MAGNETIC MEDIA PERFORMANCE:
`CONTROL METHODS FOR CRYSTALLINE TEXTURE AND ORIENTATION
`
`DAVID N. LAMBETH, WEI YANG, HENG GONG, DAVID E. LAUGHLIN, BIN LU, LI(cid:173)
`LIEN LEE, JIE ZOU, and PETERS. HARLLEE
`
`ECE and MSE Departments, Data Storage Systems Center, Carnegie Mellon University,
`Pittsburgh, PA 15213, lambeth@ece.cmu.edu
`
`ABSTRACT
`
`Currently, and during the past few years, areal recording densities have doubled
`approximately every 18 months. This has resulted in the recent introduction of products
`exceeding 4 Gigabit/in2
`. Furthermore, there have been recent laboratory demonstrations of over
`11 Gigabit/in2
`. This rapid technological pace is forcing media producers to seek new methods to
`control the media magnetic properties. It has recently been pointed out that due to the required
`signal to noise ratio and the resulting continued reduction in grain size the industry will soon be
`faced with the onset of thermal instabilities to data retention. Since the medium properties could
`limit the areal density of most recording systems, a systematic design approach toward media
`invention is necessary. Very small magnetic grains of near crystalline perfection will be required
`in order to achieve the coercivities and noise requirements for the next doubling of areal density
`(25 Gigabit/in2
`). Following this not only will crystalline perfection be required, but extremely
`uniformly sized and singly oriented grains will be required to approach 50 Gigabit/in2
`. Over the
`past few years we have taken an approach of controlled growth of the magnetic microstructure of
`thin film media to accomplish this. Here we provide an overview of the guiding media design
`philosophy and discuss materials issues, multi-layered thin film material structures and processing
`techniques which are used to control the microstructure and magnetic properties of Co-alloy
`films. Efforts toward epitaxial growth of multiple thin film layers on single crystalline Si is
`discussed as a method of achieving perfect crystallites of various highly oriented thin films. In the
`case of non-cubic materials, such as magnetic hep Co-alloys, these films have well defined axial
`directions determined by the substrate and multiple thin film epitaxial relationships.
`
`INTRODUCTION AND MOTIVATION
`
`The continuing market force driven need for improved hard disk data storage is evident by
`the sudden change in the compound annual growth rate of areal densities in the early 1990' s from
`20 to 60%. This rate is still approximately 60% and there are no signs of abatement. The single
`largest, obvious, technical factor making this possible was the change from particulate disk media
`to sputtered thin film media. This smoother media surface has allowed continual decreases in the
`head to disk fly height to the present 50 nm and less. During the mid 1990's, even as trackwidths
`were shrinking, head disk interface technologies such as robust 15 nm thin CNx or CHx
`overcoats, laser produced mechanical texture and quasi-contact slider operation enabled reduced
`head to disk spacing and enabled the ubiquitous inductive head transducer to continue to perform
`adequately. With the advance to the more sensitive, low noise, magnetoresistive, and recently the
`spin valve device, record-playback transducer technology is returning media noise to being the
`limiting factor to further increases in areal density. While various laboratory demonstrations, such
`as IBM's recent announcement of 11 Gigabit/in2 recording, have been used to herald the next
`generation of recording densities a more accurate measure of progress is provided by monitoring
`
`Mat. Res. Soc. Symp. Proc. Vol. 517 © 1998 Materials Research Society
`
`181
`
`TDK Corporation Exhibit 1013 Page 1
`
`

`
`actual product performance. As of early 1998, the highest areal density of a commercial product
`fell very close to the 60% growth curve at 4.1 Giga bit/in2 . Concurrent with the progression of
`areal density has been a remarkable increase in data transfer rate and disk rotational frequency to
`decrease access times. This bandwidth increase, along with the introduction of the spin valve
`transducer, and the higher areal densities conspire to require even lower noise media.
