d states for higher negative reverse fields. Note that s00d states are already being formed before all of the s11d states are reversed into s10d states. This overlap of the two SFDs was already indicated in the MOKE remanence hysteresis loop. In almost the same manner, s01d states can be formed from s00d states by applying a positive reverse field of about 3.5 kOe. It is worth mentioning that for all magnetic configurations of the duallayer stack, no spontaneous or tip-induced switching event was detected during the course of the experiments sseveral daysd.

D. Recording studies on dot arrays

For the recording studies, we first dc magnetized the sample in an external perpendicular field of −20 kOe. Thus, all dots were in the s00d state and aligned the recording head to scan parallel to the rows of the dots. In contrast to the continuous media se Fig. 2d where the bits can be placed everywhere on the medium, recording on patterned media requires the synchronization of the write pulses with the dot patterning.12,16 However, from the readback signal of the dcmagnetized dots the locations of the dots can be easily retrieved. In this case, the negative peaks in the readback signal correspond directly to s00d states in dot magnetization as shown in Fig. 4sad. Note that the small peaks between the s00d states were produced by the neighboring dots fsee Fig. 4sbdg due to the limited width of the read element.

In order to generate a particular magnetization state of the dot out of the four possible ones, one has to address each storage layer individually. While there are four separate states, it is of course not possible to set the hard layer without aligning the soft layer to be parallel to it with a single pulse. Therefore, for the write process, it is necessary to switch first the hard magnetic layer, thus forming s11d or s00d states, and then in a second step applying a somewhat smaller reverse field in order to switch back the soft layer solely generating s10d or s01d states, respectively. It is clear that for such a write procedure the data rate, however, will decreases accordingly. Figure 4sad shows two examples of a successful write process by positioning the recording head on top of an individual dot fin s00d stateg and applying a write pulse of 30 mA, which is sufficient to switch the magnetization orientation of the softer lower layer to form a s01d

FIG. 3. The polar MOKE loop sremanentd and the corresponding MFM images obtained on a CoPd/Pd/CoPd dot array with a dot size of 150 nm after first dc magnetizing the sample in a positive field of 20 kOe. A switching of the magnetization orientation from state s11d to s10d to s00dis shown at relatively high reverse fields while the film in the trench switches at 350 and 700 Oe.

Downloaded 09 May 2006 to Redistribution subject to AIP license or copyright, see

state s00→01d resulting in a reduced negative signal amplitude. In contrast, a write current of 50 mA is required to reverse also the harder upper layer forming the s11d state s01!11d as indicated in the readback signal, which shows the same signal amplitude compared to a s00d state but with an opposite sign, as expected. Moreover, the ability to write and read back individual data sequences, i.e., s11d-s00d-s01d, along a row of dots was demonstrated as shown in the corresponding SMRM image of Fig. 4sbd. Note that in these experiments, s10d states were not generated due to the technical difficulty in generating negative write pulses of the used experimental setup. However, these states can be clearly distinguished from s01d states having the same amplitude but with opposite sign as observed by MFM.

An aspect of the described perpendicular multiple storage layer media have to be kept in mind which is given by the fact that in a perpendicular system, the different storage layers generate demagnetization fields that favor the ferromagnetically aligned states over the mixed ferrimagnetic states as illustrated schematically in Fig. 5sad. This behavior has been observed experimentally, where the demagnetization of the two layer system results mainly in s00d and s11d states which are more likely than s10d and s01d states f88% s00d and s11d states over 12% s10d and s01d statesg as shown in Fig. 5sbd. This effect will become worse for multiple storage layer media with thinner spacer layers which are required to reduce spacing loss in a recording system. Obvi- ously antiferromagnetic sAFd exchange coupling provides a good solution to overcome this problem. Via the spacer layer thickness si.e., of Ru or Ird it is possible to control the strength of the AF exchange coupling between adjacent ferromagnetic films. Thus, it should be possible to tune AF exchange such that it exactly compensates the demagnetization fields in the way that after demagnetization, all four states have the same probability and stability. For this the interlayer thickness can be chosen as thin as 1 nm or below providing ultrathin multilevel recording media.

In conclusion, a class of magnetic recording media was introduced and its principle feasibility for magnetic data storage was demonstrated. The approach is based on a combination of patterned media and multiple magnetic storage layers. The latter is achieved using two magnetically decoupled layers, where the remanent state of each layer is independent. Therefore, the total magnetization integrated over the two different storage layers gives rise to four different signal levels in magnetization which can be used as a readout signal for magnetic recording applications. By increasing the number of storage layers, the storage density will increase accordingly allowing more feasible feature sizes, which relaxes the lithography requirements for patterned media. However,

FIG. 4. Magnetic recording on a CoPd/Pd/CoPd dot array: sad GMR readback signals after dc magnetizing the sample fs00d statesg and after applying a write pulse of 30 and 50 mA, creating, respectively, states s01d and s11d as illustrated in the drawing. Note that the small dips between s00d states in the readback signal are stemming from the stray field of the neighboring dots due to the limited width of the read element. sbd SMRM image showing a s11d-s00d-s01d data sequence along a row of dots which was previously written on a dc-magnetized dot array without disturbing the magnetization orientation of neighboring dots.

FIG. 5. sad States s00d and s11d are favored by the demagnetization fields since 1 and 2 poles cancel each other at the interface to the spacer layer sleft sided. States s01d and s10d have a high demagnetization energy and thus will decay into the energetically lower s00d and s11d states. sbd MFM image after demagnetization of a two layer system indicating that the favorable s00d and s11d states are much more likely than s10d and s01d states smarked by circlesd. Note that a field of 1 kOe was applied after demagnetizing the sample in order to saturate the trench material. This process does not affect the magnetic states of the dots.

Downloaded 09 May 2006 to Redistribution subject to AIP license or copyright, see the realization of the well-separated SFD’s along with compensated demagnetizing fields of the different storage layers might be required that will become more and more challenging with increasing storage density.

Downloaded 09 May 2006 to Redistribution subject to AIP license or copyright, see