A multi-state magnetic memory dependent on the permeability of Metglas

Current magnetic media use the direction of remanent magnetization to store information. As a result, the data can be corrupted by inadvertent exposure to magnetic fields (e.g., from cell phones, electromagnetic pulses, geomagnetic storms) or if the thermal energy of the media approaches the anisotropy energy.¹ To more robustly archive data in a “non-erasable” memory format, it is desirable to utilize an intrinsic property that depends on the microstructure, such as the magnetic permeability.² Bits of data with different permeability can be differentiated by their effects on an externally applied probe field H_probe. Higher permeability bits will attract magnetic flux and modify H_probe more than lower permeability bits. The changes in the local probe field can be determined with a read head in close proximity to the media. In addition to enhanced data stability, the incorporation of an external probe field eliminates the need for remanence in the media. This allows the use of soft magnetic media, such as Metglas alloys, with multiple permeability states dependent on the structure.

Metglas alloys are a class of magnetically soft ferromagnetic amorphous alloys with low crystalline anisotropy resulting from their lack of long-range crystalline order.^3–6 These alloys typically contain roughly 80 at. % Fe, Ni, and/or Co and 20 at. % metalloids (e.g., B, Si, P) and are formed by quenching a precursor melt into a metastable glass state. Due to the combination of high saturation induction, high permeability, low coercivity, and high resistivity, these materials have found widespread use in magnetic shielding,⁷ sensors,^8–10 and power generation and transmission.^11–13 Since the unique magnetic properties are dependent on the amorphous structure, the effect of crystallizing these materials has been studied. Nucleation and grain growth at elevated temperatures (>673 K) have been shown to deleteriously affect the magnetic softness, decreasing the relative permeability, and increasing the coercivity.^14,15 While the change in permeability reduces the effectiveness of these alloys in the aforementioned applications, it plays the key role in a memory based on such variations.

Here, we report heat-treating an amorphous ferromagnetic alloy, Metglas 2826MB (Fe₄₀Ni₃₈Mo₄B₁₈),¹⁶ to create a three state magnetic memory based on different permeability values. Because boron inhibits nucleation and molybdenum interferes with grain growth in this alloy, significant crystallization does not occur below 673 K.¹⁷ We modified the microstructure and, hence, the permeability, of a set of 250 nm thick Metglas films by an infrared (IR) laser-based heat treatment using a 10 μm spot size and power levels from 70 to 100 mW. The structure and magnetic properties of these films were compared to a set of furnace-annealed films at temperatures from 423 to 823 K for 1 h. The permeability of both sets of samples was enhanced at temperatures from 532 to 723 K (70–80 mW) before diminishing as the samples crystallized at higher temperatures. We found that lines, or bits, with three different permeabilities could be written into a Metglas film by using laser powers of 0 mW (as-deposited), 80 mW, and 100 mW. The permeability of these lines was measured using an external probe field H_probe = 32 Oe and a magnetic sensor consisting of a linear series of magnetic tunnel junctions (MTJs) positioned 10 μm away from the surface of the Metglas. The modification of H_probe by the lines was modeled using finite element analysis with Maxwell 3-D.^18,34

Amorphous films were deposited by sputtering at room temperature from a Metglas 2826MB target at 100 W and 0.7 Pa Ar onto 127 × 127 mm optical grade glass slides. The film thickness, as determined by a contact profilometer, was 250 +/− 5 nm. The slide was diced into 3 × 3 mm pieces that were heat treated using either a furnace or an IR laser. The furnace annealing was set for 1 h at temperatures T ≤ 823 K. The laser heating was performed in an optical setup with a 980-nm CW laser diode and a focal length set to 2.5 cm, resulting in a spot diameter of 10 μm. Ten micron-wide lines were written via laser heat treatment with 50 μm between centers using a motorized XY stage to move the samples past the focal point of the optics. Laser powers from 70 to 100 mW (CW) were used. The raster speed was set so that the average time the Metglas was exposed to the laser was 20 ms. All laser writing ws performed under ambient conditions.

