Buschow K.H.J. (Ed.) Concise Encyclopedia of Magnetic and Superconducting Materials

Подождите немного. Документ загружается.

where r

is the resistivity at y, r is the resistivity at

y ¼90

, and Dr is the resistivity difference between

y ¼0

and y ¼90

. Dr/r is the anisotropic magneto-

resistive ratio of the MRE. Usually a transverse

magnetic ﬁeld is applied to the MRE so that y ¼y

when the head is not reading a magnetic signal.

Assume that a magnetic signal from the medium

rotates M from y

to y, then an electric voltage

change v will be detected across the MRE as the

output signal:

v ¼ Jwr

ðsin

 sin

yÞð4Þ

where y

is the bias angle and is set to 451 for good

linearity. Equations (3) and (4) tell us that while

inductive head output is proportional to medium

velocity and not suitable for low-velocity applica-

tions, the MR head can be used for either high- or

low-velocity applications.

The second generation of MR head introduced re-

cently is the Giant Magnetoresistive (GMR) head

(see Giant Magnetoresistance). The most commonly

used implementation of the GMR head is the spin

valve conﬁguration. As depicted in Fig. 6, it has four

basic layers. The antiferromagnetic material (AFM)

provides an exchange ﬁeld to pin the magnetization

of the adjacent (pinned) layer to the direction 901

from the medium surface. A nonmagnetic spacer

(typically copper) separates the magnetic free layer

from the pinned layer. When there is no magnetic

ﬁeld from the medium, the magnetization of the free

layer parallels with the medium surface.

When a magnetic ﬁeld is applied from the medium,

the magnetization of the free layer rotates an angle y

from its original direction. y is a function of the

strength and direction of the magnetic ﬁeld. Similar

to the AMR head, the resistivity of the MRE changes

with magnetic ﬁeld, yielding

v ¼

JwDrsiny ð5Þ

Compared with the AMR head, the GMR head has

higher sensitivity, better linearity, and needs no bias.

With some modiﬁcations and enhancements, the

GMR head may be able to achieve an areal density

around 100 Mbits mm

2

. Engineers and scientists in

the magnetic recording industry and universities are

exploring new reproducing heads to extend the areal

density beyond that limit. A promising technique is

the tunneling magnetoresistive (TMR) head. It is

composed of two ferromagnetic layers separated by a

thin insulating layer. The electric current ﬂows in the

direction perpendicular to the layers, in contrast with

the GMR head where the current ﬂows in the direc-

tion parallel to the layers. The TMR head promises

to deliver higher linear density (smaller bit length B)

and higher signal amplitude, but has the disadvan-

tage of very high impedance.

2. Recording and Reproducing Process

Due to the nonlinear nature of the hysteresis loop,

the recording process is complex and we need to

adopt some approximation to simplify the analysis. If

we set the origin of a coordinate system at the center

of the gap with the x-axis on the head surface and

the y-axis pointing away from the head, then the

longitudinal magnetic ﬁeld H

of this head can be

expressed by the Karlqvist approximation:

pðg þ lA

=mA

tan

1

þ y

ðg

=4Þ



ð6Þ

where n is the number of coil turns, i is the current in

the coil, g is the gap length, l is the core length, A

the core cross-sectional area, m is the relative perme-

ability of core material, and A

is the gap cross-sec-

tional area. Equation (6) shows that the contours of

constant H

ﬁeld are circles nesting on the two gap

corners, as shown in Fig. 7. The greater the diameter

of the circle, the weaker the magnetic ﬁeld.

Figure 5

Conceptual diagram of an AMR head.

Figure 6

Conceptual diagram of a GMR head.

630

Magnetic Reco rding Technologies: Overview

Assume a magnetic medium, moving from left to

right with a distance d above the head, has a thickness

d and a magnetization M pointing to the right. At

some instant the recording current turns on and gen-

erates the magnetic ﬁeld H

above the gap as depicted

in Fig. 7. On the circumference of H

¼H

, half of

the medium material has its magnetization reversed

and half remains the same, resulting in zero total

magnetization. Since H

has a gradient, the medium

closer to the gap (inside a smaller circle) gets its

magnetization reversed more completely than the

medium further away from the gap (outside a bigger

circle). Therefore, magnetic transition is gradual in

the medium even if the change of recording current

follows a step function.

Assume the original magnetization is M and the

completely reversed magnetization is M

, this grad-

ual change of magnetization for an isolated transition

can be modeled by:

M ¼

tan

1

ð7Þ

where x is the distance from the center of transition,

and a is a parameter characterizing the sharpness of

the transition as shown in Fig. 8. For a thin-ﬁlm me-

dium, a is found to be

a ¼

ð1  S



Þðd þ d=2Þ

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

ð1  S



Þðd þ d=2Þ



dðd þ d=2Þ

pQH

ð8Þ

where S* is the medium loop squareness, and Q is the

head-ﬁeld gradient factor. For a reasonably well-

designed head, QE0.8. It is obvious that we want to

make parameter a as small as possible so that we can

record more transitions for a unit medium length. If

the head gap length g and medium thickness d are

small compared with head/medium separation d, and

the medium has a squareness of one, then the min-

imum possible value of a is:

2pH

for

4pH

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

for

4pH

ð9Þ

In order to decrease the value of a and therefore

increase areal density, we need to reduce the medium

remanence M

, thickness d, head/medium separation

d, and to increase the coercivity H

In contrast to the recording process, the reproduc-

ing process is well understood. The ﬂux density in-

duced in the head core is small, yielding a linear

process easier for mathematical treatment. The goals

for higher areal density in magnetic recording are to

reduce the pulse width (PW

) and to increase the

reproducing amplitude. The effect of various head

parameters on pulse shape can be comprehended by

studying the head reproducing voltage over an iso-

lated transition (Fig. 8) with a thin magnetic layer.

