DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0099] Embodiments of the invention are described in detail hereunder with reference to the drawings.

First Embodiment

[0100] First mentioned is the embodiment of the invention in which the free layer (first ferromagnetic layer) is thinned.

[0101] The problems with the technique of “thinning the free layer” which the present inventors have recognized in the process of achieving this embodiment of the invention are described in detail.

[0102] As so mentioned hereinabove, remarkable increase in the sensitivity of magnetoresistance effect devices is realized not only by increasing the MR ratio but also by reducing the thickness of the free layer (that is, by reducing the product of Ms×t). In a broad way, the output increases, being in reverse proportion to the product of Ms×t of the free layer. However, the present inventors' own investigations have verified that the technique of thinning the free layer brings about the following problems.

[0103] The first problem is that the bias point designing in an applied sense current is difficult. When all the magnetic fields applied to the free layer being driven are summed up and when the bias point is in the center of the linear inclination of a transfer curve, the biasing condition will be the best. However, when the free layer is thinned, the inclination of the transfer curve becomes steep. In that condition, it is extremely difficult to lead the bias point to the center of the linear region of the transfer curve. In a bad bias point condition, asymmetric signals will be formed, and in a worse condition, no output level could be taken.

[0104] The second problem is that, if the free layer is thinned to an extreme degree according to a prior art technique, the MR ratio is greatly lowered. The reduction in the MR ratio causes the reduction in the reproducing output.

[0105] FIG. 7 A and FIG. 7B are conceptual views for explaining the two problems with conventional magnetoresistance effect devices. In those drawings, shown are the transfer curves of magnetic heads each having a magnetoresistance effect device. FIG. 7A shows one case where the free layer is thick; and FIG. 7B shows another case where the free layer is thin. As mentioned above, when the free layer is thinned, the inclination of the transfer curve becomes steep (that is, Hs becomes small), and the MR ratio decreases. As a result, ΔV becomes small. From FIG. 7 A and FIG. 7 B, known are the two problems noted above.

[0106] Of the two problems, one relating to the bias point could not be recognized with ease even though the film structure is determined. Therefore, film structure designing was extremely difficult. This time, we, the present inventors carried out various simulations, and corrected the errors in the data we obtained, on the basis of our experiences. As a result, we have succeeded in correct judgment of the bias point. The calculation of the bias point is mentioned below.

[0107] The bias point is shifted by various external magnetic fields applied to the free layer. The shift could approximate the sum total of ( 1 ) current magnetic field (Hcu), ( 2 ) the static magnetic field from the pinned layer (Hpin), ( 3 ) the interlayer coupling magnetic field from the pinned layer via a spacer (Hin), and ( 4 ) the stray magnetic field (Hhard) from the hard bias film for the purpose of imparting a longitudinal bias to the magnetoresistance effect film. Of those magnetic fields ( 1 ) to ( 4 ), the hard bias magnetic field ( 4 ) is relatively small. Having noted the sum of the magnetic fields ( 1 ) to ( 3 ), we, the present inventors have assiduously studied. The calculation formulae for the bias point which we employed this time are mentioned below.

B.P.= 50×( H shift/ Hs )+50 (1-1)

H shift=− H in+ H pin± Hcu (1-2)

Hs=Hd ^free +Hk (1-3)

Hd ^free =π ² ( Ms×t ) _free /h (1-3-1)

H pin=π ² ( Ms×t ) _pin /h (1-4)

Hcu= 2 πC×I _s /h (1-5)

C =( I ₁ −I ₃ )/( I ₁ +I ₂ +I ₃ ) (1-5-1)

[0108] B.P. to be represented by the formula (1-1) is the bias point [%] to be specifically noted herein. The best bias point is 50%. Including the margin, the practicable range of the bias point will fall between 40 and 60%. If the bias point oversteps the range, asymmetric signals will be formed. In worse cases, no output could be obtainable.

[0109] Regarding the relationship between the bias point value and the asymmetry, the asymmetry will be +10% or so when the bias point is 40%, and it will be −10% or so when the bias point is 60%. As will be mentioned hereunder, the best bias point in calculation does not fall between 40 and 60%, but falls between 30 and 50% in experiences.