`These recent commercial improvements in longitudinal recording could not have been
`possible without the significant improvements already made in the recording media magnetic and
`microstructural properties. Ultimately, the achievable areal recording density should be
`determined by the media signal to noise ratio, which ifthe recording system is designed properly,
`is largely determined by the media thin film microstructure. For ideal media where the magnetic
`particles are totally non-interacting one might roughly estimate the media signal power to noise
`power ratio, SNR,p, (zero to peak signal to rms noise ratio) to be proportional to the average
`number of particles, N, sensed by the recording transducer. This assumption is based upon the
`concept that the large number of particles in a given volume is described by a Gaussian probability
`distribution and to determine the noise one merely estimates the variance to the average number
`of particles sensed. The signal power then goes as N2 while the noise power goes as N. Hence,
`for a differentiating transducer system, I 000 particles would imply a SNR,p of 27db, a value that
`is sometimes viewed as necessary for an acceptable error rate. Therefore, assuming sufficient
`transducer sensitivity, in order to maintain an adequate SNR as the areal density is doubled, N
`would remain constant and, the number of particles per unit area would need to double. For thin
`film media, in which the particle extends through the thickness of the thin film, the particle or
`grain surface diameter would have to decrease by the square root of two.
`Charap and Lu [I] recently modeled the limits of areal recording density based upon the
`concept that if a magnetic particle's anisotropy energy density-volume product is made too small
`it will spontaneously reverse due to thermal fluctuations. Their estimate of the limit of this
`product divided by the Boltzman energy is Ku V/kB T > 60. At the same time since the media
`coercivity is proportional to the anisotropy field, He a Hk = 2MJK.u
`the anisotropy energy
`density cannot be increased to a point where He would be greater than the maximum available
`transducer record fields. These field levels are determined as a fraction of the Ms of available
`head transducer materials. Based upon these boundary conditions: a maximum Ku to allow
`recording, a minimum thermally stable grain volume and a fixed number of grains to provide the
`required SNR, they estimated that a limit to long term data stability may occur around 40
`Gigabit/in2 . We would like to suggest, however, that the statistical arguments concerning the
`number of particles required may be somewhat flawed. Unlike particulate tape media, thin film
`media is essentially volume filling. That is, there are no intentional voids in the media and the
`total magnetic moment sensed by the transducer would be essentially constant if the magnetic
`easy axes of all of the grains were in the same direction. In traditional rotating thin film hard disk
`media the easy axis is designed to be random in the plane of the disk to avoid modulation of the
`signal as the disk turns. Hence, at any position on the disk a large fraction of the grains have their
`easy axes either perpendicular or considerably off axis from the transducer recording direction. In
`the extreme case of totally non-interacting particles these grains contribute little or no output flux
`to the transducer and appear as magnetic voids. Hence, even if all particles were the same size
`and regularly spaced, mis-orientation would still provide a mechanism for fluctuation in the signal.
`Clearly these mis-orientations along with the distribution of grain sizes limits the SNR. By
`narrowing the size distribution or, better still, restricting it to a singly uniform size the SNR would
`improve. However, by orienting all the easy axes to a single direction, comparable to the
`transducers' sensitive direction, the noise would become dependent only upon the size distribution
`and would be further reduced. A smaller number of grains would be required to achieve the same
`
`182
`
`TDK Corporation Exhibit 1013 Page 2
`
`

`
`SNR and recording densities could be extended beyond the current suggested thermal stability
`limit. Since grain size distributions are typically skewed and the smaller particles are thermally
`unstable they should be eliminated. By eliminating these and the variance in particle number due
`to random orientation distribution it would seem to be reasonable to expect that the noise could
`be reduced by at least a square root of two and perhaps considerably more. A factor of two
`reduction would lift the predicted areal recording density limit to well over 50 Gigabit/in2 .