For an inductive head the output voltage is:

vðxÞ¼

2m

VwM

pðg þ lA

=mA

 tan

1

g=2 þ x

a þ d



þ tan

1

g=2  x

a þ d



ð10Þ

where x is the distance between the center of the head

gap and the center of the transition, m is the relative

permeability of the core, n is the number of coil turns,

g is the head gap length, d is the head/medium sep-

aration, and x is the distance between the center of

the medium transition and the center of the head gap.

The term lA

/mA

is closely related to g for head

efﬁciency. When an MR head is used for reproducing,

Figure 7

The contour of a longitudinal magnetic ﬁeld from the

Karlqvist approximation.

Figure 8

An isolated arctangent magnetization transition from

negative M

to positive M

631

Magnetic Reco rding Technologies: Overview

the sensor is usually sandwiched between two mag-

netic shields to increase its spatial resolution to me-

dium signals. For an AMR head with small gap and

bias angle, it becomes:

vðxÞ¼

2O2EJwDrM

ptM

g þ t

 tan

1

g=2 þ x

a þ d



þ tan

1

g=2  x

a þ d



ð11Þ

where E is the head efﬁciency, J is the current density,

t is the AMR ﬁlm thickness, g is the gap length (the

distance between the MRE and the shield), and M

the saturation magnetization of the MR ﬁlm. For a

GMR head, the output is

vðxÞ¼

EJwDrM

ptM

g þ t

 tan

1

g=2 þ x

a þ d



þ tan

1

g=2  x

a þ d



ð12Þ

It is clear from Eqns. (10) and (12) that the pulse

shape versus x is similar for an inductive, an AMR,

and a GMR head. Since the MR ﬁlm thickness t is

typically one order of magnitude smaller than the gap

length g, the term (g þt)/g is approximately a con-

stant. This indicates that g, a,andd affect pulse shape

similarly for all three types of head, provided half of

the gap length needs to be used for an MR head while

the full gap length has to be used for an inductive

head. In fact, the PW

can be derived as

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

þ 4ða þ dÞ

ð13Þ

for all three types of heads. To increase linear density,

we need to reduce gap length g, transition parameter

a, and head-to-medium separation d.

The amplitude of a reproducing pulse is adversely

affected by adjacent transitions because of the oppo-

site polarity of the adjacent magnetizations. At high

linear density, the magnetization variation in the

medium approaches a sinusoidal wave,

MðxÞ¼M

sin

x ð14Þ

where l is the wavelength. The reproducing voltage in

an inductive head becomes:

vðxÞ¼

m

VwM

g þ lA

=mA

ðe



ð1  e



sin

pg=l

cos

x ð15Þ

This equation presents all the important features of

the high linear density reproducing process. The

term exp(2pd/l) is the spacing loss. It shows that

the reproducing voltage falls exponentially with the

ratio of head/medium spacing to wavelength. The

second term 1exp(2pd/l) is the thickness loss.

The name of this term is misleading because its

value increases with a greater medium thickness.

However, the rate of increase diminishes for a thicker

medium. In fact, 80% of the maximum possible value

is achieved by a medium thickness of 0.25 l. The last

term sin(pg/l)/(pg/l) is the gap loss. This term is

based on the Karlqvist approximation. If a more ac-

curate head fringe ﬁeld is used, this term is modiﬁed

to sin(pg/l)/(pg/l) (1.25 g

l

)/(g

l

). It shows a

gap null at l ¼1.12 g, and limits the shortest wave-

length producible. These three terms are plotted in

Fig. 9. The most significant loss comes from the

spacing loss term, which is 54.6 dB for d ¼l. There-

fore, one of the biggest efforts spent on magnetic re-

cording is to reduce the head/medium spacing as

much as possible without causing mechanical relia-

bility issues. The effects are similar for an MR head.

3. The Challenge for High Areal Density

Since areal density is the product of linear density

and track density, the increase in areal density trans-

lates into the reductions in pulse width and track

width. Both magnetic media and heads play impor-

tant roles in these areas. One of the most prominent

parameters affecting pulse width is the transition pa-

rameter a. Equations (7) to (12) indicate that it affects

both the recorded transition length and reproducing

pulse width. From the medium viewpoint, we can

reduce a by increasing coercivity H

, reducing M

d,or

reducing S*. Since 1990, H

has been increased from

around 1000 Oe to 3800 Oe. The primary limitation

to higher H

is the recording head. Presently it is

Figure 9

The spacing loss, thickness loss, and gap loss for high

linear-density recording.

632

Magnetic Reco rding Technologies: Overview

difﬁcult for a head to generate a magnitude ﬁeld with

the required gradient for recording on a medium with

over 4000 Oe, especially at high frequency.

