Giant magnetoresistive reproduce head.

0001The present invention relates to a magnetic reproduce head, and in particular to a magnetoresistive reproduce head employing materials exhibiting giant magnetoresistance.

0002As magnetic recording technology continues to push areal recording density limits, magnetoresistive (MR) reproduce heads appear to be the technology of choice. Recent US-A-5,084,794 and US-A-5,193,038 (incorporated herein by reference) disclose dual magnetoresistive (DMR) reproduce heads which offer improved high linear density performance compared to conventional shielded magnetoresistive (SMR) heads, as well more robust operation and simpler fabrication. Until very recently, virtually all past magnetoresistive sensor/heads, including the DMR design, have been based on the physical phenomenon of anisotropic magnetoresistance (AMR) in Permalloy (NiFe) thin films.

0003The last few years have seen increased interest and research in the area known as "giant magnetoresistance" or GMR, which is a "giant" ΔR/R response that is markedly greater in magnitude than that obtained by the AMR effect and is described below. Films observed to exhibit giant magnetoresistance can consist of multilayers of very thin (roughly 20 Angstrom) ferromagnetic films alternating with similarly thin layers of nonmagnetic conducting films typically of copper, silver, or gold.

0004The great potential of GMR films as magnetoresistive heads stems from the very large maximum change in resistance (ΔR/R as much as 10-20% in some systems) that they can exhibit in response to magnetic fields, as compared to the ΔR/R ≃ 2% typical of traditional MR films employing the usual anisotropic magnetoresistance (AMR) effect. In general, GMR films are in a high resistance state when the magnetization in the GMR multilayers are predominantly antiparallel in consecutive magnetic layers, and can be. then brought to a low resistance state by the action of an applied field which rotates each layer's magnetization into a predominantly parallel orientation roughly along the applied field direction.

0005Excluding the special case of the trilayer "spin-valve" design with an additional exchange pinning layer as taught in US-A-5,159,513 and US-A-5,206,590, multilayers which exhibit GMR in response to external applied fields generally require some intrinsic mechanism (exchange or magnetostatic in origin) which produces an effective antiferromagnetic coupling between adjacent ferromagnetic layers in order to maintain the aforementioned antiparallel alignment in zero external field. The strength of this antiferromagnetic coupling is denoted in terms of the effective coupling field H_afc, which in most GMR multilayers is usually larger in magnitude than any existing uniaxial anisotropy of strength H_k.

0006The intrinsic resistance R vs. (hard axis) applied field H response curve of a generic GMR multilayer (excluding hysteresis, if any) is shown in Fig. 1. To obtain the optimal sensitivity dR/dH with maximum dynamic range, a GMR multilayer would be operated at the indicated bias point, requiring a bias field H_b ≈ (H_afc + H_k)/√2 ≈ H_afc/√2 in order to rotate the magnetization angle by ≈45° such that sinϑ_b ≈ 1/√2. (For small, patterned GMR reproduce head elements, one should replace H_k with H_k + H_d, where H_d is the geometry dependent demagnetization field of the element, which can be quite large, for example, 100-200 Oe.) Much recent work on GMR has been directed to the development of materials with low effective H_afc and large ΔR/R in order to maximize the intrinsic sensitivity ΔR/R/H_afc. To date, the largest value of the latter, with ΔR/R of 4-5% and H_afc of 5-10 Oe, was reported by Hylton et al., reported in <u>SCIENCE</u>, vol. 261, pp. 1021-1024, 8/20/93. However, this and other reported R vs. H wafer level measurements of intrinsic H_afc and ΔR/R are generally made at current densities which are minuscule compared to those that would exist in the actual micron sized GMR reproduce heads contemplated for the application of these GMR materials. Fig. 2 indicates the spatial distribution of the internal current field H_j(x,y) generated in a thin film of total thickness, T, carrying a uniform current density J. To a good approximation, H_j(x,y) ≃ (2π/5)J x O_e, with J in amp/cm².

0007The current field H_j(x) is <i>antisymmetric</i> about the film center (x=0), that is, H_j(-x) = - H_j(x). This is of opposite symmetry to that of the uniform bias field H_b of Fig. 1, or more generally, the <i>symmetric</i> component H_sig-s (x,y) = [H_sig(x,y) + H_sig(-x,y)]/2 of any signal field H_sig(x,y) which such a uniformly biased GMR multilayer would be designed to measure using the teachings and practices of the current art. By itself, the current field H_j(x) will induce an antisymmetric bias angle distribution sinϑ j(x) ≈ H_j(x)/H_afc, which, it is important to note, is unimpeded by the size of H_d because the antisymmetric sinϑ_j(x) averages to zero. When subsequently superimposed (possibly non-linearly) on the desired uniform bias distribution sinϑ_b ≈ 1/√2, there will result an asymmetric ϑ-distribution which can be skewed through the thickness of the GMR multilayer. In cases where H_j-max = H_j(x = T/2) ≧ H_afc, there will likely occur substantial distortion of the internal GMR bias distribution in which opposite sides through the multilayer thickness will be substantially under biased (|sinϑ| « 1/√2) or saturated (|ϑ| ≈ π/2). Such bias distortion cannot be compensated for by any additional source of uniform (symmetric) bias field.

