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

Подождите немного. Документ загружается.

cycles. The Pt layers are typically 0.9 nm thick (ﬁve

atomic layers). Reducing their thickness lowers T

,as

desired, but also the Kerr angle Y

due to less mag-

netic matter in the total ﬁlm and thus diminishes the

potential advantage of the (Pt/Co) multilayers. Re-

ducing T

by alloying Ni to Co avoids this problem

but still reduces Y

at the readout temperature by

about 25%.

The M

of Pt/Co is larger than that of the RE–TM

alloys used for MO recording. This leads to stronger

demagnetizing effects. The magnetic domains tend to

form subdomains and their boundary tends to be-

come curved. Both effects increase the noise of writ-

ten marks (the write noise). The magnetization of the

ferrimagnetic RE–TM alloys is controlled by the

choice of the compensation temperature where

¼0.

The coercive ﬁeld H

of Pt/Co multilayers vanishes

at about 50 1C below T

. Wall motion during the ﬁrst

stage of the cooling in the thermomagnetic write

process is to be expected. This is a principal disad-

vantage for recording because wall motion results in

jagged domain boundaries and in an increase of the

domain length jitter (Mergel 2000, Sect. 6.6).

For MO readout the magnetization must be per-

pendicular to the substrate. The effective perpendic-

ular anisotropy constant K

eff

achieves its maximum

when the Co layer thickness corresponds to two

atomic layers (0.4 nm). In that case all the Co atoms

are forming part of Pt/Co interfaces. It was found

that a pronounced (111)-f.c.c. texture of the multi-

layer is needed to get a large K

eff

but is alone not

sufﬁcient. The most decisive feature is the sharpness

of the interfaces (Bertero and Sinclair 1994). Any

chemical intermixing reduces K

eff

The magnetization M

, the Kerr angle Y

, the

Curie temperature T

, and to a large extent, the an-

isotropy constant K

eff

are material properties deter-

mined on the subnanometer scale whereas the

strength of H

is strongly determined by the micro-

structure of the ﬁlms on the 10 nm scale.

Large values of H

are obtained for exchange-de-

coupled grains in combination with a large K

eff

. The

exchange-decoupling can be achieved by a discontin-

uous microstructure or when the grain boundaries

consist of a nonmagnetic phase.

eff

and H

are sensitive to the details of the dep-

osition process. Evaporated ﬁlms are characterized

by a pronounced (111)-f.c.c. texture with ﬂat inter-

faces leading to a high K

eff

. Furthermore, they exhibit

a porous columnar structure with physically sepa-

rated, exchange-decoupled columns or grains so that

is large.

The microstructure of Ar-sputtered ﬁlms is strongly

inﬂuenced by energetic bombardment by reﬂected (at

Pt) Ar neutrals. This leads to a low H

due to a dense

structure and to a low K

eff

due to irregular interfaces

between Pt and Co (Bertero and Sinclair 1994). Better

results are obtained when the ﬁlms are sputtered with

the heavier noble gases Kr or Xe, leading to less severe

bombardment, or when they are deposited on well

chosen and prepared underlayers.

2. Inﬂuence of Underlayers on the Microstructure

A cross-section of a typical MO disk is shown in

Fig. 2. There is a capping layer only in some disk

designs. Usually a protective layer is deposited on top

of the layer stack.

2.1 Pt Underlayers

Pt/Co multilayers exhibit good properties when

grown as a (111)-f.c.c. stacking sequence. As the sta-

ble phase of Co at room temperature is h.c.p., growth

must start with a Pt layer. The thicker this under-

layer, the better the (111) texture. However, for

substrate-incident readout, the underlayer reduces

the signal due to absorption. To give an example, a

5.5 nm thick Pt underlayer diminishes the readout

signal (carrier) by 3.5 dB (van Kesteren et al. 1991).

The best magnetic properties of Ar-sputtered (at

1.6 Pa) Pt/Co multilayers were obtained on ultrathin

about 3 nm), discontinuous Pt underlayers (Car-

cia et al. 1996). The latter were prepared by d.c.

magnetron sputtering followed by ion etching under

an rf-bias ﬁeld. The largest H

(3.5 kAm

1

) and the

largest K

eff

(1.2 10

erg cm

3

, 1.2 MJ m

3

) were ob-

tained for d

¼3.3 nm before etching. For this thick-

ness the most pronounced superstructure features

were observed in x-ray diffractograms. The micro-

structure of the (Pt/Co) ﬁlm comprised small

(10–20 nm) grains separated by voids and resembled

that observed in vapor-deposited ﬁlms. The etching

of the underlayer likely sputtered away some Pt at-

oms leaving behind highly (111) textured and ﬂat Pt

Figure 2

Layer stack comprising a (Pt/Co) recording layer. Not

all disk designs comprise a capping layer. There is

generally a protection layer on top of the stack.