This is described by Sharrock’s law:

ðtÞ¼H

1 

0:693





ð16Þ

where t is time, T is absolute temperature, k is the

Boltzman’s constant, K

is the uniaxial anisotropy

energy per unit volume, v is the grain volume of the

magnetic particle, and f

is the Larmor frequency. If

we reduce M

d, we will sacriﬁce amplitude. In addi-

tion, it is more difﬁcult to produce a uniform thin

ﬁlm than a thick ﬁlm.

Magnetic medium noise has three components:

transition noise, particulate noise, and modulation

noise. While both particulate noise and modulation

noise are independent of or decrease with linear den-

sity, transition noise generally increases with linear

density. Consequently, reduction of transition noise

becomes necessary for high areal density application.

The signal-to-transition-noise ratio is

SNRE

2pD

ð17Þ

where D

is the channel density, and D is the grain

diameter.

Obviously we need to reduce the grain size to

achieve adequate SNR if we reduce B or w to increase

areal density. A major obstacle to the reduction of

grain size is the superparamagnetic limit, which in-

dicates the thermal instability of magnetization if the

grain size is smaller than a critical volume v

¼ lnð2tf

2kT

ð18Þ

where H

is the total anisotropy ﬁeld. First-order

calculation shows that a grain diameter of 10–12 nm

is required for ten-year data retention.

Although we can increase either track density or

linear density to increase areal density, it is more ad-

vantageous for us to increase track density from the

balance of SNR and thermal stability. This is because

higher linear density requires a thinner magnetic lay-

er, resulting in shorter grains with smaller volume

and lower thermal stability, whereas a higher track

density allows us to keep the same magnetic layer

thickness, leading to higher areal density for the same

thermal stability. Consequently, the aspect ratio

(w/B) needs to reach 4 for 100 Gb in

2

, which is

presently around 8 in the 50 Gb in

2

demonstration

and was about 20 at 100 Mb in

2

When two transitions are recorded very close, the

demagnetization ﬁeld from one transition affects the

transition position of the other. This shift in transi-

tion position can cause difﬁculty in channel detection,

and is termed nonlinear transition shift (NLTS). As

linear density gets higher and higher, NLTS will

become more of a problem.

From the recording head viewpoint, the challenge

lies in writability and high frequency response. Writ-

ability is usually expressed as overwrite, which is a

measurement of a head’s capability to write over a

preexisting signal on the medium. With the increase

in medium coercivity, the recording ﬁeld needs to be

increased accordingly. However, the magnetic mate-

rial at the corners of the recording head gap may get

saturated before the required fringe ﬁeld is achieved.

This saturation will deteriorate the ﬁeld gradient,

leading to poor recording. To prevent this from hap-

pening, a high Bs layer magnetic material is deposited

on the inside surfaces of the gap, similar to the MIG

head in Fig. 2. The challenge is to ﬁnd a material that

has both high B values and high corrosion resistance.

Commonly used materials are iron and cobalt alloys.

We have implied in our analysis that the recording

ﬁeld responds instantaneously to recording voltage

pulse. In other words, the recording ﬁeld has perfect

square waves in time domain. When we deal with low

frequency, this assumption is valid. However, when

the frequency is high, it is not valid because the head

has both an inductance and the core material needs

ﬁnite time to rotate its magnetization. If the ﬁeld rise

time t is too large compared with transition time T,

then additional NLTS will be introduced. t is typi-

cally between 1 and 5 ns, indicating frequency limi-

tation occurs around 1 Gb s

1

. To reduce t, we need

short and thin magnetic yokes with high B and high

electric resistance.

From the reproducing head viewpoint, the major

amount of research and development emphasis has

been placed in the areas of signal sensitivity and sta-

bility. With narrower track width and shorter ﬂux

change length for higher areal density, the recording

head picks up less magnetic ﬂux from each transition.

To ensure adequate SNR, the recording head needs

to have a higher sensitivity to the magnetic ﬂux from

the recorded signal. The change from inductive head

to AMR head then to GMR head and maybe to

TMR head is to achieve high SNR.

The reproduced signal must be stable for the chan-

nel to detect. This requires a stable magnetic domain

in the MR ﬁlm. A hard magnetic material (hard bias)

is deposited adjacent to the leads along the MRE

width direction to stabilize the magnetic domain and

reduce large Barkhausen noise. If the strength of the

hard bias ﬁeld is not enough, the reproduced signal

may experience jumps. This phenomenon is called

baseline shift. However, if the ﬁeld is too great, the

head will suffer poor sensitivity. The head stability

can also be a result of the change of the pinned mag-

netization, and the loss of exchange ﬁeld of the AFM.

To address these concerns, synthetic antiferro-

magnetic (SAF) pinned layers are used. The SAF

layer, consisting of two ferromagnetic layers sepa-

rated by a thin ruthenium layer, provides a more

633

Magnetic Reco rding Technologies: Overview

stable pinned magnetic ﬁeld, as depicted in Fig. 10.

Various AFM materials have been employed to in-

crease its block temperature and pinning ﬁeld. In ad-

dition to sensitivity and stability, manufacturability

and durability are important. Practical issues such as

dimension control, ESD safety, and corrosion resist-

ance all need to be resolved.