`The above arguments assume non-interacting particles and it would seem to be naive to
`ignore magnetostatic or intergranular exchange interactions in a discussion of noise. However, it
`is the goal in the media design to eliminate the intergranular exchange by isolating each grain.
`Hence, noise induced by magnetostatic interactions can be limited if the media and recording
`system are designed correctly, as discussed next.
`
`MAGNETIC RECORDING PHYSICS ISSUES
`
`While the playback signal is extremely dependent upon the head to medium spacing and
`transducer resolution (gap length), it is also proportional to the magnetic medium film thickness
`and the spatial gradient of the magnetization along the recording track. Hence, one would be
`inclined to argue that the larger the remnant-film thickness product (M,-0) the better. This is not
`the case however, as the recorded transition length increases with this quantity and magneto(cid:173)
`resistive playback transducers are easily driven into a non-linear regime of operation if the sensed
`field is too large. Hence, there is a maximum, and optimal M,-0, for the media determined by the
`head sensitivity function and the head to media spacing. For currently commercial high
`performance media M,-0 has been reduced to less than 0.5 milli-emu/cm2. This is beneficial to
`media design for two reasons: One is that, for exchange coupled media or for a system with a
`poor record head field gradient, the apparent media transition noise can be directly correlated to
`the demagnetization fields at the transition. The second reason is that the down-track (linear)
`recording density is limited by the finite flux reversal length. This length is measured by the
`transition parameter, ax, which is determined either via transition demagnetization forces or by the
`finite record head field gradient. The media noise is always lower if the coercivity is greater than
`the demagnetization field (He> Hd). In the ideal limit where the head is in contact with an zero
`thickness media and has a zero gap length the recorded magnetization profile would be a step
`function provided that He > Hd and that the media were homogenous. For the more realistic
`media microstructure the transition length would be determined by the characteristics of the set of
`grains that lie at the immediate location where the ideal transtion would have been. The transition
`length is then nominally the average grain size and the transition location variance (noise) is
`determined by the grain location, size and orientation distributions. In other words, even if media
`microstructure were made ideal, it would still be easy to incorrectly design a recording system to
`induce apparent media noise.
`As an example for disussion consider Figure 1 which shows the media noise power spectra
`of several differing media determined by the noise power spectral integration technique. These
`datum were obtained using a Read Rite Tripad head with a .22 micron gap and flying just above
`medium contact at approximately 25 nm (at 7.1 m/sec.). In each case it is noted that the noise
`power initially increases linearly with density as the noise power is dominated by the flux
`transition location jitter and increases linearly with recording frequency as transition density
`increases. The fact that there is noise even at zero frequency indicates the discrete non-zero size
`of the randomly oriented magnetic switching units or grains. The initial slopes of these curves are
`determined by the grain size, intergranular exchange coupling, the maximum demagnetization field
`(determined by M,-0), and the finite head field gradient. At higher recording densities it is
`
`183
`
`TDK Corporation Exhibit 1013 Page 3
`
`

`
`common that the noise increases supra-linearly as the individual flux transitions interact during the
`recording process and begin to interpercolate. The onset of this supra-linearly noise behavior
`typically occurs as the flux transition density approaches the transition length, a,. From a practical
`standpoint, but to some extent dependent upon the signal processing encoding technique, this
`noise limits the data transition spacing to approximately na,. At extremely high flux transition
`densities the noise again decreases as the media appears to be AC erased. A media with highly
`exchange coupled grains will show little DC frequency noise as the magnetization of all grains
`tend to align, while when AC erased the percolation effect is large. This DC result is a clear
`example that thin film media is volume filling and that this noise is not simply controlled by
`particle counting statistics. For a media with little exchange coupling the DC noise will correlate
`to the number density and orientation of the grains. When the media is designed with very small
`grain sizes and such that the Hd < He the head field gradient during the record process determines
`a, and not the demagnetization fields associated with the media. The lowest noise curves of
`Figure 1 indicate that these media have sufficiently small grains (low slope) and a small enough
`4nM, /He that the supra-linear increase in noise does not occur until well after the maximum
`measured kfci. Hence, the grain size (isolated magnetic switching unit size) in combination with
`the random in-plane orientation determines the slope of this curve, as well as, the DC erased noise
`power shown at low frequencies. Whereas, media with large grains will show a large low flux
`density noise slope and if the Hd > He the supra-linear noise behavior will set on early. The
`higher noise curves represent this type of media. The lowest noise media of this set is comparable
`to today's best commercial media. If a medium noise is not head field gradient limited, but limited
`
`4
`
`·····•····· 1900 Oe CoCrTa/Cr
`0.5 Gbisqin
`-m - 2280 Oe CoCrPt
`I Gb/sqin
`···· •· ·· 3000 Oc SmCoiCr
`(CMlJ)
`- -•- - 3000 Oe CoCrPt/Cr
`(NSIC- IBM)
`· · • · · 2750 Oe CoCrPtiCriNiAI
`fCMU - IBM)
`
`/
`
`•
`,.......