800

Magneto-optic Recording: Total Film Stack, Layer Conﬁguration

mesas that served as independent nucleation centers

for the multilayers.

For thicker or thinner underlayers the Pt/Co mul-

tilayers are structurally less coherent. On thicker un-

derlayers the surface structure becomes continuous.

When the underlayers are thinner, the grains are sep-

arated but the surface becomes very rough.

2.2 Dielectric Layers

The stack design usually contains a dielectric layer

with an index of refraction n E2 to get an enhance-

ment of the magneto-optic effects by interference.

For RE–TM media only nitrides like Si

or AlN

can be used due to the tendency of the RE elements to

oxidize. For Pt/Co, oxides are also possible and ZnO

and In

have been attempted. The dielectric un-

derlayers inﬂuence the magnetic properties of the

Pt/Co multilayers deposited on them. K

eff

increased

to more than 7 10

erg cm

3

(0.7 MJ m

3

) when the

underlayer (ZnO and In

) was sputter-etched at

100 V (Carcia and Suzuki 2000).

ZnO is the most effective in creating a (111)-f.c.c.

texture and prevents degradation of the recording

properties upon annealing above 300 1C. It has a

hexagonal structure and grows preferentially with its

c axis (0001) perpendicular to the substrate. The mis-

match between the ZnO(0001) and the (Pt/Co)(111)

f.c.c. planes is only 1.6% and the multilayers grow

epitaxially on ZnO.

The structural and magnetic properties of sput-

tered (Pt/Co) multilayers depend on the ZnO depo-

sition temperature T

ZnO

(Carcia et al. 1993). K

eff

was

maximum when ZnO was deposited at 150 1C. For

ZnO

p 200 1C a continuous microstructure was ob-

served and the magnetic reversal was by wall motion.

For T

ZnO

¼500 1C the microstructure is particulate

with a better (111)-f.c.c. texture and the magnetiza-

tion reverses by coherent rotation within the sepa-

rated grains. However, the perpendicular anisotropy

is reduced, correlated with a severe multilayer rough-

ening. This proves that the quality of layering is more

important for a large K

eff

than the texture.

3. Magnetization Reversal

3.1 Magnetic After-effect

In magnetic after-effect experiments, a magnetically

saturated specimen is subjected to an inverse mag-

netic ﬁeld with a value smaller than H

and the mag-

netic relaxation (slow reversal) is observed as a

function of the lag time.

For (Co

0.6

0.4

)[0.5 nm]Pt[X0.6 nm] multilayers

sputtered on a Pt buffer layer, the magnetic after-

effect depends on the sputter pressure (Meng et al.

1999). Sputtering at low Ar pressure results in dense

ﬁlms. The reversal is dominated by wall motion

indicating that the wall propagation ﬁeld is smaller

than the nucleation ﬁeld. The magnetization pattern

is characterized by concatenated small domains. In

discontinuous ﬁlms, grown at high Ar-pressure, a

homogeneous nucleation all over the ﬁlm is observed.

Domain wall motion is inhibited and the magnetiza-

tion reversal occurs by coherent rotation in the iso-

lated grains. The hysteresis curves are sheared,

according to the distribution of nucleation energies.

3.2 Mark Structures

The marks in Pt/Co layers can be investigated by

scanning magnetic force microscopy (van Kesteren

et al. 1991). In this technique, a ferromagnetic tip is

scanned across the object surface to obtain an image

of the magnetization pattern. Pt/Co layers are well

suited because they need no protection layer, in con-

trast to RE–TM layers. This has the advantage that

the parameters measured in a recorder, namely the

writing conditions and the carrier and noise levels,

can be directly related to the magnetic microstructure

of the domains on the same disk.

For layers with high magnetization, marks were

written even without an applied ﬁeld, a consequence

of the demagnetizing ﬁeld in the heated area. They

contained subdomains because the decrease of the

magnetostatic energy is larger than the increase of the

total wall energy. Subdomains decrease the readout

signal. They furthermore lead to ﬂuctuations in the

readout signal and an increase of the write noise.

With increasing external ﬁeld, the subdomains disap-

pear, the carrier increases, and the noise decreases.