0008For future ultra high storage density applications, with sub-micron track widths, conventional AMR heads and/or spin-valve GMR designs will likely need to operate at current densities J ≧ 10⁷ amp/cm² to achieve adequate signal levels. Since reproduce signal scales directly with current density, GMR multilayer heads will need to operate at similar J if they are to take full advantage of their large intrinsic ΔR/R. For a GMR multilayer of total thickness T = 400A, this corresponds to H_j-max ≧ 25 Oe. It follows that such a GMR multilayer with H_afc ≦ 25 Oe, will, when operated at such competitive current densities, be expected to suffer the aforementioned distortion of their bias magnetization distribution that will inevitably lead to a significant reduction in the usable ΔR/R, as well as loss of linear dynamic range and accompanying increase in nonlinear distortion of the reproduce signal.

0009Excluding the aforementioned DMR design, conventional MR head designs using either single layer AMR films, GMR spin-valves, or (uniformly biased) GMR multilayers will all require magnetic shield structures with MR-shield gaps of order 1/10 micron in order to obtain adequate linear reproduce resolution for future ultra high density recording systems. In addition to imposing significant technological difficulties such as insuring adequate electrical insulation of the MR head from very closely spaced conductive magnetic shields, such shielding structures also limit design options for biasing the MR element. Regarding the uniformly biased GMR multilayer, the only likely feasible bias technique known in the present art involves sense current self-bias by the inclusion of additional biasing layers, for example, current shunt and/or magnetically-soft-adjacent-layers. Even for convention AMR films with small anisotropy fields of H_k ≃ 4-5 Oe, it is well known that these adjacent-layer bias techniques will require ever increasing bias current densities of >10⁷ amp/cm² to achieve adequate bias levels as the MR element height is reduced to 1 micron or less to accompany similarly small future reproduce track widths. This can further exacerbate the detrimental effects of the additional antisymmetric self-bias component of GMR multilayers described above. Although the latter problem can be eased, as implied above, by development and use of GMR materials with H_afc of order 30-100 Oe, this in turn will add to the level of required current density to achieve adequate (uniform) bias of the shielded GMR multilayer, which then exacerbates internal bias problem, and so forth. The presence of additional biasing layers will in general also lower signal levels due to their electrical shunting of the MR layer(s).

0010The desirability of achieving a uniform bias state of sinϑ ≃ 1/√2 in a GMR multilayer is that it is the optimum bias state for maximizing reproduce sensitivity/dynamic range for detecting the <i>symmetric</i> component of any signal field. A brief, limited summary of the difficulties in achieving such a bias state have been outlined above. However, an approximately antisymmetric bias state, such as the internal sense current self-biased state of a GMR multilayer, is similarly optimal for detecting the <i>antisymmetric</i> signal field component H_sig-a(x,y) = [H_sig(x,y) - H_sig(-x,y)]/2. At high linear magnetic recording densities, both H_sig-s and H_sig-a are similar in amplitude, and either one may be detected for the reproduction of information recorded on a magnetic medium. The DMR design, as taught in the above-incorporated '794 and '038 patents, is an example of an antisymmetrically self-biased MR head which exploits an intrinsic sensitivity to H_sig-a in order to be able to achieve, <i>without magnetic shields</i>, comparable high density signal levels and superior linear resolution to the most narrow gap shielded MR heads that may be practically fabricated.

0011The present invention is aimed at overcoming the recognizable problems of employing GMR multilayer films to replace conventional AMR films in an otherwise conventional shielded-MR reproduce head design for future high density magnetic recording applications.

0012More particularly, the present invention is directed to a GMR multilayer reproduce head which exploits the internal sense current self-biasing of the GMR multilayer to establish a "GMR-DMR" head, that is, an unshielded GMR multilayer which is completely self-biased at reasonable current densities, and which achieves the functionally equivalent high density reproduce performance advantages intrinsic to the conventional DMR reproduce head design as taught in the above-incorporated '794 and '038 patents, but with the much greater intrinsic sensitivity offered by the "giant" magnetoresistivity of GMR materials.

0013Further, the magnetostatic stability of the fully antisymmetrical GMR-DMR bias magnetization state, which is additionally strongly energetically favored by the intrinsic antiferromagnetic interlayer-coupling of GMR multilayers, should obviate the need for complex single-domain-stabilization/Barkausen-noise-suppression means known in the art that are difficult to manufacture, and should make the GMR-DMR structure more naturally suited to the sub-micron sizes (both track width and element height) expected of future ultra high density reproduce heads than any practical alternative MR head design yet known.