`• •
`.. ,.-
`
`/Y
`
`•
`-
`-~ ..
`
`~
`
`• •
`----
`•
`
`,,_ .r
`
`50
`
`200
`150
`100
`Linear Density (KFCI)
`
`250
`
`300
`
`Figure 1. Normalized media noise vs. flux reversal density for media with various degrees
`of exchange coupling, Mro, and head field gradient induced noise.
`
`184
`
`TDK Corporation Exhibit 1013 Page 4
`
`

`
`by the statistical nature of the medium grains, then the required average grain size, assuming
`random size and orientation distributions, for 10 Gigabit/in2 technology is estimated to be about
`15-20 nm.
`
`MEDIA CONSTRUCTION AND THE ROLE OF CRYSTALLINE TEXTURE
`
`To attain the highest recording densities the coercivity should be maximized. In order to
`accomplish this the media designer could choose a magnetic alloy to achieve a higher anisotropy
`constant, lower the 4nMs which also decreases the flux transition demagnetization effects, or gain
`better control of the microstructure to achieve the maximum potential anisotropy energy density
`of a given alloy. While studies have been performed on other alloys, in modern hard disk media
`hep Co alloys are almost exclusively chosen due to their corrosion resistance and high anisotropy
`constants. Second and third elements such as Ni, Cr, Ta, Nb, B or Pt are chosen to promote
`In current
`diffusion of non-magnetic elements to the grain boundaries during film processing.
`products Cr and Ta are widely used and have proven to be especially useful at providing magnetic
`grain to grain isolation, while Pt appears to increase the anisotropy constant, but is not as
`effective at providing isolation. Likewise, these non-ferromagnetic elements usually lower the
`magnetization by dilution. The Co alloy is typically deposited onto a thin film underlayer structure
`which induces both an hep Co phase and orients the crystalline c-axis by epitaxial growth. Perfect,
`defect free and isolated, hep crystalline grains of appropriate size insure that domain walls do not
`nucleate at grain boundaries, crystalline flaws or stacking faults to lower the coercivity, while the
`orientation of the c-axis determines the maximum achievable coherent rotation coercivity. Via
`modeling, Yang [2] has predicted the hystersis loop dependence upon the orientation of the c(cid:173)
`axis with respect to the film plane. For ideal Stoner-Wohlfarth particles with easy axes (c-axes)
`parallel to the applied record field the coercivity would be equal to the anisotropy field, Hk =
`2Ku1Ms, while for a random ensemble of particles with c-axes in the film plane the predicted
`coercivity is reduced to 0.51 of HK. However, if the c-axes are randomly oriented in all three
`dimensions the grains with axes out of plane, (or even only somewhat dispersed about the plane)
`have their magnetization forced back into the plane via demagnetization fields and these additional
`fields further reduce the coercivity. For the rotating longitudinal recording hard disk media
`format the singly directed ensemble is currently impractical and so the random two-dimensional
`structure is the most desirable. Hence, the choice of the underlayer texture upon which to
`perform epitaxial growth is critical in determining the Co alloy c-axis orientation.