For laser intensity modulation, a uniform magnet-

ization of the written marks was only observed for an

external write ﬁeld H

ext

430 kA m

1

(375 Oe). On

these disks, magnetic ﬁeld modulation always pro-

duced irregular domains. This was attributed to the

lack of a wall gradient force. The wall gradient force is

due to the temperature proﬁle and tends to shift the

domain wall into regions of higher temperature, i.e., to

the center of the hot spot. This effect balances the

expanding demagnetizing force and favors circular do-

mains. For constant laser power, the temperature gra-

dient is much smaller than for modulated laser power.

4. Recording Experiments

In order to estimate the recording performance of a

disk, a carrier is written. A carrier consists of peri-

odically written marks and produces a signal with

50% duty cycle (up and down signal have the same

length). The readout signal is analyzed in a spectrum

analyzer. The carrier-to-noise ratio (CNR) is deter-

mined as the ratio of the signal and the noise at the

write frequency. It depends on the linear velocity of

the disk and the residual bandwidth of the spectrum

analyzer.

801

Magneto-optic Recording: Total Film Stack, Layer Conﬁguration

4.1 Magnetic Field Sensitivity

In Fig. 3, the recording performance of (Pt/Co) and

GdDyFeCo media is compared (Teragaki et al.

1993). The carrier of Pt/Co saturates at a higher

modulation ﬁeld, 200 Oe (16 kA m

1

), instead of

100 Oe (8 kAm

1

) for GdDyFeCo. The reason is the

higher M

leading to stronger demagnetizing effects.

The higher magnetic ﬁeld H

is necessary to obtain

complete marks without subdomains although the

magnetic reversal starts at a lower ﬁeld due to a

higher Zeeman energy m

A high magne tic ﬁeld sensitivity was obtained in

high-M

ﬁlms when the formation of subdomains was

suppressed by a thin (2–5 nm) RE–TM capping layer,

RE-rich and with a high T

(4280 1C, Ishida et al.

1997). This works for Tb

¼100 Oe)

and for Pt/Co (140 Oe). The CNR obtained with a

light wavelength l ¼680 nm was 2 dB higher for the

RE–TM medium. The ferrimagnetic RE–TM-based

media have the advantage that the magnetization in the

recording relevant temperature region can be optimi-

zed towards best recording performance by the choice

of the compensation and the Curie temperatures.

4.2 Competition of Carrier-to-noise Ratios

For Pt/Co multilayers, the readout signal at smaller

wavelengths is higher than for conventional RE–TM

alloys due to a stronger Kerr effect. However, this

does not imply automatically a better performance of

Pt/Co media and a higher storage density on these

media in ‘‘blue’’ recording. The performance of a

drive-medium system depends not only on the signal

level but also on the various noise levels of the disk

and the system noise.

A competition between two disk families, one with

a higher signal level, the other with a lower medium

noise level is illustrated in Fig. 4. For a large system

noise (‘‘1’’), disk 2 with the higher signal level yields a

relatively better performance, whereas disk 1 with the

lower medium noise performs better for a lower sys-

tem noise (‘‘2’’). Therefore, the various noise sources

have to be considered in detail when predictions on

the recording performance of a medium class are

made.

4.3 Noise Sources

The system noise is measured on a stationary disk and

arises from the light source, the detector, and the

electronic circuit. In the ﬁrst recording experiments

with a green laser, the detector noise was initially very

strong but its inﬂuence on the CNR could be reduced

by increasing the laser intensity. Later a specially de-

signed PIN photodetector with a high sensitivity was

used (Kaneko et al. 1993).

The disk noise is the difference between the noise of

the rotating and the stationary disk in a magnetically

initialized state without magnetic marks. It originates

from ﬂuctuations of the polarization due to spatial

ﬂuctuations of the magneto-optical properties of the

recording layer or due to the roughness of underlay-

ers or the substrate. In Fig. 3, the disk noise is larger

for the Pt/Co medium. This is due to additional

polarization noise introduced by a nonperfect (111)

texture of grain orientations.

The write noise is the additional noise measured

on a rotating disk with written marks. Both subdo-

mains in the mark and irregularities of the mark edge

contribute to the write noise. A higher write noise is

expected for Pt/Co media because the domain length

jitter is about 40 nm (Teragaki et al. 1993) compared

Figure 4

Comparison of signal and noise levels for two competing

MO media. Disk 1 has the higher signal level whereas

disk 2 has the lower disk noise level. It depends on the

level of the system noise which disk exhibits the higher

CNR. The CNR is given by the distance between the

signal and the highest noise level.