[0110] FIG. 8 is a graph of calculated bias point values versus head reproducing signal waves. As shown, when the bias point falls between 30 and 50%, the asymmetry is relatively small, and the signal profiles are good. However, when the bias point is outside the range, the asymmetry becomes great, as in FIG. 8 , and the signal profiles are not practicable.

[0111] As in the formula (1-2), Hshift, is the sum of the magnetic fields [Oe] applied to the free layer. As in FIG. 7 , Hs is the inclination of the transfer curve.

[0112] FIG. 9 is an explanatory view indicating magnetic fields acting on the free layer.

[0113] Hd ^free is an antimagnetic field for the free layer at a certain MR height. h is the MR height [μm]. Hpin is the pinned layer stray magnetic field from the pinned layer to the free layer. (Ms×t) _free is the product of the total saturation magnetization, Ms, and the thickness, t, of the free layer. (Ms×t) _pin is the product of the saturation magnetization and the thickness of the net pinned layer (for Synthetic AF, the difference in the magnetic thickness between the upper and lower pinned layers).

[0114] Hcu is the current magnetic field applied to the free layer. Is is the sense current [mA]. The coefficient, C, in the formula (1-5-1) is the ratio of the partial current flow running through the upper layer overlying the free layer to that running through the lower layer underlying it.

[0115] FIG. 10 is a conceptual view indicating the partial current flows I ₁ to I ₃ running through the layers.

[0116] For simplifying the calculation, the influences of the edges of the ABS plane and those of shields are not taken into consideration. In our experiences, we, the present inventors have found that the bias point values as estimated in calculation are shifted to the minus side by about 10% from those in actual heads. Considering the fact that the usable bias point range falls plus/minus 10% of the best bias point value, it can be said that the good bias point range will fall between 30% and 50% to be obtained in calculation. Accordingly, when the bias point falls between 30% and 50%, as obtained according to the calculation shown above, it could be judged that the bias point values within the range are good for practical use.

[0117] The problems with some conventional spin valve films are described in detail hereunder, with reference to the bias point calculation formulae mentioned above.

Comparative Case 1: Ordinary Spin Valve (with Neither High-Conductivity Film nor Synthetic AF)

5 nanometer Ta/2 nm NiFe/0.5 nm Co/2 nm Cu/2 nm CoFe/7 nm IrMn/5 nanometer Ta (1)

[0118] The formula (1) indicates the laminate structure of the spin valve, in which are shown the elements constituting the layers and their thicknesses (nanometers). The film of this Comparative Case is a modification of a prior art spin valve film in which only the free film is thinned. The bias points in this film constitution are calculated.

[0119] Of the bias point formulae (1-1) to (1-5) noted above, the current magnetic field of the formula (1-5) is the most difficult to obtain. This is because the current flow ratio, C, of the formula (1-5-1) is difficult to obtain. In the thinned film, the specific resistance of each layer is influenced by the crystallinity and the current distribution, and significantly differs from the specific resistance of the bulk. For practicable calculation as much as possible, we, the present inventors took the following measure and succeeded in obtaining the accurate current flow ratio, C.

[0120] For obtaining the specific resistance of each layer, a spin valve film sample having the constitution noted above is prepared. For obtaining the specific resistance of a predetermined layer, a few samples in which the thickness of the layer is varied by plus/minus 2 nanometers are prepared. In those samples, the relationship between the thickness of the layer and the conductance is obtained through linear extrapolation. The reason for the process is because actual data could not be obtained according to the well-known technique of obtaining the specific resistance of thin, single-layer films. In order to minimize the influences of crystallinity and those of current distribution, we, the present inventors have found through our experiments that the best way for accurate determination is to prepare film samples in which even the overlying and underlying layers are of practicable materials and to determine the conductance difference within the small thickness range as mentioned above.

[0121] The specific resistance of each layer as obtained according to the method is influenced little by the crystallinity and, in addition, could cancel the influences of current distribution thereon. Therefore, the data of the current flow ratio, C, of the formula (1-5-1) obtained according to this method are much more accurate than those obtained in a simple conductor in which is used the specific resistance of single-layer films. According to the method, it has become possible to calculate and estimate the current magnetic field with high accuracy, which, however, is difficult in the prior art technique.