`In addition
`these underlayer structures are critical in determining Co alloy crystalline quality and the grain
`size. Historically [3], and even though a number of other elements have been investigated, bee Cr
`and Cr alloys have been used almost exclusively for this purpose.
`Most current hard disk thin film magnetic media are constructed upon a highly polished
`NiP electrolessly plated AlMg, glass, or glass ceramic substrate by sputter deposition of a
`sequence of metallic layers. The exact structure depends upon the substrate, but usually consists
`of sequential depositions of a non-magnetic seed layer and underlayer followed by a magnetic Co(cid:173)
`alloy, followed by a ceramic-like protective coating (principally carbon, CHx or CNx), and finally a
`very thin lubricant. The seed layers that have been used include both oxides and metals depending
`upon the substrate and the manufacturer. The purpose of the seed and the underlayers are to
`buffer the substrate surface and to initiate the crystalline growth and texture of the very thin
`magnetic layer. Hence, their composition, interaction with the substrate, and the processing
`conditions are important in determining the microstructural interface to the magnetic layer.
`In
`addition, the objective of the seed and underlayer is to establish a controlled grain size for the
`growth of other epitaxial layers.
`
`185
`
`TDK Corporation Exhibit 1013 Page 5
`
`

`
`CoCrPt (0002) peaks
`
`4000
`
`3000
`
`If the atoms of a metallic thin film have sufficient surface mobility during the deposition
`process they arrange into a minimum surface energy configuration. For most simple lattice
`structures this implies a close packed atomic surface configuration. For fee or hep lattices the thin
`film surface will have a (111) or (0002) texture, respectively, while for the bee lattices a (110)
`texture results. Hence, for bee Cr alloys the (110) texture is commonly observed. As an example
`of this process consider
`Figure 2 which shows a
`series of x-ray diffraction
`8-28 scans of a CoCrPt
`film sputter deposited upon
`very
`thin, but varying
`of
`sputter
`thicknesses
`deposited Cr on a glass
`substrate.
`The glass
`substrate represents a high
`energy oxidized
`surface
`upon which arriving Co
`atoms would have limited
`mobility. By first placing a
`very thin Cr layer on this
`oxidized
`surface
`the
`mobility of the Co atoms
`can
`be
`significantly
`increased.
`However,
`below 5 nm the Cr is so
`thin that little or no crystalline texture has evolved and so it could, perhaps, be thought of as
`amorphous or very conducive to strain relaxation. Nevertheless, its presence alters the interfacial
`energy between the glass substrate and the Co alloy sufficiently to allow the thicker Co alloy film
`to seek its lowest energy close packed arrangement. Hence, while the (0002) texture is initially
`
`2000
`
`1000
`
`" ' - - - - - - - - 40 Acr
`- - - - - - - -3 0 Acr
`- - - - - - - - 20Acr
`..._ ____ ___ 5 Acr
`----~--- IO A Cr
`....... no Cr
`
`Figure 2. Close packed (0002) texture evolution of 40 nm thick
`CoCrPt grown on very thin Cr.
`
`Quad-crystal
`Co(10·1 )/Cr(11 O)
`
`Bi-crystal
`
`Co(11 ·0)/Cr(200)
`
`Uni-crystal
`
`Co(1 O·O)/Cr(112)
`
`Figure 3. Epitaxial orientational relationships between bee Cr and hep Co.
`
`186
`
`TDK Corporation Exhibit 1013 Page 6
`
`

`
`2.88 Cr El
`EV Ag • Cr
`
`Figure 4. Orientational relationships for
`Cr(l 10)/ Ag(l 11)/Si(l11 ).