Figure 3

Carrier C and noise N of a Pt/Co and a GdDyFeCo

medium. Laser wavelength 780 nm. Schematically after

Teragaki et al. (1993).

802

Magneto-optic Recording: Total Film Stack, Layer Conﬁguration

to 15 nm in the best RE–TM media (see Mergel

2000).

The sum of write and disk noise is called medium

noise. It arises from geometrical irregularities on the

disk. As it is readout by the optical system, it is sub-

jected to the optical transfer function. As a conse-

quence, the CNR depends on the mark length that

determines the spatial frequency at which the CNR is

measured. This effect is important when judging the

relative performance of RE–TM and (Pt/Co) media.

For marks shorter than 0.5 mm, Pt/Co performs bet-

ter because the medium noise is suppressed below the

system noise by the optical transfer function (Kaneko

et al. 1993). This corresponds to ‘‘system noise 1’’ in

Fig. 4. RE–TM performs better for longer marks due

to a lower medium noise, ‘‘system noise 2.’’ There-

fore, RE–TM media are expected to perform better

even for shorter marks when the system noise is re-

duced (Mergel 2000).

5. Concluding Remarks

Pt/Co media exhibit a high magnetization such that a

higher magnetic ﬁeld is necessary to suppress demag-

netizing effects and to obtain saturated marks. The

modulation ﬁeld can be reduced by a thin RE–TM

capping layer with high T

on top of the recording

layer. Then, however, the advantage of intrinsic cor-

rosion resistance is lost.

A discontinuous microstructure with exchange-

decoupled grains is necessary to obtain a high coer-

cive ﬁeld H

. This can be achieved by ion-etched Pt

nucleation layers on top of the dielectric layer. Ion-

etched dielectric underlayers can promote the spatial

coherence and the ﬂatness of the interfaces of Pt/Co

multilayers, the key feature to obtain a large perpen-

dicular magnetic anisotropy. ZnO is especially effec-

tive because the mismatch of the lattices is only 1.6%

and enables heteroepitaxial growth.

The higher signal level of Pt/Co for shorter wave-

lengths is counteracted by a currently higher medium

noise level, both relative to RE–TM. In such a sit-

uation, the measured signal-to-noise ratios depend on

the details of the recording and readout process. The

currently reported domain length jitter is 40 nm as

compared to 15 nm for RE–TM resulting in a higher

write noise. Therefore, a higher storage density than

for RE–TM media cannot be expected in ‘‘blue’’ re-

cording when the system and disk noise are sup-

pressed below the write noise by improved disk

manufacturing.

The reason for the large jitter is the temperature

dependence of the coercive ﬁeld, H

(T). H

(T) van-

ishes at 50 1C below the Curie temperature such

that the domain walls can easily move in the ﬁrst

stage of the cooling in the thermomagnetic write

process and become jagged under nonoptimal re-

cording conditions.

See also: Chemical Stability and Life Time; Magneto-

optical Disks: Thermal Effects; Magneto-optical Ef-

fects, Enhancement of; Magneto-optic Multilayers;

Magneto-optic Recording Materials: Chemical Stabi-

lity of Life Time; Magneto-optic Recording: Overwrite

and Associated Problems; Super-resolution: Optical

and Magnetic Techniques

Bibliography

Bertero G A, Sinclair R 1994 Structure–property correlations in

Pt/Co multilayers for magneto-optic recording. J. Magn.

Magn. Mater. 134, 173–84

Carcia P F, Coulman D, McLean R S, Reilly M 1996 Magnetic

and structural properties of nanophase Pt/Co multilayers.

J. Magn. Magn. Mater. 164, 411–9

Carcia P F, Li Z G, Reilly M 1993 The magnetic and micro-

structural properties of Pt/Co multilayers grown on ZnO.

J. Appl. Phys. 73, 6424–6

Carcia P F, Suzuki T 2000 New MO recording media based on

Pt/Co and Pd/Co multilayers and Pt–Co alloys. In: Gambino

R J, Suzuki T (eds.) Magneto-optical Recording Materials.

IEEE Press, Piscataway, NJ, Chap. 3

Gambino R J, Suzuki T (eds.) 2000 Magneto-optical Recording

Materials. IEEE Press, Piscataway, NJ

Ishida M, Katsyuama T, Kawase T, Nebashi S, Shimoda T

1997 High magnetic ﬁeld sensitivity of TbFeCo layer and

Pt/Co multilayers with an ultra-thin RE-rich RE–TM layer.