0014Accordingly, a GMR-DMR sensor or reproduce head in accordance with the invention comprises a substrate, a thin film, multilayered structure formed on said substrate; said structure further comprising alternating electrically conducting layers of magnetic and non-magnetic film layers formed such that the magnetic film layers are substantially antiferromagnetic coupling across the sandwiched nonmagnetic film layers, said antiferromagnetic coupling aligning the magnetization vectors of adjacent planar regions of consecutive magnetic film layers to be substantially antiparallel to one another when not subjected to either an external magnetic field or the field induced by current flow therein, said multilayer structure exhibiting a change in resistivity when subjected to an external magnetic field causing the magnetization vectors of said adjacent planar regions of said magnetic film layers to rotate into a substantially more parallel orientation thereby exhibiting a giant magnetoresistive effect, means for producing a current flow through said thin film, multilayer structure, wherein the magnetic fields generated by said current flow maintain a bias magnetization state of said structure in which the net component of magnetization along the axis transverse to that of said current flow of a first set of sandwiched magnetic and nonmagnetic film layers that are formed on one side of the center-most film layer is substantially equal in magnitude and opposite in direction by comparison to that of the net component of magnetization along the axis transverse to that of said current flow of the second set of sandwiched magnetic and nonmagnetic film layers that are formed on the other side of the center-most film layer, so that the layered structure operates in the fashion of a dual magnetoresistive sensor or reproduce head, and means for sensing the variations of resistivity of said magnetoresistive head in response to an applied magnetic field.

0015Preferably in the GMR-DMR sensor or reproduce head, both of the outer-most magnetic film layers are formed substantially thinner (approximately 50% thinner) than the relatively uniform thickness, inner magnetic film layers of the multilayer structure in order to reduce self-bias level variance between the outer-most and inner magnetic film layers.

0016The magnetic and nonmagnetic film layers may be equal or unequal in number but preferably would be formed as two mirror image sets of N magnetic and N-1 nonmagnetic film layers on either side of a center-most nonmagnetic film layer or magnetic film layer. The center-most film layer, particularly the center-most nonmagnetic film layer, may be formed of a greater thickness sufficient to ensure that the net components of magnetization along the axes transverse to that of the current flows of the first and second sets of N sandwiched magnetic and nonmagnetic film layers are equal and opposite.

0017The GMR-DMR multilayer is unshielded and completely self-biased at reasonable current densities and achieves the functionally equivalent high density reproduce performance advantages intrinsic to the conventional DMR reproduce head design, but with the much greater intrinsic sensitivity offered by the "giant" magnetoresistivity of GMR materials. The required perpendicular biasing is attained solely by the interaction of the magnetic film layers across the nonmagnetic film layers under the influence of the magnetic fields arising from the flow of the sense current in the GMR-DMR structure. The inherent high resolution of the prior art DMR head is retained in the operation of the GMR-DMR head, and is effectively combined with the high output signal capability provided by GMR behavior of the thin film layers. The GMR-DMR sensor or reproduce head can be advantageously constructed and operated with sub-micron element heights suitable for reproducing ultra high linear density recorded tracks widths at or below the one micron range. <ul id="ul0001" list-style="none"><li>Fig. 1 is a sketch of the intrinsic R vs. H response curve (excluding any possible hysteresis) of a generic GMR multilayer, along with a sketch of the orientation angles of the magnetization of any two adjacent magnetic layers;</li><li>Fig. 2 is a sketch illustrating the spatial distribution of the internal magnetic field generated inside a thin film conductor carrying an approximately uniform current density;</li><li>Fig. 3 is a cross-section view of a GMR-DMR reproduce head structure, in accordance with a first embodiment;</li><li>Fig. 4 is an exploded, perspective view in part illustrating a self-biased 2N magnetic layer GMR-DMR multilayer reproduce head, in accordance with the first embodiment;</li><li>Fig. 5 is a graph depicting experimentally measured R vs. H curves for a 2N=10 layer GMR multilayer subjected to varying sense current densities, along with theoretical simulations;</li><li>Fig. 6 is a theoretical graph of the relative efficiency of several designs of GMR-DMR heads, vs. GMR-DMR element height L;</li><li>Fig. 7 is a theoretical graph of the sense current density required to achieve approximately optimal bias of several designs of GMR-DMR heads, vs. GMR-DMR element height;</li><li>Fig. 8 is a graph of theoretical computations of the intrinsic spectral response function for several designs of GMR-DMR reproduce heads, along with a comparison to that for a conventional DMR with equivalent geometric/magnetic properties;</li><li>Fig. 9 is a cross-sectional view of a further alternative embodiment; and</li><li>Fig. 10 is a cross-sectional view of a still further alternative embodiment.</li></ul>

0018The figures are not necessarily to scale.