`
`Si
`
`0 Si
`
`not present when little or no Cr is deposited, it appears quite strong as the Cr thickness becomes
`sufficient to allow the Co to wet the substrate, then again disappears as the Cr becomes thick
`enough to develop and maintain a texture upon which the Co alloy is forced to epitaxially grow in
`different directions. The fact that the Cr has not developed a strong texture of its own at 5 nm
`thickness results in the Co
`texture being random.
`A
`similar effect, of this natural
`Co texture evolution, would
`if
`the
`underlayer
`occur
`allowed wetting of the Co to
`the surface or did not present
`a crystalline structure suitable
`for any epitaxial growth. Ta
`films,
`which
`appear
`amorphous when very thin, are
`known to perform a similar
`fee Ni
`function
`for
`( 111)
`alloys and are typically used as
`initializing seed layers for spin
`valve
`transducers.
`Other
`crystalline metals will perform
`this task if the lattice spacing
`or crystalline symmetry is largely different from that of the following layer. Likewise, if the
`atomic mobility is limited by substrate to film interfacial energy or by competition between the
`deposition rate and the atomic surface relaxation, or by impinging atomic kinetic energy, then
`other textures may appear. Consequentially, if the film nucleation process produces island like
`growth with high aspect ratios then the sides of the islands can represent a large fraction of a
`nucleating grain surface area and the lowest surface energy { 110} planes of a bee will not be
`parallel to the substrate surface, but to the island sides. This results in a (002) texture in addition
`to the common (110) texture. Historically the most commonly sought Cr texture has been the
`(002). To a limited extent this texture can be induced by deposition at elevated temperatures and
`at high deposition rates which induce a high aspect ratio island growth [ 4]. On occasion,
`especially for very thick films, as a powder diffraction pattern might begin to appear, the (112)
`texture would even be observed. Hence, the epitaxial growth relationships between Co alloys and
`the various Cr textures have been discussed in depth [ 5 and references therein]. The more
`relevant texture and orientation relationships are summarized as:
`
`Bi-crystal:
`Quad-crystal:
`Uni-crystal:
`
`Co(l 12.0)[0001] II Cr(002)[110] or Co(l 12,0)[0001] II Cr(002)[110]
`Co(!Oll)[l210] 11 Cr(l IO)[llO] or Co(!Oll)[l210] II Cr(l 10)[110]
`Co(!OlO)[OOOI] II Cr(l 12)[110]
`
`Figure 3 illustrates these three Co textures. While the Cr (002) and Cr (112) textures induce the
`Co c-axis into the film plane the most easily formed Cr (110) texture results in the c-axis being
`inclined at ±28 degrees with respect to the surface. Hence, a lower coercivity would be
`anticipated from the Co grown on the Cr (110) texture as the c-axis is not parallel to the
`recording plane. Also we see that there are multiple directions that the Co c-axis can be placed
`upon the Cr (002) and the Cr (110) textures. Hence, upon a single (002) textured Cr grain two
`possible c-axis orientations can grow (bi-crystal) while upon a single (110) textured Cr grain four
`
`187
`
`TDK Corporation Exhibit 1013 Page 7
`
`

`
`Cr(110)
`
`Cr(11 O)/Ag(111)
`
`Ag(111 )/Si(111)
`
`Figure 5. TEM diffraction patterns for Cr(llO) and Ag(lll) epitaxially grown on Si(lll).