[0122] The data of the specific resistance of each layer as obtained according to the method mentioned above are as follows: NiFe has 20 μΩcm; CoFe 13 μΩcm; spacer Cu 8 μΩcm; IrMn 250 μΩcm. If the underlayer of Ta (tantalum) is thick, its specific resistance will greatly vary through crystallization. The cap Ta is much influenced by the surface oxides. Therefore, their accurate data could not be obtained. The specific resistance of the Ta layer is presumed to be 100 μΩcm. Based on those data, the current flow ratio of each layer is obtained, and the current magnetic field Hcu is calculated according to the formula (1-5).

[0123] The value of Hin is 25 Oe, as measured. Hpin is obtained according to the formula (1-4).

[0124] In the film constitution of this case, the height is shortened while the thickness of the pinned layer is thick. Therefore, the stray magnetic field, Hpin, from the pinned layer to the free layer is large. In addition, since more current flows above the free layer than below it, the current magnetic field Hcu applied to the free layer is large. Accordingly, for designing the bias point, it may be employable to control the bias point by canceling the large Hpin by the large current magnetic field Hcu.

[0125] The bias point values as calculated on the basis of the data obtained above are shown in Table 1. The sense current is 4 mA. 1

[0126] As is known from Table 1, the bias point falls between 61 and 70% when the MR height falls between 0.3 and 0.5 micrometers, and this oversteps the calculated best bias point range.

[0127] FIG. 11 is a conceptual view showing the condition of the bias point in Comparative Case 1. It is understood that reducing the MR height results in shifting of the bias point to the antiferromagnetic site (to the site larger than 50%). The MR height inevitably fluctuates, owing to the mechanical polishing. It is understood that the MR height distribution lowers the yield. Qualitatively, the reason is because the bias point is controlled in the extremely unstable method where the large pinned layer stray magnetic field Hpin is canceled by the large current magnetic field Hcu.

[0128] Except for the bias point, the film of this Comparative Case has still another essential problem. The problem is that the ultra-thin free layer to which the present invention is directed lowers the MR ratio. Through our experiments, we, the present inventors have found the fact that the MR ratio in the devices having a thinned free layer is extremely lowered after thermal treatment, and this is a serious problem with the devices. For example, in the film constitution of Comparative Case 1, the MR ratio in the as-deposited condition is around 11%, but is 5.6% after thermal treatment. That is, the latter is about ahalf of the former. In that condition, spin valve films practicable in high-density recording/reproducing systems could not be realized.

[0129] In the spin valve film of this Comparative Case, the layers are all thinned, and the sheet resistance of the film is around 30 Ω, and is large. In view of the electrostatic discharge (ESD), the film is not practicable. As well known, ESD increases with the increase in the resistance.

[0130] From the above, it is understood that the film of Comparative Case 1 is not practicable at all in high-density recording heads.

Comparative Case 2: U.S. Pat. No. 5,422,591 (with Spin Filter but No Synthetic AF)

5 nanometer Ta/x nm Cu/1.5 nm NiFe/2.3 nm Cu/5 nm NiFe/11 nm FeMn/5 nanometer Ta (2)

[0131] In order to improve MR in the ultra-thin free layer therein, a spin valve film has been proposed, in which a high-conductivity layer is laminated on the free layer at the side opposite to the side of the nonmagnetic spacer layer. For example, referred to are U.S. Pat. Nos. 2,637,360, 5,422,591 and 5,688,605.

[0132] The film (2) is an example of the spin valve film based on U.S. Pat. No. 5,422,591. In this spin valve film, the Cu layer adjacent to the free layer at the side opposite to the side of the spacer Cu is thickened, whereby the mean free path for up-spin is prolonged to increase the MR ratio. However, if the Cu layer is too much thickened over the mean free path, it will be a simple shunt layer. Therefore, in this film, the MR ratio peak appears at a certain Cu layer thickness. Based on this phenomenon, one problem with the film of Comparative Case 1, that is, the reduction in the MR ratio owing to the ultra-thin free layer could be overcome in some degree.