Both Eqn. (13) and Eqn. (15) show the importance

of head-to-medium spacing d. To reduce PW

and

increase amplitude, we need to minimize d.Ifd is too

small, however, mechanical interactions will occur on

the head and/or the medium, leading to severe dam-

age. The science and art for the balance between areal

density and head/medium interface durability be-

longs to the ﬁeld of tribology. For low duty-cycle

applications such as in a ﬂoppy drive, the head is in

contact with the medium surface. For high duty-cycle

applications such as in a hard drive, the head ﬂies

over the disk surface during operation. The ﬂying

height is between 10 and 20 nm. This small and stable

ﬂying height is achieved by using an air-bearing

surface on the head, a ﬂat and smooth surface on

the disk, and a pressurized air ﬁlm in between. The

scientific and engineering foundation is the hydro-

dynamic lubrication, governed by the Reynolds

equation.

See also: Magnetic Recording Devices: Head/Medi-

um Interface; Magnetic Recording: Rigid Media,

Tribology; Magnetic Recording: Rigid Media,

Recording Properties

Bibliography

Bertram H N 1994 Theory of Magnetic Recording. Cambridge

University Press, Cambridge

Fan G J 1961 A study of the playback process of a magnetic

ring head. IBM J. Res. Dev. 5, 321–5

Karlqvist O 1954 Calculation of the magnetic ﬁeld in the ferro-

magnetic layer of a magnetic drum. Trans. R. Inst. Technol.

(Stockholm) 86, 3–28

Maller V A J, Middleton B K 1973 A simpliﬁed model of the

writing process in saturation magnetic recording. IERE Conf.

Proc. 26, 137–47

Mallinson J C 1993 The Foundations of Magnetic Recording.

Academic Press, San Diego, CA 2nd edn

Mee C D, Daniel E D (eds.) 1996 Magnetic Storage Handbook

(2nd edn). McGraw-Hill, New York

Miyata J J, Tartel R R 1959 The recording and reproduction of

signals on magnetic medium using saturation-type recording.

IRE Trans. Elec. Comp EC-8, 159–69

Potter R 1974 Digital magnetic recording theory. IEEE Trans.

Magn MAG-10, 502–8

Wang S, Taratorin A M 1999 Magnetic Information Storage

Technology. Academic Press, New York

Watkinson J 1994 The Art of Data Recording. Focal Press,

Oxford

Webster J G (ed.) 1999 The Measurement, Instrumentation and

Sensors Handbook. CRC Press, Boca Raton, FL

Westmijze W K 1953 Studies on magnetic recording. Philips

Res. Rep. 8, 161–83

Williams M L, Comstock R L 1972 An analytical model of the

write process in digital magnetic recording. AIP Conf. Proc.

5, 38–42

Y. Li and A. K. Menon

Read-Rite Corporation, Fremont, California, USA

Magnetic Recording: Flexible Media,

Tribology

The data densities demanded of magnetic media have

increased monotonically and much work has been

done to respond to these demands. The tribology of

these systems has, however, been largely neglected

and problems which occurred were solved largely by

a combination of good design, lower contact pres-

sures, better lubricants, better materials for heads and

media, and by mature use of natural tribological

laws. Thus by a combination of good system design,

Figure 10

Sectional views of various spin-valve structures:

(a) standard, (b) bottom synthetic, and (c) dual

synthetic.

634

Magnetic Reco rding: Flexible Media, Tribology

exploitation of materials technology, and good for-

tune, the tribology of the systems has, until recently,

seldom been a limiting factor in their development.

Good fortune can no longer be relied upon and there

is an urgent need for more effort to be made in un-

derstanding the basic mechanisms of contact in these

very complex systems.

As recording densities increase, recording wave-

length is reduced and the trend is to thinner magnetic

coatings, two or more layered structures, thinner

polymer substrates, zero head/media spacing, and

smaller head dimensions. Magneto-resistive (MR)

read head elements are employed which may run at

temperatures of 70 1C or more. In future, giant mag-

neto-resistive (GMR) heads will probably be used

which will necessitate the use of ultrathin head coat-

ings. In particulate media this necessitates the use of

magnetic layers of different structure, with different

resin binders with higher glass transition tempera-

tures and new, more effective lubricants and lubrica-

tion systems. In addition, magnetic particle loadings

must increase and particle size must be reduced. All

of these factors produce acute tribological contact

problems and are now significant limiting factors in

the development of dynamic magnetic storage tech-

nology, and will be so increasingly in the future.

With the miniaturization of components and the

consequential changes detailed above, any factor

which adversely affects the intensity of the output

recorded signal or which produces recording errors

becomes of prime importance. For example, any

process or mechanism which introduces spacing be-

tween the media and the recording head or which

leads to the physical removal of material from the

magnetic media must be understood and its effects

eliminated. The signal falls exponentially with an in-

crease in this spacing.

The materials aspects of the problem will be dis-

cussed later, but it is worth considering some basic

tribological laws which have aided the magnetic me-

dia cause. A simple law describing the wear of ma-

terials under a wide variety of conditions is the

Archard wear law:

o ¼ kW=P

ð1Þ

where o is the volume of material removed per

unit distance of sliding, the wear rate; W is the

applied load; and P

is the material ﬂow pressure, or

hardness.

From this simple law one can see that increasing

head and media hardness will decrease wear, as will

reduction in head loads. Here, however, the industry

is further helped by nature, in that k does not remain

constant, but can fall by almost an order of magni-

tude for a ceramic head vs. a g-Fe

media at low

load. However, the friction coefﬁcient increases at

low pressures, particularly with very smooth surfaces.