IEEE Trans. Magn. 33, 3229–31

Kaneko M, Sabi Y, Ichimura I, Hashimoto S 1993 Magneto-

optical recording on Pt/Co and GdFeCo/TbFeCo disks using

a green laser. IEEE Trans. Magn. 29, 3766–71

Meng Q, Schuurhuis J R, Ladder J C, Meyer P, Jamet J P,

Ferre

J 1999 Magnetization reversal processes in CoNi/Pt

multilayers. J. Magn. Soc. Jpn. 23 (51), 87–90

Mergel D 2000 Domain dynamics and recording physics in

magneto-optical recording media. In: Gambino R J, Suzuki

T (eds.) Magneto-optical Recording Materials. IEEE Press,

Piscataway, NJ, Chap. 6

Mergel D, Hansen P, Raasch D 1992 The physical basis of

magneto-optical recording. Proc. SPIE 1663, 240–9

Teragaki Y, Kusumoto Y, Sumi S, Torazawa K, Tsunashima S,

Uchiyama S 1993 Magnetic ﬁeld modulation recording on

Pt/Co magneto-optical disks. IEEE Trans. Magn. 29, 3796–8

van Kesteren H W, den Boef A J, Zeper W B, Spruit J H M,

Jacobs B A J, Carcia P F 1991 Scanning magnetic force mi-

croscopy on Co/Pt magneto-optical disks. J. Appl. Phys. 70,

2413–22

D. Mergel

University of Essen, Germany

Magneto-optical Disks: Thermal Effects

In thermomagnetic recording on magneto-optical

(MO) data storage media, focused light energy is di-

rected to a moving magnetic ﬁlm structure to control

the local temperature distribution. The magnetic ma-

terial typically has a strong native anisotropy which

803

Magneto-optical Disks: Thermal Effects

restricts the magnetization alignment perpendicular

to the ﬁlm plane, or nearly so. Owing to the strong

temperature dependence of the magnetic properties

of the medium, domain reversals that represent a

pattern of recorded data can be induced, usually with

the aid of an external bias magnetic ﬁeld. Subsequent

to recording, readout of written data is performed by

reﬂecting a focused beam of reduced power from the

impressed domain pattern to interrogate the magnetic

state of the recorded track by using the MO Kerr

effect (see Faraday and Kerr Rotation: Phenomeno-

logical Theory). The resultant heating of the medium

during reading may or may not be instrumental for

the proper performance of the readout function, de-

pending on the design of the medium.

1. Regime of Thermal Transport in MO

Recording

Much has been published on thermal performance of

MO media (McDaniel and Victora 1997, Mansuripur

1995). Normally, the thermal physics involved in

treating recording and readout is well within the limits

of classical heat conduction in homogeneous media.

Therefore, the well-known analysis of classical refer-

ences (Carslaw and Jaeger 1959) applies, and the

complications of microscale heat transport or phase

change can be ignored. Even with these simpliﬁca-

tions, an MO media designer routinely uses numerical

analyses of heat conduction by computer simulation

due to the relatively complicated thin ﬁlm and disk

substrate structures involved. Non-numerical thermal

analysis methods for moving multilayer thin ﬁlm

stacks, even those on planar substrates, may be too

complicated or lack generality for routine engineering

design practice (Madison and McDaniel 1989).

Heat conduction in the solid MO ﬁlm struc-

tures obeys the standard diffusion-like differential

equation:

T 

Qðx; y; z; tÞ

¼ Aexpfb

½ðx  vtÞ

þ y

g

 expðazÞCðtÞð1Þ

Here T(x,y,z,t) is the temperature ﬁeld of interest,

k ¼k/rc is the thermal diffusivity, k is the thermal

conductivity, r is the material mass density, c is the

material mass specific heat, v is the moving media ve-

locity, and Q is the thermal power input per volume.

In the case of laser beam heating for recording on a

typical MO medium thin ﬁlm structure as shown in

Fig. 1, Q should be represented as the Poynting vector

energy deposition rate from absorption of the incident

electromagnetic beam. The far right side of Eqn. (1)

can be used for a simpliﬁed case in which a moving

beam of Gaussian proﬁle undergoes Beer’s law ab-

sorption in a single, thick ﬁlm. a is an absorption co-

efﬁcient, b is an inverse beam size parameter, and C(t)

is an incident beam power-versus-time proﬁle.