[0133] However, the film constitution of the spin valve film (2) based on U.S. Pat. No. 5,422,591 has two problems. One is the problem with the bias point, and the other is the problem with the thermal stability for the MR ratio.

[0134] First, regarding the viewpoint of the bias point, U.S. Pat. No. 5,422,591 has no disclosure of direct description or indirect suggestion. The constitution of the film (2) could not be used at all in practical heads. The reasons are mentioned in detail hereunder.

[0135] In the same manner as in Comparative Case 1, the current magnetic field, Hcu, is obtained on the basis of the experimental data of the specific resistance of each layer. In this case, the specific resistance of Ta is presumed to be 100 μΩcm, and the experimental data of FeMn, NiFe, spacer Cu and subbing Cu are 250 μΩcm, 20 μΩcm, 8 μΩcm and 10 μΩcm, respectively. The sense current is 4 mA. Though not described in U.S. Pat. No. 5,422,591, Hin has been found to fall between 15 Oe and 25 Oe through our experiments. In this case, Hin is 20 Oe.

[0136] The bias point is calculated for high-density recording heads for which the size of the device is as follows: track width=0.5 μm, MR height=0.3 to 0.5 μm. The data are in Table 2. 2

[0137] In this constitution, the pinned layer stray magnetic field from the pinned layer to the free layer is extremely large, and the bias point is readily shifted to the plus side. As is known from the data of the bias point in Table 2, in the samples with no spin filter effect in which the subbing Cu thickness is zero, the bias point falls between 111% and 126% at the MR height of from 0.3 to 0.5 μm. This means that the samples produce no output. This situation gets worse when the subbing Cu layer thickness is thick as shown in table 2.

[0138] FIG. 12 is a conceptual view of the bias point versus Hin, Hpin and Hcu on a transfer curve. As Hpin is large, the bias point oversteps the level when the current is zero. This constitution is so designed that the bias point is forcedly reduced to 50% by means of an applied current magnetic field. In this constitution, however, since the underlayer is a high-conductivity layer of Cu, I ₃ in FIG. 10 shall be large and the current magnetic field, Hcu, to be obtained according to the formula (1-5) shall be small. In other words, the bias point controlling method is to cancel the large Hpin by the small Hcu that is opposite direction to Hpin, which is impossible to attain a good bias point value. It is further known from Table 2 that the increase in the subbing Cu thickness results in further increase in the bias point.

[0139] Through our experiments mentioned above, we, the present inventors have found that the bias point designing is quite difficult in the constitution of U.S. Pat. No. 5,422,591, and that forming the high-conductivity Cu layer as the underlayer of the comparative case makes the bias point more impracticable.

[0140] From the viewpoint of the thermal stability for the MR ratio, the film of U.S. Pat. No. 5,422,591 is not a practical one. As so described in comparative case, the MR ratio in the as-deposited film surely increases owing to the spin filter effect. However, after the thermal treatment for head fabrication, we, the present inventors have found that the MR ratio in the film of the comparative case is greatly reduced as the phenomenon peculiar to ultra-thin free films. This is a serious problem, if high output for high-density recording is intended.

[0141] In fact, the MR ratio in the as-deposited film of one example of U.S. Pat. No. 5,422,591 (this is film (2)) is 1.8% when the subbing Cu thickness is 1 nanometer. However, after the thermal treatment in our simulation, the MR ratio in the U.S. Pat. No. 5,422,591 lowered to 0.8%. As will be mentioned hereunder, the reason for the MR ratio reduction will be essentially because of the antiferromagnetic film of FeMn in U.S. Pat. No. 5,422,591. Even if the ultra-thin free layer, with which it is difficult to realize high MR, is used in the spin valve film so as to forcedly increase the MR ratio in the film owing to the spin filter effect, the film will be quite useless in practical applications. It is understood from the data noted above that the simple spin filter effect could not realize spin valve films having an ultra-thin free layer and exhibiting high MR ratio.