The magnetic properties of the heat treated samples were investigated via an alternating gradient magnetometer (AGM). The results of measuring major (H_max = 1000 Oe) and minor M-H (H_max = 50 Oe) hysteresis loops on both the laser and furnace heated samples are shown in Fig. 1. The maximum permeability, saturation magnetization (M_s), and intrinsic coercivity (H_c) after furnace annealing and laser heating are plotted in Fig. 2. One sees that furnace annealing of the as-deposited sample over the temperature range of 523–723 K significantly increased the permeability and raised M_s by almost 20%. However, increasing the annealing temperature to 823 K lowered M_s and magnetically hardened the material by increasing H_c. Although the heating of the laser-written lines was less uniform, there was a similar trend. Laser writing at 70–80 mW enhanced the permeability and possibly M_s, while writing at 90–100 mW increased H_c and diminished M_s.

The microstructure was investigated using x-ray diffraction (XRD) measurements with a Bruker-AXS D8 Diffractometer equipped with a general area detector diffraction system (GADDS). XRD scans in the region of the highest intensity peaks for selected laser and furnace treatments are plotted in Fig. 3. The broad diffraction peak in the as-deposited samples, notated as either 298 K or 0 mW, indicated that they are amorphous. After annealing at 723 K, diffraction peaks appeared at 2θ = 44° and 51°. Earlier measurements on samples with the same composition observed similar peaks.¹⁹ The peak at 2θ = 44° was identified as the most intense peak of the magnetic γ-NiFeMo phase and the peak at 2θ = 51° was identified with the (511) (NiFeMo)₂₃B₆ paramagnetic cubic boride phase.²⁰ At 823 K, the increased intensity of these peaks indicates the growth of these phases. The nucleation and growth of these crystalline phases support prior suggestions of a primary and secondary crystallization temperature at ∼700 K and ∼810 K associated respectively with the γ-NiFeMo and cubic boride phases.²¹ The laser-written samples show a similar trend where γ-NiFeMo becomes significant at 80 mW and (NiFeMo)₂₃B₆ at 90–100 mW accompanied by coarsening of the γ-NiFeMo nanocrystals. Instrument-corrected values of the FWHM⁻¹ of the (111) γ-NiFeMo peak are plotted in Figs. 2(b) and 2(d). We attribute the increases in FWHM⁻¹ above the crystallization temperature to grain growth of the γ-NiFeMo crystals. Inserting these FWHM⁻¹ values into Scherrer's equation indicates crystal dimensions less than 10 nm for all samples.²² 

To verify the as-deposited composition as well as the absence of significant diffusion on the scale of the laser-written lines, such as those shown in Fig. 3(c), energy dispersive spectroscopy (EDS) analysis was performed using a Hitachi 4500 field emission SEM. The EDS results reveal Ni, Fe, B, and Mo atomic percentages in the as-deposited films within 5% of the Metglas 2826MB target composition. Other studies on Metglas have used evaporation techniques, leading to films with little to no B and Mo glass formers and subsequently higher coercivities at all temperatures.^23,24 Furthermore, after laser annealing, there were no significant changes in atomic compositions within the lines or dramatic increases in the oxygen peak that indicate oxide formation within the bulk of the film.

The changes in the permeability are thus correlated to changes in the microstructure. For both furnace and laser heat treatments, the relative permeability reached a maximum at a point where the FWHM⁻¹ of the γ-NiFeMo peak began to significantly increase due to the onset of nanocrystallization and grain growth. This occurred at ∼723 K in the furnace treatment and 80 mW in the laser treatment. The initial increase in the permeability from the as-deposited Metglas to ∼723 K (80 mW) is due to two effects of thermally induced atomic rearrangements. The first effect at lower temperatures is a decrease in strain anisotropy due to relaxation of the initial sputtered film, as seen in prior studies.^19,25 The second effect is an increase in the magnetic moment due to a local ordering of Ni and Fe atoms from the as-deposited state to a face-centered cubic (fcc) permalloy structure that has a smaller lattice constant than the γ-NiFeMo phase seen at higher temperatures.^26–28 At the glass transition temperature, which has been reported in the range of 648–698 K, such rearrangement is kinetically favorable.²⁹ For annealing above 723 K (80 mW), the coercivity increased exponentially with the grain size according to the Herzer model for nanocrystalline materials.³⁰ Our measurement at increasing temperature above 723 K are consistent with earlier studies that showed M_s was also reduced due to the formation of the paramagnetic cubic boride phase (NiFeMo)₂₃B₆.³¹ Both the increase in H_c and the decrease in M_s diminish the permeability.