0019Fig. 3 illustrates a cross-section view of a GMR-DMR reproduce head or sensor structure 10, in accordance with a first embodiment. On top of substrate 12 (and over a suitable, optional, buffer layer 14) are alternatingly deposited 2N electrically conductive magnetic layers, including outer magnetic layers 16, 18, and interior magnetic layers 20, with 2N-1 nonmagnetic, electrically conducting layers 22. The nonmagnetic conducting layers 22 are nominally of the same composition and thickness, as are all the interior magnetic layers 20. The outer magnetic layers 16 and 18 are nominally of similar composition to interior layers 20, but have approximately one-half the magnetic thickness of layers 20. A conductive passivation layer 30 of preferentially high electrical resistivity is then deposited on top of outer layer 18. Patterned current lead structures 32 and 34 are subsequently deposited atop passivation layer 30 using photolithographic techniques well known in the art.

0020The composition and thicknesses of the layers are in general chosen to optimize the properties of the GMR-DMR sensor, for example, maximize the coefficient ΔR/R consistent with an appropriate magnitude of coupling field strength H_afc. The magnetic layers 16 and 18 are also labeled "-N" and "N", respectively, and interior magnetic layers 20 are labeled in order extending outward from the central, nonmagnetic conductive layer 24 (shown in Fig. 4). The fabrication of this structure is relatively straightforward, and it requires no additional depositions of bias and/or stabilization layers and essentially only one final masking step in order to form the patterned current lead structures 32, 34.

0021Fig. 4 illustrates the basic structure of the GMR-DMR element layers in a partial perspective view, and also defines various geometric and magnetic parameters characteristic of the embodiment, where the number N may be on the order of ten. The substrate 12, buffer layer 14, passivation layer 30 and patterned current lead structures 32, 34 are not shown for simplicity. The center-most non magnetic layer 24 is the starting point for the numbers 1-N and -(1-N) appended to the magnetic layers in the multilayer structure. The interior magnetic layers 20 and nonmagnetic layers 22, 24 have thicknesses t and g, respectively, that are typically on the order of about 2 nm. The thicknesses t_N = t_-N of the outermost magnetic layers are approximately one half that of the interior magnetic layer 20 thickness t. For high density reproduce heads, the stripe height L of the GMR-DMR sensor structure 10 is about 1 micron.

0022Magnetic layers 16, 18, 20, are generally deposited in the presence of a magnetic field in order to induce an easy axis of uniaxial magnetic anisotropy longitudinal to the stripe (along the z-axis), with anisotropy strength characterized by an anisotropy field H_k typically about ten Oe or less. The magnetic layers 16, 18, 20, are typically alloys of Fe, Co, and/or Ni, and may possibly consist of bilayers of more than one alloy composition. The nonmagnetic conducting layers 22, 24 are generally deposited of Cu, Au, or Ag or alloys thereof.

0023In order to estimate the performance and device physics of the GMR-DMR reproduce head described in the embodiment of Figs. 3 and 4, it is important to have some understanding of the nature of the interlayer coupling between magnetic layers of the GMR multilayer. It is assumed here that the coupling mechanism is predominantly exchange coupling between adjacent magnetic layers. The interaction energy (per unit surface area) is taken to have the form: K_afc cos(ϑ_n - ϑ_n+1) + K_bqc cos²(ϑ_n - ϑ_n+1), from which one defines the "antiferromagnetic coupling field" H_afc ≡ 4K_afc/(M_st) and "biquadratic coupling field" H_bqc ≡ 8K_bqc/(M_st), where M_s is the saturation magnetization of the magnetic layers. There is currently a growing body of evidence as to the importance of including a biquadratic component to the interlayer exchange coupling equations, and that H_bqc can be comparable (or even larger) in magnitude than the antiferromagnetic coupling field H_afc which was earlier thought to be alone sufficient to characterize the interlayer exchange interaction.

0024For example, Fig. 5 shows R vs. H measurements (expressed as the %(δR/R) resistance change for various mean stripe current densities J. The externally applied field H was transverse to the stripe (along y-axis) except in the case of the bottom curve, where H was longitudinal. The experimental results where taken on a 2N = 10 layer NiFeCo-Cu GMR multilayer with approximate parameter values 4πM_s = 8000 G, H_k = 4 Oe, t = g = 2 nm, and L = 12 micron provided by the MINT Center of the University of Alabama. The data have been corrected for the reduction in |δR|_max/R = ΔR/R due to purely thermal effects (from Joule heating) that were independently measured in a temperature controlled oven. The remaining one third reduction of the observable %(ΔR/R), from 6% to 4%, at a current density J = 1.25 x 10⁷ amp/cm², is a direct manifestation of the aforementioned effects of internal sense current self-biasing. This is confirmed by comparison to micro-magnetic simulations also shown in Fig. 4. Using H_afc = 40 Oe (roughly fit to the full-width-at-half-maximum of the low-J data) and assuming H_bqc = 0 (dotted curves) or H_bqc = H_afc (dashed curves), theoretical results for the case H_bqc = H_afc accurately predict the observed reduction in ΔR/R due to sense current self-biasing for all J values shown.