As a consequence of this effect, dynamic and static

friction problems arise under low-load conditions,

which are not noticed at higher loads.

1. The Media

Flexible magnetic media consist of a thin magnetic

coating applied to a polymeric substrate carrier ﬁlm.

The magnetic coating may be applied as a slurry

containing magnetic particles (particulate media) or

as an evaporated magnetic metal ﬁlm (metal evapo-

rated media). In addition to the magnetic side of the

media there is often a backside coating applied to

tape to give desired mechanical and tribological

properties, to prevent adhesion when rolled, and to

give antistatic protection. Traditionally, the base coat

has been produced from polyethylene teraphthalate,

but for the very thin substrates used in the continued

drive towards miniaturization, materials with higher

elastic modulus such as polyaramide and polyethyl-

ene naphthalate are required.

1.1 Particulate Media

In particulate media, the magnetic coating consists

essentially of acicular magnetic particles bound to-

gether by means of a polymeric resin. Thus the mag-

netic layer in contact with the read/write heads will

have the tribological properties of a ﬁlled polymer.

This consists not only of the magnetic pigment (at up

to 80% by weight, 50% by volume) and the polymer

binder, but also a head-cleaning agent (HCA) (typ-

ically 4–8% by weight of a relatively large abrasive

particle such as Al or Zr oxide) and a possible com-

bination of wetting agents, cross-linking agents, sol-

vents, antifungal agents, carbon black (to prevent

surface charging), and the essential lubricant (typi-

cally fatty acid/fatty acid esters). Until recently, oxide

magnetic pigment (Co doped g-Fe

) was used in

ﬂexible recording media. The necessity for higher co-

ercivity and squareness has meant that current de-

velopment is concentrated on passivated metal (Fe)

acicular particles (MP media) with a secondary in-

terest in barium ferrite hexagonal platelets.

Magnetic constraints demand that particle sizes and

layer thicknesses are continually decreasing with par-

ticles now being produced of about 70 nm in length

(4:1 acicular ratio) with projected layer thicknesses of

about 10 particles. The mechanical integrity of such a

layer would not provide sufﬁcient durability and

hence tapes must be produced where the magnetic

layer is formed on a thicker nonmagnetic undercoat.

The most popular current method of providing lubri-

cation in particulate media is to incorporate the lu-

bricant in the magnetic layer. However, as layers

become thinner the volume of lubricant may not be

sufﬁcient to provide the required durability over the

projected lifetime of the media. Hence lubricant is in-

corporated into the undercoat and reaches the surface

through pores in the magnetic layer.

635

Magnetic Reco rding: Flexible Media, Tribology

Considering the tribology of this system, the most

important part of the contact layer is the binder

which must have the ability to accept high magnetic

particle loading, provide good wear resistance, good

substrate and particle adhesion, low surface energy

(low friction), be resistant to temperature and hu-

midity effects, and exhibit sufﬁcient ﬂexibility for the

continual deformation required of a tape or ﬂoppy

disk. To satisfy these requirements the polymeric

binders consist of at least two phases, a ‘‘hard’’ glassy

phase to give the system mechanical integrity and a

‘‘soft’’ rubbery phase to provide the ﬂexibility re-

quired. The blend is cross-linked to increase phase

separation and to ensure that the glass transition

temperature is below the operating temperature of

the media. Many present polymer blends use a dis-

persing or wetting agent to ensure polymer–particle

adhesion and homogeneous dispersion, but dispers-

ing binders are being developed with modiﬁed func-

tional end groups. This results in better binder

particle bonding, higher cohesive strength, higher

particle loading, and better wear properties. One dis-

advantage may be that such binders are more polar

and this may increase the adhesive forces between

heads and media.

1.2 Metal Evaporated Media

In order to satisfy high areal recording density re-

quirements, thin metal ﬁlms such as Ni–Co alloys

evaporated under vacuum onto a ﬂexible polymer

substrate were introduced. This media has potentially

more effective magnetization than particulate media

due to the increased grain density. The media consists

of single, or possibly, multilayers of evaporated metal

magnetic layers on a polymer substrate, covered by a

protective carbon overcoat which is in turn topically

lubricated. Tribologically the system is inherently sim-

pler than the particulate media, but presents its own

unique problems. Major among these is that, unlike

particulate media, surface coverage of the lubricant

must be complete, must remain so for the projected

lifetime of the media, must afford complete corrosion

protection, and should not be more than about 5 nm

thick in order that spacing losses are not excessive.

Metal evaporated (ME) thin ﬁlm ﬂexible media has the

advantages of the porosity and surface chemistry of the

particulate media, thus the demands on the lubricant

present a severe challenge. The lubricants are the per-

ﬂuoropolyethers, often with end group modiﬁcations.

It should also be borne in mind that these thin metal

ﬁlms are far less compliant than particulate-ﬁlled poly-

mer ﬁlms and should not be subject to excessive strain

if cracking and catastrophic failure are to be avoided.