Comparative Case 3: JP-A 10-261209

5 nanometer Ta/3 nm Cu/1 nanometer Ta/5 nm NiFe/2.5 nm Cu/2.5 nm Co/10 nm FeMn/5 nanometer Ta (3)

[0142] In the film (3) disclosed in JP-A 10-261209, the Cu shunt layer disposed adjacent to the free layer via Ta therebetween is, being different from the layer as intended for the spin filter effect for the MR ratio as in U.S. Pat. No. 5,422,591 of Comparative Case 2, intended for stabilizing the asymmetry by reducing the current magnetic field Hcu and by retarding the bias point fluctuation owing to sense current. This idea will be effective in the region where the free layer is relatively thick as in the film (3), but is quite ineffective in the case of ultra-thin free layers to which the present invention is directed, in view of the bias point and the MR ratio. Based on this idea, practicable films having an ultra-thin free layer could not be obtained at all. The reasons are mentioned below.

[0143] First, regarding the bias point, when Hs is extremely reduced by the use of the ultra-thin free layer, as in the film (2) of Comparative Case 2, the best bias point could not be realized even though the current magnetic field Hcu is reduced, if the pinned layer stray magnetic field is large. The advantage of the structure of the film (3) is that, when the free layer is thick, or that is, when Hs is relatively large, then the best bias point having been once obtained depends little on the sense current. However, when the free layer in the film constitution of (3) is much reduced, it is naturally impossible to realize the best bias point. In other words, when the thickness of the free layer in the film constitution of (3) is reduced to 4.5 nanometers or smaller so as to make the free layer applicable to high-density recording systems, the bias point shall be shifted to the plus side.

[0144] To verify the fact, the calculated bias point data of the film constitution are shown in Table 3. 3

[0145] Hin for the data calculation is 10 Oe. As in Table 3, it is understood that the bias point in the film constitution of Comparative Case 3 is naturally shifted to the plus side even when the thickness of the NiFe film therein is 5 nanometers, and the film constitution is not well designed. In this, the bias point is much more shifted to the plus side when the thickness of the free layer of NiFe is thinned to be 3 nanometers.

[0146] FIG. 13 is a conceptual view showing the determinant factors for the bias point in Comparative Case 3. As illustrated, since the current magnetic field Hcu only is reduced while Hpin is still large, no bias point is obtained at all in the region where the free layer is thick. Specifically, since the best bias point appears at the site where the total sum of the current magnetic field Hcu, the interlayer coupling magnetic field Hin and the pinned layer stray magnetic field Hpin is zero, the film designing of such that the current center is made nearer to the free layer while only the current magnetic field is made zero, as in the constitution of (3), is quite meaningless.

[0147] The second problem with the constitution of (3) is that the film (3) could not have high MR ratio necessary for high-density recording. Specifically, in the constitution of (3), since a diffusion-preventing layer of a material having a relatively high resistance is put between the high-conductivity layer and the free layer, the film (3) could not enjoy the spin filter effect for MR, such as that in U.S. Pat. No. 5,422,591 noted above, when the free layer is an ultra-thin one. In the region where the free layer has a thickness of at most 4.5 nanometers and where the present invention is especially effective, as will be described in detail hereunder, the film (3) will be useless as the MR ratio therein is lowered.

[0148] For the two reasons mentioned above, the idea of the film constitution (3) is exclusively for the region where the free layer is relatively thick, and it is understood that the idea is useless in preparing practicable films having an ultra-thin free layer.

Comparative Case 4: Synthetic AF with No Spin Filter

5 nanometer Ta/2 nm NiFe/0.5 nm CoFe/2 nm Cu/2.5 nm CoFe/0.9 nm Ru/2 nm CoFe/7 nm IrMn/5 nanometer Ta (4)