0025In particular, the H_bqc = H_afc simulation also closely matches the "triangular" shape of the low-J δR/R(H) curves which is both common to many experimentally studied GMR multilayer systems, and which cannot be explained by the parabolic curve that is theoretically predicted when only antiferromagnetic interlayer coupling is present (that is, H_bqc = 0). For the high-J data, the experimental curves show a developing asymmetry about H = 0 which is indicative of a asymmetric non-uniformity of film properties through the thickness of the particular GMR sample used in the experiment, and which is not predictable by the theoretical simulation. However, for a longitudinal applied field, the measurement is made insensitive to such a non-uniformity (since H is here perpendicular to the current field H_J) and reasonably good agreement between theory and experiment is again obtained when H_bqc = H_afc, with notably poorer agreement when H_bqc = 0.

0026Using an analytical approximation to the micro-magnetic model used for the simulations of Fig. 5, one can show, for the structure shown in Fig. 4, that the mean bias level 〈sinϑ_n〉 (where 〈 〉 denote an average bias level over element height L) in the nth magnetic layer (-N ≦ n ≠ 0 ≦ N), due to sense current self-bias can be accurately estimated from solutions of the cubic equation:<maths id="math0001" num="(1a)"><math display="block"><mrow><msub><mrow><mtext>〈sinϑ</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>〉[1 + (H</mtext></mrow><mrow><mtext>bqc</mtext></mrow></msub><msub><mrow><mtext>/H</mtext></mrow><mrow><mtext>sat</mtext></mrow></msub><msub><mrow><mtext>) (2〈sinϑ</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><msub><mrow><mtext>〉² -1)] = H</mtext></mrow><mrow><mtext>Jn</mtext></mrow></msub><msub><mrow><mtext>/H</mtext></mrow><mrow><mtext>sat</mtext></mrow></msub><msub><mrow><mtext> (〈sinϑ</mtext></mrow><mrow><mtext>n</mtext></mrow></msub><mtext>〉) < 1)</mtext></mrow></math><img file="EP0684484A2_D0001.tif" /></maths> where<maths id="math0002" num="(1b)"><math display="block"><mrow><msub><mrow><mtext>H</mtext></mrow><mrow><mtext>Jn</mtext></mrow></msub><mtext> = sgn(n) (2|n|-1) (g + t) (πJ/5),</mtext><mspace linebreak="newline" /><msub><mrow><mtext> (H</mtext></mrow><mrow><mtext>Jn</mtext></mrow></msub><mtext> in Oe with J in amp/cm²)</mtext></mrow></math><img file="EP0684484A2_D0002.tif" /></maths><maths id="math0003" num="(1c)"><math display="block"><mrow><msub><mrow><mtext>H</mtext></mrow><mrow><mtext>sat</mtext></mrow></msub><msub><mrow><mtext> = (H</mtext></mrow><mrow><mtext>afc</mtext></mrow></msub><msub><mrow><mtext> + H</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>) / (1-ε), ε = (2Λ/L)tanh(L/2Λ)</mtext></mrow></math><img file="EP0684484A2_D0003.tif" /></maths><maths id="math0004" num="(1d)"><math display="block"><mrow><msub><mrow><mtext>Λ = [(1/10 (2N-1)²4πM</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><msub><mrow><mtext>tg + 2A/M</mtext></mrow><mrow><mtext>s</mtext></mrow></msub><msub><mrow><mtext>) / (H</mtext></mrow><mrow><mtext>afc</mtext></mrow></msub><msub><mrow><mtext> + H</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msup><mrow><mtext>)]</mtext></mrow><mrow><mtext>½</mtext></mrow></msup></mrow></math><img file="EP0684484A2_D0004.tif" /></maths><maths id="math0005" num="(1e)"><math display="block"><mrow><msub><mrow><mtext>t</mtext></mrow><mrow><mtext>N</mtext></mrow></msub><msub><mrow><mtext>/t = r</mtext></mrow><mrow><mtext>N</mtext></mrow></msub><msub><mrow><mtext> ≃ (N-1)/(2N-1) [1 + N/ (2N-1)² (1+H</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext>/H</mtext></mrow><mrow><mtext>afc</mtext></mrow></msub><mtext>)/(1+g/t)]</mtext></mrow></math><img file="EP0684484A2_D0005.tif" /></maths> and A is the intralayer exchange stiffness of the magnetic layers. Although derived for the case H_bqc≦ H_afc + H_k, Eq. 1 can also give reasonably accurate predictions when H_bqc > H_afc + H_k under practical conditions of low L (for example, ≦ 1 µm) and moderately high J (for example, ≧ 10⁷ amp/cm²). The restriction of Eq. 1e to multilayers where t_N/t = r_N ≈ (N-1)/(2N-1) ≈ 1/2 is made to compensate for the fact that the outermost ±Nth magnetic layers are exchange coupled to only one adjacent magnetic layer, and thus experience only about half of the interlayer exchange torques as do interior magnetic layers. The condition t_N = r_Nt quenches out what otherwise (for example, when t_N = t and H_afc » H_sat - H_afc) could result in oscillations in 〈sinϑ_n〉 vs. n that are not accounted for in Eq. 1a, and which could be detrimental to the operation of a GMR-DMR head. This effect is roughly analogous to properly terminating a coaxial transmission line to quench out reflections.