2. Contact and Forces at Head–Media Interfaces

When two surfaces interact, the real area of contact is

dependent on the physical properties of the surface,

but is generally lower than the apparent area of con-

tact. This is because contact will only occur at as-

perity tips. When contact occurs, these asperities may

deform either plastically or elastically and it is im-

portant to know the mode of deformation before

wear and friction can be understood. Bhushan (1984)

adapted the work of Greenwood and Williamson

(1966) and deﬁned a plasticity index c for such that:

c ¼



1=2

ð2Þ

where E* is the complex modulus of elasticity, Y is

the yield strength or hardness, s is the standard de-

viation of asperity heights, and r

is the mean radius

of those asperity peaks. He suggested that for poly-

mers if co1.8 the contact is elastic, and if c42.6 the

contact is plastic. Consideration of particulate media

reveals that most contacts are elastic. From the above

relationship, therefore, the real area of contact, A

can be calculated for elastic contact from:

E3:2

ð3Þ

The classical analysis of Halliday (1955) also shows

that the contact must be almost entirely elastic for

particulate media and elastic/plastic for metal evap-

orated media.

Thus for particulate media, the forces generated

when contact occurs are largely adhesive and the real

areas of contact are governed by the elastic properties

of the media. For metal evaporated media the contact

is more complex, but, considering the elastic/plastic

combination, adhesive forces probably still dominate

contact.

3. Surface and Contact Temperatures

Surface and contact temperatures can have a sub-

stantial effect on the nature and magnitude of wear

and friction processes. There have been many at-

tempts to measure and calculate surface tempera-

tures. Bhushan (1987) reported a detailed thermal

analysis in which he predicted contact temperature

rises for typical head–tape contacts of 7–10 1C,

although he concedes that if magnetic particles are

exposed the contact temperatures can be very much

higher at 600–900 1C. There is some correlation of

these results with IR measurements made on a sim-

ulated head/tape interface by Gulino et al . (1987).

Our calculations, however, suggest that contact or

hot spot temperatures much higher than Bhushan’s

ﬁgures can occur in practice.

In this model the heat generated due to friction is

equated to the sum of the heat lost due to conduction

into the head surface, conduction into neighboring

636

Magnetic Reco rding: Flexible Media, Tribology

asperities, and that due to forced convection to the

air. Taking atomic force micrographs of the surface as

a basis for a realistic model, the surface was modeled

into a network of nodes, each node representing an

asperity. It was assumed that the asperities were sub-

ject to random Raleigh distribution of volumes and

base radii. On the basis of this model, maximum con-

tact temperatures for a typical linear tape system were

calculated to be over 100 1C, although the mean sur-

face temperature was low. If it is assumed that all the

load is carried by one asperity, and we have evidence

to show that this may occasionally occur, then very

high temperatures are predicted.

4. Friction

Friction is an important parameter in the design of

magnetic recording systems. With the continual move

towards miniaturization and the use of low power

drives, friction which is too high can cause speed var-

iations, high interfacial temperatures, or, ultimately,

failure of drive motors. In tape systems, it is not only

head–media friction which must be considered, but

the tape has to negotiate a complex transport system

and backcoat friction can lead to problems. Dynamic

friction must therefore be understood and controlled,

but equally as important is the static or ‘‘start-up’’

friction.

4.1 Dynamic Friction

The frictional force, F, consists of an adhesive com-

ponent, F

, plus a deformation component, F

,which

in turn consists of a plastic deformation or ploughing

term and a hysteresis loss term, hence, F ¼F

þF

Considering the viscoelastic nature of the magnetic

layer, the nature of the contact, and the surface to-

pography, then, except at very low speeds and at high

temperatures, the deformation term in the friction

equation is negligible. Hence friction may be regarded

to a ﬁrst approximation as being due to adhesion. All

the frictional energy is dissipated in the contact region

and the frictional force may be written as:

FEF

¼ tA

ð4Þ

where t is the interfacial shear stress. t is not constant,

but will increase with pressure and will, in general, fall

with increased temperature within the operating range

of the media. Friction will also be speed dependent

due to the time dependence of the elastic properties

of the polymer and the increase in temperature with

increase in velocity.

With particulate media, a fatty acid/fatty acid ester

type lubricant is normally incorporated into the bulk

of the magnetic coating. The coating thus acts as a

reservoir to replenish lubricant molecules removed

by wear. Using angle resolved x-ray photoelectron

spectroscopy (XPS) we have measured the thickness

and areal coverage of lubricants on a wide variety of

tape and diskette surfaces and found current lubri-

cants to be about two or three molecular layers thick,

with coverages as low as 10% of the total surface

area. Taking into account the presence of the lubri-

cant, the frictional force can be written as:

F ¼ A

ftð1  aÞþt

ga ð5Þ

where t

is the shear stress in the lubricant and a is

the fractional areal coverage of the lubricant. The

lubricant provides an easily sheared layer separating

the two surfaces and will also produce some plastic-

ization at the surface of the polymer.

Polymer surfaces have low surface energy and this

together with the low areal coverage means that the

frictional force will not change substantially with lu-

brication, but the presence of this lubricant is essen-

tial to reduce wear. Since the lubricant ﬁlm will not

affect the frictional force, this force must be governed

by the properties of the magnetic layer. Thus a high

hardness, high complex modulus material should be

used to reduce friction, with due regard to a suitable

glass transition temperature and toughness. Rougher

surfaces produce lower frictional forces; however,

due to the need for minimum spacing loss, surfaces

need to be as smooth as possible.

Friction increases with time of running, at least

over the ﬁrst few passes. This is to some extent due to

the surface becoming smoother, but this is not en-

tirely consistent with the assumptions that contact is

elastic. Our ﬁndings indicate that the outer layers also

change chemically. The chemical changes are con-

ﬁned to the outer polymer covering on the magnetic

particles. Assuming the contacts to be mainly adhe-

sive, the changes may be sufﬁcient to inﬂuence the

frictional force.