[0149] In this Comparative Case, employed is a Synthetic AF structure for the purpose of improving the pinning characteristics. The two ferromagnetic layers are antiferromagnetically coupled to each other via Ru (ruthenium) One of the two ferromagnetic layers is pinned to the other via an antiferromagnetic layer existing therebetween. Even if the uni-directionally anisotropic magnetic field Hua is too small for normally pinned structures, but if it is on a certain level, such a small Hua is applicable to the Synthetic AF structure, and the pinning resistance of the Synthetic AF structure is high. As previously mentioned hereinabove, in the Synthetic AF structure, the magnetization directions of the upper and lower ferromagnetic layers as coupled via Ru are opposite to each other, and the coupling magnetic field is several kOe and is much larger than the medium magnetic field for head operation. Approximately for the magnetic moment running outward, therefore, it is considered that the difference in Ms×t between the upper and lower pinned layer will be equal to the net moment. Specifically, it is possible to reduce the influence of the magnetic field strayed from the pinned layer to the free layer, and it is expected that the reduction is advantageous for the bias point (see U.S. Pat. No. 5,465,185).

[0150] For example, in this Comparative Case, it is considered that the net pinning thickness will be equivalent to the pinned layer having a thickness of 0.5 nanometers, and it is possible to realize the pinned layer stray magnetic field equivalent to the thin pinned layer, which, however, could not be realized in normal pinning structures. Ideally, when the upper and lower pinned layers are controlled to have the same product of Ms×t, then the pinned layer stray magnetic field could be zero. It has heretofore been considered that only the reduction in the pinned layer stray magnetic field will be satisfactory for good bias point designing for spin valve films for high-density recording. This time, however, we, the present inventors have found that, if only the Synthetic AF structure is employed, it is impossible to realize stable bias points in ultra-thin free layers applicable to high-density recording. The matters we have found are described in detail hereunder.

[0151] FIG. 14 is a conceptual view showing the determinant factors for the bias point in this Comparative Case 4. Specifically, in the constitution of this Comparative Case, the free layer is positioned, being much far from the current center of the current distribution in the spin valve film. In this, therefore, the current magnetic filed Hcu is extremely large. Hin is at most 20 Oe or so, and the pinned layer stray magnetic field is extremely reduced owing to the Synthetic AF structure employed. This means that the film constitution of this Comparative Case 4 is nearly in the just bias condition in the absence of current. When current is applied to the spin valve film of this constitution, and when the applied current is increased, then the film will much overstep the just bias condition owing to the increased current magnetic field Hcu.

[0152] The calculated bias point data of this Comparative Case are shown in Table 4. 4

[0153] Hin is 20 Oe. As so anticipated, it is understood from Table 4 that the bias point could not fall the range of from 30 to 50% irrespective of the direction of the current flow.

[0154] For obtaining the just bias in this constitution, one means may be taken into consideration, which comprises minimizing as much as possible the pinned layer stray magnetic field, or that is, making the upper and lower pinned layers have the same thickness in the Synthetic AF structure thereby to make the pinned layer stray magnetic field nearly zero. It further comprises enlarging Hin as much as possible so that the just bias could be in the current magnetic field for canceling the enlarged Hin. However, this means is undesirable. The enlarged Hin has some negative influences on external magnetic field response. The enlarged Hin shifts the linear region of the response and, in addition, reduces it. It will be good to control Hin to be small. However, fabricating spin valve films while the value is forcedly controlled to be an unnaturally large and constant one is extremely difficult in mass production and is unfavorable.

[0155] Since no high-conductivity layer is provided on the free layer at the side opposite to the side of the spacer adjacent to the free layer, the MR ratio is lowered when the free layer is an ultra-thin one, for the same reasons as in Comparative Case 1. Therefore, the film constitution of this Comparative Case 4 could not ensure satisfactory output, when applied to high-density recording heads. This is the essential problem with this film constitution.

[0156] For the two reasons of bias point and high output mentioned above, the spin valve film merely incorporating the Synthetic AF constitution could not realize the use of ultra-thin free layers therein for high-density recording.

[0157] As has been described in detail hereinabove, we, the present inventors have clarified, through many simulations in various current magnetic fields, that the film structures of Comparative Cases 1 to 4 could not attain stable bias point and satisfactorily high output for spin valve films having an ultra-thin free layer for high-density recording. Through further studies and investigations, we have achieved the present invention. The constitution of the invention is described in detail hereunder.