Heat diffusion in a thin ﬁlm structure is frequently

described using characteristic lengths and times for

the thermal transport described by L ¼(kt)

0.5

and

t ¼L

/k. The various ﬁlm materials in typical MO

media structures have very different diffusivities, as

indicated in Table 1. Therefore, characteristic lengths

and times are most applicable to either single ﬁlms, or

as effective values derived from numeric analyses that

weigh together composite results from a realistic, but

more complicated structure. Characteristic times in

thermal diffusion analysis can also be estimated by

relating heat ﬂow to an electrical circuit analogue. If

Figure 1

First-surface ﬁlm stack geometry deposited on a grooved substrate. The geometry of the recording track (land L), the

inter-track groove G, the groove depth D , and the groove side wall angle Y help control heat diffusion.

804

Magneto-optical Disks: Thermal Effects

one relates temperature to voltage, heat ﬂux to current,

thermal resistance R

¼L/kA (where L and A are

effective conductive path length and cross-sectional

area, respectively) to electrical resistance, and heat

capacity (cm,withm the heat conductor’s mass) to

electrical capacitance, then the time for the charging of

an electrical RC network (t ¼RC) becomes t ¼R

cm.

This leads again to the expression t ¼L

/k.

We note that the thermal constants of the thin ﬁlms

shown in Table 1 may differ considerably from those

of the corresponding bulk material, and they may be

adjusted considerably by altering the ﬁlm composi-

tions or deposition processes. One expects higher

phonon and electron scattering in thin ﬁlms due

to the effects of layer boundaries and ﬁlm growth

morphology.

In thermomagnetic recording, the thermal energy

deposited by an incident light beam should provide

locally concentrated heating so that well-controlled

magnetic domain reversals are established along a data

track. Lateral heat spreading due to media motion

(distributed dosing) and thermal diffusion must be

limited to enable high resolution writing without ex-

cessive cross-write and cross-erase of previously re-

corded adjacent information. The means available to

aid in this purpose are (i) surface topography of the

thin ﬁlms and substrate, (ii) thermal design of the thin

ﬁlm stack, and (iii) the writing strategy employed to

apply the radiant energy. In Fig. 1, thermal conﬁne-

ment grooves provide an example of (i), while (ii) is

illustrated by the reﬂector ﬁlm, which can serve as a

heat-sinking layer to minimize lateral heat diffusion.

Regarding writing strategy (iii), probably the most im-

portant fundamental to remember in the heating of

moving thin ﬁlm structures with focused light beams is

the following: The most efﬁcient heating with the highest

resultant thermal gradients is always achieved through

the use of short duration, low duty-cycle irradiation

pulses (McDaniel and Victora 1997, McDaniel 1999).

2. Fundamentals of Thermal Design for MO Disk

Thin Film Structures

It is evident from Eqn. (1) that the minimum lateral

extent of a heated region on the MO medium is the

width of the incident focused light beam. Disk mo-

tion under the beam and lateral thermal diffusion can

only enlarge the heated zone. Therefore, to achieve

high-resolution, high-performance recording, the MO

disk designer must counteract these tendencies with a

combination of means as indicated in the previous

section. Fortunately, each of the three avenues for

control of lateral heat spreading (topography, ﬁlm

design, writing strategy) can be accurately modeled

with a numerical computer simulation. The most

powerful and ﬂexible of the simulation tools for this

purpose provides a solution T(x,y,z,t) in two steps.

First, the three-dimensional geometry of the substrate

with ﬁlm coatings (see, for example, Fig. 1) is exposed

to a radiant impulse from a prescribed focused beam,

and a fully time-developed ‘‘thermal impulse solu-

tion’’ is obtained from a ﬁnite-element or ﬁnite-

difference solver operating on Eqn. (1). As long as the

thermal problem is deﬁned to be linear (material

thermal properties independent of temperature; no

material phase changes), superposition of the impulse

solution in space and time with appropriate weighting

according to the input light beam power–time proﬁle

will yield the desired ﬁnal solution T(x,y,z,t ), even for

a moving beam (McDaniel and Victora 1997, Mans-

uripur 1995).

Substrate and ﬁlm topography departing from per-

fect planarity can increase the impedance to lateral

heat ﬂow. This is particularly true if the dominant

thermal conduction path(s) is(are) altered or inter-

rupted. For example, in Fig. 1 the side walls of the

thermal conﬁnement grooves are likely to have re-

duced thickness ﬁlm layers (as measured normal to

the local surface) if the incident material species dur-

ing deposition travel predominantly in the z direc-

tion. Consequently, it is expected that lateral heat

conduction can be shaped with features such as those

depicted in Fig. 1 (McDaniel 1999), Peng and Mans-

uripur 2000).