In the case of ME media, the lubricant layer cov-

erage is complete and hence a ¼1. This is a case of

classical boundary lubrication, where the friction

force is wholly determined by the shear stress of the

lubricant/head interface. Here, the deformation com-

ponent of friction is negligible. If lubrication fails,

however, friction forces can become catastrophically

high.

4.2 Static Friction—Stiction

In all materials there is a distinction between dynamic

and static friction, with the force required to initiate

motion, static friction, being higher than that re-

quired to maintain it, dynamic friction. In magnetic

recording systems where the head is often parked in

contact with the media, this difference can be partic-

ularly pronounced and can lead to acute start-up

problems. This results in a jerky and uneven motion

of the tape or disk. There are a number of reasons for

this high initial friction. As mentioned before the

637

Magnetic Reco rding: Flexible Media, Tribology

ﬁlled polymer magnetic layer is in a viscoelastic state

at normal operating temperatures. In this state the

strain is rate dependent. This means that the real area

of contact increases with time and increasing static

friction. In addition to this, van der Waals forces

operate between the smooth surfaces of the magnetic

layer and the head and these will increase the sur-

face area above that predicted by theory. There is

some controversy over the importance of electrostatic

forces, but it is likely that they account for increases

in friction at very low loads.

Very high static friction forces are most apparent

under conditions of high humidity. This may be due

to a meniscus effect at the head–media interface or

the effect of water plasticization of the polymer in the

contact zone. Meniscus effects are possible since high

friction is also observed if too much lubricant is

present; it is more likely, however, that increased

moisture levels instigate hydrolysis and plasticization

takes place. To overcome these problems similar

polymer properties of high complex modulus and

high hardness are required with an additional

requirement for the binder to be hydrophobic.

5. Wear at the Head–M edia Interface

Although the ultimate life of head and tape are of

importance and determined by absolute wear, the

actual wear rates are, in engineering terms, very small

indeed and in general it is the effect of wear on signal

performance that has a greater inﬂuence on useful life

than material removal rates. Any effect which intro-

duces a spacing between the magnetic media and the

read/write sensors or which results in loss of data is

of prime importance. Thus media degradation and

wear, head wear, differential wear, head surface

modiﬁcations, and transfer ﬁlm formation must all

be considered. Wear is classiﬁed as adhesive, fatigue,

abrasive, corrosive, or erosive. Although it is clear

that the contact is largely adhesive in head–media

contact, it is probable that most of these forms of

wear apply at some time.

5.1 Particulate Media

(a) Transfer ﬁlm—head stain

The ﬁrst stage of wear usually consists of material

transfer from media to head. Our x-ray spectroscopic

measurements show that at the surface the magnetic

particles are covered with polymeric resin to a depth

of about 5 nm. Thus the initial head–media contact is

between head and the polymer surface of the media.

The adhesive forces formed at this contact may be

either physical, involving van der Waal bonds, or

chemical, involving electrostatic, ionic, or covalent

bonding. Typical van der Waal bond strengths indi-

cate that their contribution is not significant. The

chemical bonds are probably more significant and

these will depend on the nature of the head material.

For polymer/metal contacts covalent bonding occurs

and for polymer/ceramic contact, ion-dipole bonds

are formed. Each of these bond types will give strong

adhesive junctions.

Within the polymer binder, atoms are covalently

bound (bond strengths up to about 10 eVatom

1

) and

chains bound together via dipole–dipole interactions.

Dipole–dipole interactions are typically a few tenths

of an eVatom

1

, although interchain bond strengths

approach those of the covalent bonds due to cross-

linking. Bond strengths of the adhesive junctions

formed at the various polymer/head interfaces can be

greater than those of the binder/particle interfaces

within the magnetic layer and this, if the conditions

are right, leads to adhesive transfer from media to

head. (Transfer is not observed for relative humidities

above about 45–60%, where water saturation of ac-

tive sites on the head material surfaces prevents the

formation of adhesive junctions.) Transfer is partic-

ularly evident on MR metal elements due to high

temperature enhancement of covalent bond forma-

tion and reduction of surface saturation due to ex-

pulsion of water vapor and other contaminants. Thus

transfer ﬁlms occur readily between polymer/metal

and polymer/ceramic surfaces and these can substan-

tially affect wear properties of the junctions and may

introduce spacing losses. This phenomenon is known

as ‘‘head stain.’’

Transfer does not occur in homogeneous layers,

but as a discontinuous lumpy ﬁlm, with thicknesses

corresponding to one magnetic particle. The thick-

ness of these layers indicates that the depth of ma-

terial removed from the media in this early phase of

wear is not sufﬁcient to produce data errors in cur-

rent media.

These transfer ﬁlms have been variously described

as friction polymer, brown, blue, or white stain. In

XPS and Auger electron spectroscopic studies we

have found no friction polymers. Our XPS analysis

(careful measurement of Fe 2p binding energies) in-

dicates that the layer consists of magnetic pigment

from the media and we conclude that color is deter-

mined by thickness and substrate.