[0158] FIG. 15 is a graph of the free layer thickness dependence of the bias point in the spin valve films of the invention, as compared with that in the spin valve films of the above-mentioned Comparative Cases. It is understood that the spin valve films of the Comparative Cases are all problematic in the bias point. The best bias point for spin valve films falls between 30and 50%. For satisfactorily high sensitivity, the bias point must fall within the defined range at a low Ms×t free layer.

[0159] However, the bias point in the Comparative Cases significantly oversteps the preferred range in the condition of low Ms×t. In addition, in the Comparative Cases, the bias point fluctuation relative to Ms×t is extremely large, and it is understood that the bias point is difficult to control in those Comparative Cases.

[0160] As opposed to the films of the Comparative Cases, it is understood that, in the film of Example 1 of the invention, the bias point fluctuation is extremely small relative to Ms×t, and the bias point is all the time within the preferred range.

[0161] In FIG. 15 , the calculated bias point values in Comparative Case 1 do not fall within the range between 30% and 50% even in the region where Ms×t is not smaller than 5 nanometer Tesla. This is because, in the actual region of low recording density for which a free layer having Ms×t of at least 5 nanometer Tesla is used, the MR height is large. Concretely, this is because the MR height for low recording density heads is larger than the range of from 0.3 μm to 0.5 μm for the high recording density to which the present invention is directed.

[0162] Anyhow, it is obvious that, in the region where Ms×t is at most 5 nanometer Tesla, the bias point designing for the films of the invention is much superior to that for the films of the Comparative Cases.

[0163] FIG. 16 is a graph of the MR ratio in the structures of Comparative Cases 1 to 4 with the product of Ms×t only in the free layer being reduced. In this, the MR ratio as plotted in the vertical axis is nearly proportional to the vertical axis of FIG. 9 for the transfer curve. For comparison, the data of Examples 1 and 2 of the invention to be mentioned hereunder are also plotted in FIG. 16 .

[0164] The film samples of Comparative Cases 1 to 4 and those of Example 1 of the invention were prepared, in which the thickness of the free layer of NiFe was varied for the varying Ms×t. The film samples of Example 2 were prepared by varying the thickness of the free layer of CoFe. All the samples were thermally annealed at 270° C. for 10 hours in a magnetic field of 7 kOe, and their data were measured.

[0165] In Comparative Case 2 and Examples 1 and 2, the high-conductivity layer is of Cu, having a thickness of 2 nanometers. The points of Ms×t in the free layer as indicated by the arrows in FIG. 16 are for the films ( 1 ) to ( 4 ) of Comparative Cases mentioned above. For the Ms×t in the free layer in all samples, Ms of NiFe is 1 T and Ms of CoFe is 1.8 T. All the free layer thickness are expressed in terms of thickness with Ms of 1 Tesla.

[0166] In the films of Comparative Cases 1, 3 and 4 where no high-conductivity layer is provided on the free layer, MR ratio is greatly lowered with the reduction in Ms×t in the free layer. These films could hardly ensure high output capable of satisfying high-density recording.

[0167] In the film of Comparative Case 2 having a high-conductivity layer, the free layer Ms×t dependence of the MR ratio is relatively small. However, since the antiferromagnetic layer in this film is of FeMn, not containing a noble metal, the thermal stability for MR ratio in thermal treatment is low. With such small MR ratio, the film could not ensure high output for high-density recording.

[0168] In the films of Comparative Case 2 and Comparative Case 3, if a layer of Co or CoFe having a thickness of 0.5 nanometers is inserted between the spacer of Cu and the free layer of NiFe, the MR ratio will increase by 1 to 2% above the data in the graph of FIG. 16 . Even if so, however, the Ms×t dependence of the MR ratio is still the same as that in the single-layer NiFe free layer. Anyhow, small MR ratio will do well in the region where Ms×t in the free layer is small.

[0169] On the other hand, in the films of the invention where a high-conductivity layer is provided adjacent to the free layer and the antiferromagnetic layer contains a noble metal, the thermal stability for the MR ratio after thermal treatment is good. The films of the invention produce high output well applicable to high-density recording. In particular, the difference in the MR ratio between the films of the invention and those of Comparative Cases is obvious in the region where Ms×t is smaller than 5 nanometer Tesla.