`possible c-axis orientations of Cr can grow (quad-crystal). These bi-crystals and quad-crystals
`can never have both c-axes parallel to the applied field simultaneously. Hence, by the Stoner(cid:173)
`Wolhfarth model one would anticipate the coercivity to be compromised. Likewise, when two or
`more Co variants do appear on a single Cr grain it is less likely that they will be isolated from one
`another by grain boundary diffusion as when compared to two Co grains located on two separate
`Cr grains. Hence, the boundary between the two variants can be thought of as a crystalline defect
`at which the magnetic spin orientation must be twisted and this provides for wall nucleation or an
`incoherent spin rotation center to initiate the switching process. This also compromises the
`maximum achievable particle anisotropy and hence, coercivity. Even worse, the four possible
`directions of the Co(lOll )//Cr(l 10) texture relation places the Co c-axes ±28 degrees from the
`film plane and this in combination with the perpendicular demagnetization field always lowers the
`coercivity to an in-plane field. On the other hand, in the absence of severe Co alloy compositional
`flaws, the uni-crystal Co(1010)//Cr(l 12) texture relationship only allows a single orientation upon
`a Cr grain and conceivably could result in a higher coercivity if aligned to the recording field. In
`addition, the surface atomic spacings (a/sqrt(3) = 2.50; sqrt(2) x a= 4.07A) of this Cr texture
`closely matches both the c (-4.07A) and a (-2.50A) axes lattice spacing of the Co alloy
`simultaneously while the other Cr textures only match well in one of the two directions. Since
`only an uni-crystal can grow on an individual underlayer grain the anisotropy energy is not
`compromised by multiple growth variants. Unfortunately, this Cr texture is seldom seen, as
`processing at low temperatures or with an applied substrate bias results in Cr (110) texture while
`high temperatures and deposition rates can partially induce the Cr (002) texture. It is believed
`that the (112) texture of the bee derivative, NiAI, provides a high coercivity template while the
`strong intermetallic Ni-Al bonding provids a small uniform grain size template [6,7].
`In addition, it should be mentioned that extended Co alloy stacking faults, caused by
`compositional inhomogeneity or epitaxial lattice mis-match, could appear as small regions of the
`Co fee phase. This cubic phase has a considerably lower anisotropy than the hep phase and since
`it is in intimate exchange contact with the remaining hep portion of the crystallite it may locally
`reduce the anisotropy energy and, hence, the coercivity. Processing at an elevated temperature
`and possibly with substrate bias during the Co deposition helps to provide the atomic mobility to
`minimize this crystalline disorder. The first criterion for the selection of Cr as an underlayer for
`Co was the close atomic lattice spacing match. Hence, these stacking flaws are exaggerated, or
`epitaxial growth does not even result, if the Co-alloy additives create too great a lattice mis(cid:173)
`match. Due to its large atomic size Pt solutes significantly increase the Co lattice constant and to
`
`188
`
`TDK Corporation Exhibit 1013 Page 8
`
`

`
`1.2
`
`Easy Axis
`Direction
`
`! 0
`"
`
`1.2
`
`Easy Axis
`Direction
`
`I
`
`0
`
`-1.2 ~~~~~~~~~~~~
`-1.510•
`0
`
`-1.2 ~~~~~~~~~~~~
`-1.s10•
`1.s10•
`
`Applied Field (Oe)
`
`Applied Field {Oe)
`
`-1.2 ~~~~~~~~~~~~
`-1.510 4
`
`Applied Field (Ce)
`CoCr20Pt,:ITi/glass
`
`-1.2 ~~~~~~~~~~~~
`1.s10•
`-1.s10•
`0
`Applied Fleld (Oe)
`Cocr .. Pt,:/Ti/Ag/Si(111)
`
`Figure 6. Comparison of easy and hard axes hystersis loops for CoCrPt (0002) films.
`
`correct for this V, Ti, and Mo have been alloyed into Cr to expand its lattice constant
`appropriately.
`
`HIGHLY ORIENTED MAGNETIC THIN FILMS
`
`From the previous discussions, we see that it is desirable to grow Co-alloy and underlayer
`crystalline grains of considerable perfection and appropriate texture to avoid compromising the
`crystalline anisotropy. Likewise, to maximize the coercivity and to minimize the media noise it is
`desirable to control the easy axes orientations. By utilizing single crystal Si as a substrate we have
`developed a model system to approach these goals. Silicon is reasonably inexpensive and could
`conceivably be used as a media substrate. By first removing the Si02 surface layers a metallic
`epitaxial layer growth can be obtained. In particular we have found that fee metals such as Ag,
`Au, Cu, and Al can grow on various Si surface orientations with a very high degree of epitaxy.