Next, consider the ﬁlm design issues. The basic

physics of thermal transport involves optical absorp-

tion of the incident radiation (depending on how ﬁlm

stack reﬂectance R and absorbance A are split, given

the constraint R þAE1 in an opaque structure) on a

time scale of o10

12

s, followed by diffusion of the

Table 1

Materials and their properties for a nominal quadrilayer MO structure (see Fig. 1).

Material Thickness (nm) nk

opt

k (W m

1

) c (J kg

1

) r (kg m

3

) k (10

6

1

)

Air N 1 0 0.03 1150 1.15 22.7

SiN 85 2.02 0.01 2.0 740 3400 0.79

MO 20 2.4 3.6 7.0 360 7900 2.46

AlX 40 1.8 6.8 25 900 2700 10.3

Polycarbonate B10

1.58 0 0.15 300 1200 0.42

glass B10

1.5 0 1.0 860 2300 0.51

805

Magneto-optical Disks: Thermal Effects

resultant thermal energy by conduction in all direc-

tions on time scales ranging from B10

10

to 10

6

Light absorption occurs predominantly in the ﬁrst

ﬁlm that the incident light encounters that has a large

value of k

opt

(see Table 1). Heat sinking to a reﬂector

layer at the base of the ﬁlm stack is made increasingly

effective by (i) increasing the heat capacity (volume)

of the reﬂector, (ii) increasing the thermal conduc-

tivity of the reﬂector, and (iii) decreasing the thermal

resistance between light absorption location and the

reﬂector.

Intuitively, we expect an effective heat sink layer to

increase the heat ﬂow in the axial (z) direction away

from the light absorption zone, and to diminish the

lateral ﬂow. Surrounding the thin, but relatively high

thermal conductivity MO layer with insulating die-

lectric ﬁlms further suppresses the relative importance

of lateral heat ﬂow. Simple analysis of characteristic

times for these respective ﬂow directions tells us that

the axial channel is at least an order of magnitude

faster than the lateral ﬂow channel. This is the pri-

mary reason that radiation power input by short-du-

ration, low duty-cycle pulses couples energy

efﬁciently to the fast axial ﬂow channel, and mini-

mizes the (detrimental) effect of the slow lateral ﬂow

channel in spreading heat.

3. Issues in Advanced MO Film Designs and

Limits of Thermomagnetic Writing

Advanced MO ﬁlm designs for the attainment of

higher performance or areal recording density usually

have more complex magnetic ﬁlm structures than that

shown in Fig. 1. In these cases, there typically are two

or more magnetic layers with thermally controlled

magnetic coupling to each other. Examples of these

design structures include light-intensity modulated

direct overwrite (LIM-DOW), magnetically induced

super-resolution (MSR), and magnetic ampliﬁcation

MO system (MAMMOS). The total thickness of the

metallic magnetic layers in these designs (B50–

100 nm) is much greater than that of the simple,

single-layer structure illustrated in Fig. 1 and Table 1.

Consequently, the effectiveness of a design strategy

utilizing a heat-sinking layer at the base of the ﬁlm

stack is significantly compromised. The incident light

is totally absorbed in the thick magnetic layers, and

the resultant heat will diffuse laterally with diminished

control from the more distant underlying heat sink.

Another inevitable trend for high-performance

MO storage media will be the necessity to increase

the recording and playback rate. Data throughput

rates for rotating memory products are driven higher

by increasing the disk rotation rate or increasing

the linear bit density along the track, or both. Notice

that the former strategy increases the relative linear

speed of the write/read beam and the storage medium,

while the latter requires improved along-track writing

and reading resolution. The resolution limitations

when using an optical stylus whose lateral extent is

WBl/NA (wavelength/numerical aperture) are well

known and obvious. But increasing the linear speed

of recording combines with the higher linear bit den-

sity to shorten the bit transit time to tBL

bit

/v. Even-

tually, t becomes shorter than the shortest thermal

characteristic time of the recording medium, regard-

less of the aggressiveness of the ﬁlm thermal design or

the writing strategy. At this point, one has lost con-

trol of shaping the thermal environment for thermo-

magnetic writing, and the viability of a recording

process based on impressed thermal gradients

becomes questionable.