0027Like the conventional DMR, a GMR-DMR sensor responds primarily to the antisymmetric component of a transverse signal field H_sig-a(x,y) = [H_sig(x,y) - H_sig(-x,y)]/2. The parameter A physically represents the GMR-DMR flux-propagation length for signal flux generated by the H_sig-a(x,y) from a recorded magnetic medium, and the quantity ε in Eq. 1c approximately represents the linear flux-efficiency of a given GMR-DMR design relative to a maximum possible efficiency ε = 1. A plot of flux efficiency ε vs. element height L, for N = 5 and 10, H_afc = 20, 40, and 80 Oe, and otherwise assuming representative parameter values of 4πM_s = 10⁴ G, H_k = 5 Oe, t = g = 2 nm, and A = 10⁻⁶ erg/cm, is shown in Fig. 5. Because the GMR-DMR bias level |〈sinϑ_n〉| is nonuniform with |n| (Eqs. 1.a,b), the choice of J to achieve an optimal bias point will in general be an application-specific compromise to minimize nonlinear distortion from under-bias or over-bias of inner or outer magnetic layers, respectively, as well as achieve adequate signal levels which otherwise scale linearly with J. The "optimal" bias current J_opt is likely to be in the range of that which achieves the condition |〈sinϑ_n〉| ≃ 1/√2, where 2N/3 ≦ n' ≦ N. Using Eq. 1a, this criterion inverts to the condition H_Jn' = H_sat/√2, which is both simple and robust from a design standpoint since it is expected to be <i>independent</i> of the value of H_bqc. Using, for present demonstration purposes, the more conservative choice n' = N, Fig. 7 shows plots of J_opt vs. L assuming the same parameter values as were used for Fig. 6.

0028Future ultra high density magnetic recording systems (particularly disk systems) will demand submicron reproduce track widths, which in turn will require comparably small (and preferably smaller) element heights for the longitudinal head geometries which are presently the most probable designs for future MR reproduce heads. The results of Figs. 6 and 7 are indicative to the fact that the GMR-DMR is naturally suited to this sub-micron element height regime, in terms of both achieving decent intrinsic flux efficiencies ε of around 50% or more, and with high, but non-excessively large bias/sense current densities J_opt of around 2 x 10⁷ amp/cm². Further, the magnitude of J_opt is less sensitive to element height (for L &lt; 1 µm) in comparison to other well known self-biased heads (for example, soft-adjacent-layer biased MR or conventional DMR), where, in practical designs, J_opt ∝ 1/L² at sub-micron L, and where impracticable high values of J approaching 10⁸ amp/cm² may be required to achieve optimal bias, which is then further subject to larger variability due to fabrication tolerances on establishing such small element heights. These GMR-DMR characteristics stem from the physical smallness of the length parameter Λ &lt;&lt; 1 µm (Eq. 1d) in typical cases. While this condition tends to reduce efficiency ε, it also makes J_opt and H_sat less subject to the L-dependent exchange and magnetostatic (shape anisotropy) limitations on bias level amplitude achievable at a given J (Eqs. 1.c,d). Further, the somewhat lower intrinsic flux efficiency of the GMR-DMR in comparison to more conventional designs (with ε ≃ 1) will usually be in good part compensated for by the lack of electrical shunt loss from the additional non-MR conductive biasing layers generally required of these same alternative MR head designs.

0029The reproduce signal level and linear resolution of any MR head design may be quantitatively described in terms of the intrinsic spectral response function F(λ) of the head, which can be defined from the relation δR (λ)/ΔR = (M_rδ/M_s T_M) [(1-exp(-2πδ/λ))/(2πδ/λ)]exp(-2πd/λ) F(λ), which describes the amplitude of the linearized magnetoresistive response from playback of a sinusoidally recorded magnetic medium. M_rδ is the amplitude of the recorded magnetic medium's remanence-thickness product, λ is the recorded wavelength, d is the head-medium spacing, and T_M = (N-1/2)t is the total magnetic half-thickness of the GMR-DMR. Spectral response function F(λ) is particularly useful for comparative purposes, since it is reflective of a given head's intrinsic reproduce capabilities, independent of details of the medium, head-medium interface, and or the record process.

0030Fig. 7 shows numerical computations of F(λ) for several L = 0.7 µm element height GMR-DMR head designs with H_afc = 40 Oe, N = 5 or 10, H_bqc= (0, 1, or 2) x H_afc, J_opt = (2.1 or 1.4) x 10⁷ amp/cm² for N = 5 or 10, respectively (Fig. 6), and with other parameter values as cited above. The peak values of F(λ) for curves with N = 5 are about 56% as large as for N = 10, which is in good agreement with the value of 61% predicted from the ratio of relative efficiency shown in Fig. 7. The null observed in F(λ) at λ = G' = 0.060 µm for N = 10 and H_bqc = 0 shows the effective DMR reproduce gap G' to be reasonable close to the G' ≃ 4N(g + t)/3 = 0.053 µm that one would in general estimate given the linear bias profile |〈sinϑ_n〉| ∝ 2|n| - 1 predicted in Eq. 1a. (The nonlinear H_bqc term in Eq. 1a yields a more uniform bias profile, which reduces the effective G' somewhat.)