Once the laser conditions responsible for the trends in permeability were identified, 1000 × 10 μm wide lines were written in another initially amorphous Metglas film on glass. The lines, acting as bits, had three possible permeabilities dependent on using either 0 mW, 80 mW, or 100 mW of laser power. These samples were placed below the lens of a Keyence VHX-1000 microscope with the Metglas facing the moveable stage of the microscope. Figure 4(a) shows a view through the microscope of a series of 80 and 100 mW lines written on the as-deposited Metglas film. A NdFeB magnet provided H_probe oriented in the plane of the Metglas and perpendicular to the laser-written lines. A linear series of ten 40 × 10 μm MTJs was used to read changes in the probe field. This MTJ reader was attached to the microscope stage and aligned parallel to the written lines.³² The set of lines was read by translating the reader underneath these films at a separation distance, denoted as the fly height h, of 10 μm.

Figure 4(b) illustrates the modeled system for reading the media. The pseudo colors shown in the figure indicate the predicted magnetic flux modification of H_probe = 32 Oe obtained by inserting the magnetization data for the laser-treated Metglas into Maxwell 3-D, a finite element code. The modeling predicts that local increases in permeability shunt the field away from the reader, while decreases in permeability allow flux to emerge from the media, resulting in an increase in the measured field. Memory states are assigned in order of increasing permeability, with a “0” standing for the crystallized 100 mW lines, a “1” for the as-deposited lines, and a “2” for the 80 mW lines.

Experimental changes in the field were determined by the change in percent tunneling magnetoresistance (ΔTMR) of the MTJ reader as it passed underneath the bit lines. The TMR of the MTJ reader as a function of applied field is shown in Fig. 5(a). Since the highest sensitivity of this reader was at 32 Oe, we adjusted H_probe to this value. The ΔTMR values measured as the reader was swept underneath the media shown in Fig. 4(a) are plotted in Fig. 5(b). The as-deposited Metglas on either side of the written lines is set to a ΔTMR of 0%. Having a higher permeability than the as-deposited Metglas, the 80 mW lines shunted an additional 4 Oe away from the reader, lowering the TMR% by ∼0.6% compared to the as-deposited Metglas film. The lower permeability 100 mW lines allowed 12 Oe to emerge from the Metglas, increasing the TMR% by ∼1.6%. The experimental data trend is consistent with the simulated field data shown in Fig. 4(b) for different permeability media. A similar scan of the same Metglas sample was observed after intentionally degaussing these bits from a starting field of 1 T or heating to at least 423 K for 1 h. Thus, these three different permeability states are not affected by such high fields and large temperatures changes.

In summary, we have demonstrated a three state, magnetic permeability memory system using soft magnetic media. By changing the microstructure of Metglas, we can write bits of different permeability with a laser that are resistant to temperatures of at least 423 K for 1 h and to fields of at least 1 T. The bits of different permeability are read by introducing an external probe field to the media and using a magnetic read head to detect local modifications in the field. While this ternary-state memory system may immediately boost storage capabilities by roughly 50% over binary encoding, the ability to tune the permeability at a finer level will increase storage densities at similar bit sizes, analogous to multiple resistance levels in phase change memory.³³ Furthermore, we can decrease the bit size by adjusting the thermal writing, decreasing reader dimensions/fly height, and increasing the reader sensitivity.

We wish to thank Matthew Ervin at the U.S. Army Research Laboratory for support in EDS measurements.