`Furthermore, these fee metallic quasi-single crystal thin films can then easily be used to epitaxially
`grow other quasi-single crystal films of similar or differing crystalline structure. By proper choice
`of lattice constants various textures can be achieved with limited induced lattice strain. Multiple
`layers allow a transition from what would be thought of as an impossible lattice strain situation to
`one of little strain, a high degree of texture and orientation results as determined by the substrate
`[8]. As examples consider a few of the texture relationships we have obtained:
`
`189
`
`TDK Corporation Exhibit 1013 Page 9
`
`

`
`Bi-crystal: Co(l tlO)/Cr(l 00)/ Ag(l 00)/Si(l 00)
`
`Quad-crystal: Co(lO!l)/Cr(llO)/Ag(lll)/Si(lll)
`
`Perpendicular: Co(0002)ffi(0001 )/ Ag(l 11)/Si(l11)
`
`Soft cubic: NiFe(lll)/Cu(lll)/Ag(lll)/Si(lll)
`
`Uni-crystal: Co(10!0)/Cr(112)/Ag(110)/Si(110).
`
`In each of these the Ag to Si lattice match occurs at a unit cell ratio of 4 to 3, respectively. The
`very long range order of the single crystal Si is believed to promote the order of the metal
`contacting layer. The following metal layers' crystalline axes may then either align to the principle
`Si axes or be rotated through fixed angles depending upon the lattice parameter relationships. In
`the case of Co growth on the quasi-single crystal Cr thin films the orientational relationships listed
`in the earlier discussion of texture apply. The excellent epitaxial relationships between the single
`crystal substrate and the bi-crystal Co(l l~O) were described earlier[8], however, it was not clear
`from that work that the extensive array of other epitaxial relationships would exist.
`Consider Figure 4 showing one of the three orientational relationships of bee Cr(! I 0) on
`fee Ag(I 11) used to promote the quad-crystal structure. It is worth noting that these are the
`close packed lattice planes and since the atomic spacings of the Cr and the Ag match reasonably
`well the high energy Cr(l 11) would not be anticipated to grow. Since there are three possible
`orientations for the Cr(! IO) on the Ag(! I I) surface and since there are four possible orientations
`(quad-crystal) for the Co(IOll) growth on Cr(I 10) there are actually twelve possible Co
`orientations on the Si substrate. This represents enough orientations that it is tempting to
`advocate that a quad-crystal rotating disk made from a Si(ll I) substrate would have little
`rotational signal modulation. However, this would be far from the uni-orientational structure
`NiFe SOnm/Cu 100nm/Ag 100nm/Si(111)
`
`.......
`.....
`.....
`~ u;
`
`.......
`.....
`.....
`.....
`Ci
`c(
`
`1.0x105
`
`J!J c
`~
`0
`0
`
`.......
`.....
`.....
`.....
`'5'
`0
`
`.......
`.....
`.....
`.....
`Qi
`LI.
`z
`
`0
`
`20
`
`30
`
`40
`
`N'
`N
`N
`Ci
`c(
`
`A
`
`60
`
`70
`
`80
`
`90
`
`50
`
`29
`
`Figure 7. X-ray diffraction scans ofNiFe(lll)/Cu(lll)/Ag(lll)/Si(lll).
`
`190
`
`TDK Corporation Exhibit 1013 Page 10
`
`

`
`Si (110)
`
`5.43 A
`
`®
`
`[001) t
`T ®
`1
`I---- 7.68A --l
`
`Ag(llO)
`
`[001) '
`
`4.09A
`
`T
`J_
`
`r--5.78A -l
`
`Cr (112)
`
`[lfO]