0031As was inferred earlier from Eqs. 1, Fig. 8 shows that the reproduce properties of the GMR-DMR (most notably the efficiency, or peak amplitude of F(λ) to be reasonably insensitive to H_bqc, particularly when H_bqc ≦ H_afc + H_k. The latter condition should also help eliminate the possibility of hard axis hysteresis from the nonlinear nature of the biquadratic coupling term, and thus is a more generally desirable criterion for the future development of practical GMR multilayer materials. As was shown in Fig. 5, the presence of a moderate degree biquadratic coupling can also beneficially serve to "linearize" the intrinsic R(H) response curve (that is, make the slope dR/dH less bias-point dependent), and thereby reduce the well known even order distortion from the parabolic R(H) common to AMR or GMR with H_bqc = 0. This can be particularly beneficial to the GMR-DMR design, where the aforementioned non-uniformity in bias level precludes simultaneously having all magnetic layers individually biased to an optimum bias point.

0032Fig. 8 also shows a comparison with F(λ) for a conventional AMR-DMR with the same values of 4πM_s, H_k, T_M, and G' as for the N = 10, H_bqc = 0, GMR-DMR structure, and which is biased to the equivalent mean bias level of |(sinϑ)| = 1/2 (|〈sinϑ_n〉| = 1/√2) = 0.354. The AMR-DMR result was computed using the analytical equations described by Smith et al., <u>IEEE Trans. Magn.</u>, vol. 28, pp. 2295-2297, 1992. Although 25% lower in relative efficiency, the spectral shape of F(λ) for the GMR-DMR structure is virtually identical to the equivalent AMR-DMR head, indicating their functional equivalence regarding the underlying device magnetics that yields the high sensitivity and unsurpassed intrinsic linear resolution of the DMR design, as taught in the above-incorporated '038 patent. When further accounting for the fact that GMR multilayers of similar magnetic characteristics to that assumed above, with ΔR/R ≈ 10%, have already been made in the laboratory (e.g, the case shown in Fig. 5), and with expected improvements in the future, the GMR-DMR reproduce head potentially offers an order of magnitude improvement in reproduce sensitivity over the conventional AMR-DMR reproduce head for ultra high density magnetic recording applications.

0033The GMR-DMR reproduce head also offers additional benefits, not included in the above descriptions, which become very important in small track width MR reproduce heads where the three-dimensional effects of anisotropic signal flux propagation (via magnetization rotation) through the active area of an MR head become important. Firstly, because adjacent magnetic layers in either half of the self-biased GMR-DMR are magnetized at approximately supplementary angles, that is, ϑ_n + ϑ_n+1 ≃ π for N >> 1 (see Fig. 4), the axes of rotational flux propagation (oriented at angle ϑ_n ± π/2) of adjacent pairs of magnetic layers will propagate the signal flux (up from the medium, through the active area of the MR element) equally well from either side of the recorded track, thereby eliminating the well known and undesirable asymmetry in the cross track profile of a single, narrow-track MR element.

0034Secondly, because the dual pair of ± Nth magnetic layers in opposite halves of the self-biased GMR-DMR are magnetized antiparallel, that is, ϑ_-n = ϑ_n ± π, their axes of rotational flux propagation are necessarily parallel. In this circumstance, each dual pair individually (or collectively for all odd, or even n) constitute a fully antisymmetric, or "A-DMR", bias configuration, in which dual pairs form a minimum reluctance, closed loop flux return path for the signal flux generated by the antisymmetric component H_sig-a of the recorded medium's signal field.

0035As was taught by Smith et. al., <u>IEEE Trans.</u><u>Magn.</u>, vol. 29, March, 1994, the A-DMR configuration is not only key to maximizing narrow-track reproduce sensitivity, but crucially important to provide the common mode rejection of longitudinal signal fields needed minimize adjacent track cross-talk interference and on-track signal distortion due to mistracking. The vanishing of the slope d(δR)/dH about H = 0 in the bottom curve in Fig. 5 is a demonstration of such common mode rejection of longitudinal fields. Unlike the conventional DMR, in which difficult to fabricate means would likely be required to initialize and stabilize the A-DMR bias state, the antiferromagnetic nature of the interlayer exchange coupling intrinsic to GMR multilayers should automatically and optimally serve precisely this function at zero additional cost or complexity.