These results are entirely consistent with the mech-

anisms where the thin surface polymer is adhesively

bound to the head. This results in cohesive failure

within the media and laminae of media that are one

particle thick transferring and adhering to the head

surface. Once the ﬁlm has been formed contact is

between it and the media, although one would expect

the situation to be dynamic with continual removal

and replacement of the stain. After formation, the

junction strength between load bearing surfaces of

the transfer ﬁlm and the polymeric media is not suf-

ﬁcient to duplicate the transfer process and hence

ﬁlms will not normally increase in thickness. In tribo-

logical terms, the formation of a transfer layer is

beneﬁcial in reducing wear. However, the layer may

638

Magnetic Reco rding: Flexible Media, Tribology

introduce unacceptable spacing losses and may, un-

der certain circumstances, short out MR elements

and shields. A function of the head-cleaning agent

added to the magnetic layer is to limit the thickness of

this stain and to reduce spacing losses. (Another and

perhaps less well understood function of the head

cleaning agent is to act as a bearing surface and ef-

fectively protect the soft media against wear.) During

this phase of wear, surfaces of the particulate media

generally become smoother and errors are reduced

(Hempstock et al. 1998).

(b) Delaminative wear and head clogging

The second stage of wear of particulate media is due

to fatigue and results from continuous cyclic stressing

of the media by contact with the head. When two

solids are in sliding contact, the maximum stress oc-

curs not at the asperity junction, but below this sur-

face. When an asperity approaches an opposing

asperity the subsurface is in compression; as the as-

perities separate the subsurface is in tension. Hence

for repeated contact the subsurface immediately be-

low the asperity contacts is subject to cyclic stressing.

This leads to the initiation of subsurface fatigue

cracks which grow to the surface and produce de-

laminative removal of particles, which leave crater-

like recesses behind (Hempstock and Sullivan 1998).

This delaminative particle removal process is analo-

gous to that described by Suh (1986) for metals. For

fatigue-induced crack initiation there needs to be

defects in the matrix. Stress concentrations at these

defects will then enhance crack initiation and pro-

pagation. There are three possible candidates for this:

the boundary between hard and soft phases of the

copolymer binder, the boundary between head clean-

ing agent particles and binder, or the boundary be-

tween acicular magnetic particles or agglomerates of

these particles and the binder.

Although this wear phase is relative rare, it has

many effects on signal performance and error

growth. The removal of delaminative wear particles

from the media can result in loss of bits of informa-

tion. Hempstock and Sullivan (1997) have observed a

correlation between such surface damage and in-

crease in dropouts (hard errors). The released debris

then forms third body particles which may damage

head and media and lead to head clogging. Head

clogging is due to loose debris expelled from the

contact area agglomerating on the exit side of

the head. The agglomerates are then picked up by

the tape and eventually transferred to the head where

they accumulate in the gap region, leading to spacing

losses with a consequent decrease in the reproduced

signal. Head clogging can produce a transfer ﬁlm

which is somewhat like the ‘‘stain’’ described above.

The differences are that the ﬁlm is very much thicker,

more easily removed by wiping with solvents, and

is more likely to be produced at high humidity, the

latter probably due to water assisted breakdown

within the magnetic layer, due to a chain scission

mechanism. We have also seen such an effect and

have shown that this can also cause gap damage

which affects the high-frequency response of the sys-

tem. The wear phase eventually leads to the produc-

tion of more debris and an avalanche catastrophic

failure mode observed by ourselves and others.

It should be noted here that increasing the magnetic

particle loading to a ﬁgure of about 50% by volume

leads to an increase in complex modulus and decrease

in wear. Increasing loading beyond this level, how-

ever, will reduce the modulus and increase wear.

There are two forms of abrasive wear. Two-body

abrasion occurs when a hard, rough surface slides

over a softer one resulting in ploughing action and

removal of material. Three-body abrasion occurs

when hard particles become entrapped between two

sliding bodies, resulting in the same effect. An essen-

tial requirement for abrasive wear is that plastic

deformation should occur beneath a penetrating as-

perity or abrasive particle. Our calculations indicate

that localized contact temperatures can be quite high

and could lead to softening of the polymer matrix,

thus aiding plastic ﬂow, but reference to plasticity

criteria (Greenwood and Williamson 1966) reveals

that with current surface roughnesses two-body abra-

sive wear should not occur either in the media or at

the head.

However, agglomerates of wear debris entrapped

between head and media may be of sufﬁcient size to

give rise to three-body abrasion. This is almost cer-

tainly the origin of scratches seen on tape and dis-

kette surfaces after wear. For polymeric materials the

abrasive wear rate is proportional to S

1

, where S

is the breaking strength of the material and e the

elongation to break. The addition of magnetic par-

ticles will increase S, but reduce e and hence the ﬁlled

polymer is no more resistant to abrasion than the

polymer itself. This is not true of other forms of wear

where the increased hardness and elastic modulus

may reduce wear. In general, abrasive wear of the

media is not too important when considering loss of

signal, but may cause errors. The same cannot be said

when considering the head.

(d) Effect of lubricants

Wear of media and heads is reduced by introducing a

boundary lubricant between the opposing surfaces

and by reducing the polar molecule content of the

binder. Sullivan and Sharma (1992) found lifetimes

reduced by two orders of magnitude in the media

without lubricant. Friction values were relatively un-

affected. Our angle-resolved XPS measurements have

shown areal coverages between 10% and 40%, and

the thickness of the polymer/lubricant covering the

639

Magnetic Reco rding: Flexible Media, Tribology