[0170] The magnetoresistance effect device of the invention is described in detail hereunder.

[0171] FIG. 1 is a conceptual view showing the sectional constitution of the magnetoresistance effect device of the invention. As illustrated, the magnetoresistance effect device of the invention comprises a high-conductivity layer 101 , a free layer 102 , a spacer layer 103 , a first ferromagnetic layer 104 , a coupling film 105 , a second ferromagnetic layer 106 and an antiferromagnetic film 107 , all laminated in that order.

[0172] In the device with that constitution, the free layer 102 is much thinned. On the transfer curve where Hs is small due to ultra thin free layer, Hcu, Hpin and Hin are all small Hpin−Hin=Hcu, and a good bias point can be realized in the device. In general, in ultra-thin free layers, high MR ratio is difficult to realize. The device of the invention has overcome this problem. The device has good thermal stability for MR ratio, and therefore can realize high-output heads.

[0173] Specifically, even though having the ultra-thin free layer for high-density recording, the spin valve film constitution of the invention can realize good bias point and can maintain high MR ratio. Therefore, the film stably produces high output. Concretely, in bias point designing, the condition of Hpin−Hin=Hcu is realized, and the film has good bias point. Reducing all Hpin, Hin and Hcu is important for stably realizing the condition of Hpin−Hin=Hcu.

[0174] Regarding Hpin, the film has a so-called Synthetic AF structure where the two ferromagnetic layers are antiferromagnetically coupled to each other. In this, therefore, Hpin is derived from only the difference in the magnetic thickness between the two layers of the first and second ferromagnetic layers, and can be reduced.

[0175] From the formula (1-4), it is understood that reducing (Ms×t)pin in the pinned layer is effective for reducing Hpin.

[0176] However, for bias point designing in ultra-thin free layers, only the reduction in Hpin is meaningless. For this, the reduction in the current magnetic field Hcu is indispensable. Therefore, the nonmagnetic high-conductivity layer is provided on the free layer at the side opposite to the side of the spacer layer, whereby the center of the current distribution in the spin valve film can be near to the free layer and Hcu can be thereby reduced. This is because, in the formulae (1-5) and (1-5-1), when I ₃ is increased for the top-type spin valve film (I ₁ is increased for the bottom-type spin valve film) and the current flow ratio C is lowered, then the current magnetic field Hcu can be reduced. Another significant function of the nonmagnetic high-conductivity layer is to maintain high MR ratio in the ultra-thin free layer, to which the invention is directed, owing to the spin filter effect. Specifically, by providing the nonmagnetic high-conductivity layer, the difference in the mean free path of up-spin can be kept large between the parallel condition and the antiparallel condition for the magnetization directions of the free layer and the pinned layer adjacent to the spacer.

[0177] For stably realizing the condition of Hpin−Hin=Hcu, the reduction in Hin is also important. For realizing the high MR ratio owing to the high-conductivity layer as provided adjacent to the ultra-thin free layer (spin filter effect), it is important to thin the spacer. However, in general, Hin will increase with the reduction in the spacer thickness and with the reduction in the free layer thickness. Overcoming this problem, it is important to use the device of the invention at Hin falling within a range of from 0 to 20 Oe or so.

[0178] FIG. 2 is a schematic view of the transfer curve given by the spin valve film of the invention. Even in the small transfer curve for which the ultra-thin free layer is used to give small Hs, all of Hpin, Hcu and Hin are reduced, and it is possible to design the condition of Hpin−Hin=Hcu. Therefore, it is possible to settle the bias point in a good site of around 50% (good bias point around 40% in the value calculated by our method). In addition, since the film incorporates the high-conductivity layer exhibiting the spin filter effect, it still maintain high MR ratio even in the ultra-thin free layer. The value in the vertical axis in FIG. 2 is satisfactorily high.

[0179] Next, the determinant factors for the bias point, namely the parameters of Hpin, Hin and Hcu are described in detail.

[0180] First, low Hcu is referred to. As previously described hereinabove, the high-conductivity layer is provided on the free layer at the side opposite to the side of the spacer, whereby the value C in the formula (1-5) is reduced and the current magnetic field Hcu is reduced. One co