0036The effective reproduce gap G' = 0.06 µm from the GMR-DMR example of Fig. 8, is probably already beyond practical linear resolution limits of conventional shielded MR designs (using AMR or GMR), due to the need to electrically insulate the MR element from conductive magnetic shield structures located a distance ≦ G' from the MR and/or its current lead structures. The GMR-DMR reproduce gap, G' ≃ 4N(g + t)/3, can readily be reduced as small as necessary to meet any foreseeable linear resolution requirements at ultra high linear recording densities of the future. Because the thicknesses t, and especially g, will likely be chosen to optimize ΔR/R, H_afc, and/or other intrinsic properties of a GMR multilayer material, layer number N is likely to be the design parameter available for adjustment to optimize G' in the embodiment of Figs. 3 and 4.

0037The alternative GMR-DMR embodiment shown in Fig. 9 would allow some additional flexibility to design the reproduce gap G' more independently from the effective magnetic thickness T_M ≃ Nt, which is more optimally chosen in accordance with the recording medium so to maintain a value of (M_rδ/M_sT_M) consistent with adequate signal levels at sufficiently low nonlinear distortion. The alternative embodiment of Fig. 9 is essentially the same as shown in Figs. 3 and 4, with the exception that the central, nonmagnetic, conductive layer 24 has a thickness g₀ > g. The thicker central conductive layer 24 may be made of the same material as the other nonmagnetic conducting layers, or if g₀ >> g, from an alternative higher electrical resistivity material to minimize the necessary signal loss from electrical shunting by a "thick" central conductor. In this latter case, the exchange coupling between the N = ±1 magnetic layers will be eliminated, and magnetostatic coupling between opposite halves of the GMR-DMR structure will be weakened. Consequently, a minimum shunt current down the central conductive layer 24 may be also necessary to maintain unipolarity of 〈sinϑ_n〉 in each half of the multilayer on opposite sides of layer 24. It would also be necessary to establish that the desirable properties of the second half GMR multilayer deposited on top of the material of layer 24 are maintained, and that they be approximately equivalent to that of the first half deposited earlier. The latter condition may be aided by using a buffer layer 14 of similar composition to central, nonmagnetic conductive layer 24. The center-most nonmagnetic film layer 24 is preferably of a greater electrical resistivity than the remaining nonmagnetic film layers 22.

0038In yet another alternative embodiment, shown in Fig. 10, there are 2N nonmagnetic conducting layers 22, and 2N+1 magnetic layers 16, 18, 20, including a central magnetic layer 28. In this embodiment, the central magnetic layer 28 is chosen to be an n = 0 magnetic layer with thickness t₀ > t, which maintains (possibly modified in strength or form) exchange coupling between adjacent n = ±1 magnetic layers. Remaining unbiased by H_J (that is, 〈sinϑ_n〉 ≃ 0), a central magnetic layer 28 can also be useful in stabilizing an antisymmetric bias state 〈sinϑ_-n〉 = -〈sin ϑ_n〉 (and particularly 〈sinϑ₋₁〉 = -〈sinϑ₁〉) in cases where H_bqc > H_afc + H_k. However, in this, or possibly in the preceding alternative embodiment, the bias angles will now satisfy the condition ϑ_n + ϑ_-n = 0, and not the true A-DMR bias condition described above. Nonetheless, for large N >> 1, ϑ_n + ϑ_n+1≃ π, which means that 0_-n ≃ ϑ_n±1 ± π, and the odd n layers of one half of the multilayer can form dual "A-DMR" pairs with the even n layers from the other half (and visa-versa), and a close approximation to the optimal A-DMR state will still be maintained.

0039Hence, from the multiple viewpoints of self-bias capability at sub-micron element heights, with high, but nonextreme optimal sense current densities (that is, J_opt ≦ 3 x 10⁸ amp/cm²) with reasonable element height tolerance, with "built-in" stability of the optimum bias configuration state at no additional cost in fabrication complexity or signal loss, with high sensitivity (ε ≧ 50%, and ΔR/R ≈ 10%) and unsurpassed linear resolution capability, the GMR-DMR reproduce head is potentially uniquely better suited for future ultra high density magnetic disk recording applications at sub-micron element heights and track widths than any currently known alternative MR head designs.

0040Regardless of the specific details of how the GMR-DMR structure is fabricated or processed, or the origin of the interlayer antiferromagnetic coupling (exchange or magnetostatic), the essential DMR-like self-biasing characteristics should remain valid for GMR films with continuous or noncontinuous film layers and/or randomly oriented anisotropy similar to those described in the above-referenced Hylton et al. article.

0041Other possible alternative designs to improve bias uniformity (across either half thickness), including variable thickness magnetic layers and/or conductor layers to intentionally vary H_afc (or to make it uniform) across the thickness of the layers, as well as changing the direction of induced anisotropy, are also within the scope of the present invention.

PARTS LIST FOR FIGURES 1 - 10

0042GMR-DMR reproduce head or sensor structure 10 substrate 12 buffer layer 14 outer magnetic, conducting layers 16, 18 interior magnetic, conducting layers 20 nonmagnetic electrically conducting layers 22 central, nonmagnetic, conductive layer 24 buffer layer 26 central magnetic layer 28 conductive passivation layer 30 patterned current lead structures 32, 34