See also: M agneto-optic Recording Materials: Chemi-

cal Stability and Life Time; Magneto-op tic Recording:

Overwrite and Associated Problems; Multilayer Op-

tical Film Modeling; Super-resolution: Optical and

Magnetic Techniques

Bibliography

Carslaw H S, Jaeger J C 1959 Conduction of Heat in Solids.

Oxford University Press, Oxford

Madison M R, McDaniel T W 1989 Temperature distributions

produced in an N-layer ﬁlm structure by static or scanning

laser or electron beam with application to MO media.

J. Appl. Phys. 66, 5738–48

Mansuripur M 1995 The Physical Principles of MO Recording.

Cambridge University Press, Cambridge

McDaniel T W 1999 Thermal design of high performance mag-

neto-optical data storage ﬁlms. Proceedings MORIS ’99.

J. Magn. Soc. Jpn., 23 Suppl. S1, 251–256

McDaniel T W, Victora R H (eds.) 1997 Handbook of MO Data

Recording. Noyes, Park Ridge, NJ

Peng C, Mansuripur M 2000 Thermal cross-track cross talk in

phase-change optical disk data storage. J. Appl. Phys. 88,

1214–20

T. W. McDaniel

Seagate Research, San Jose, California, USA

Magneto-optical Effects, Enhancement of

The fundamental magneto-optical Kerr and Faraday

effects, described phenomenologically in Faraday

and Kerr Rotation: Phenomenological Theory, are in-

trinsically very small. However, with modern light

sources and signal-processing techniques the various

phenomena, discovered originally by Michael Fara-

day and John Kerr (Faraday 1845, Kerr 1877), can be

easily observed. Indeed, with optical modulation

techniques, it is possible to detect magneto-optic sig-

nals associated with thin layers of magnetic materials

that are effectively one atom thick or less. However,

806

Magneto-optical Effects, Enhancement of

for commercial applications aimed at mass markets,

such as magneto-optic memories, the costly use of

sophisticated electronic detection techniques and the

need to read magnetically stored information very

rapidly requires one to optimize the sensitivity with

which magneto-optical signals can be detected. The

relationship between the size of the magneto-optical

effects and the sensitivity with which they can be de-

tected is discussed below. The inﬂuence that the var-

ious optical and magneto-optical parameters of the

material have on the detection process is determined

and various techniques are outlined for increasing the

size of magneto-optical effects and, in particular,

maximizing the sensitivity with which they can be

detected. In this article it is not possible to treat all

magneto-optic conﬁgurations (see Faraday and Kerr

Rotation: Phenomenological Theory) and we shall

concentrate on the polar Kerr effect at normal inci-

dence, since this is most applicable to the develop-

ments in magneto-optic data storage systems.

However, much of the underlying philosophy of en-

hancement applies to the alternative situations and,

where appropriate, this will be pointed out.

In most instances, whether one is working in re-

ﬂection (Kerr effect) or transmission (Faraday effect),

the magneto-optical signal arises from a magnetiza-

tion-sensitive polarization change, usually from a lin-

ear state to an elliptical state, with the major axis of

the polarization ellipse rotated slightly with respect to

the incident polarization direction. This of course

only applies to the polar and longitudinal cases and

not to the transverse case. It is shown in Faraday and

Kerr Rotation: Phenomenological Theory that the po-

larization ellipse is the result of the combination of

two orthogonal vectors, one optical and the other

magneto-optical. These are the Fresnel reﬂection co-

efﬁcient, r, and the Kerr coefﬁcient, k, respectively.

Under normal circumstances, in thin absorbing ﬁlms,

the condition 7k757r7 applies, with the result that a

very small, complex magneto-optical rotation,

may be deﬁned as

¼ y

þ ie



cos D

þ i



sin D

ð1Þ

where D

is the phase difference between k and r, and

and e

are the Kerr rotation and ellipticity, re-

spectively. Intuitively, it seems obvious that the larger

this complex rotation the easier it is to detect, and in

general this is true. However, as is shown below, one

may reduce the reﬂection coefﬁcient, r, to a value

comparable to k with the result that very large ro-

tations could, in principle, be obtained, albeit with

very low accompanying reﬂectivity. It should be

made clear, however, that despite the large effective

magneto-optical rotations that would result, it is

undesirable to have this accompanied by a sample

surface that reﬂects little or no light at all. The role

of sample reﬂectivity, R( ¼7r7

), and the associated

magnitude of complex Kerr rotation in determining

the sensitivity by which a magneto-optic signal can be

detected are clearly of importance, and the way in

which each contributes to determining the signal-

to-noise ratio in any detection system requires careful

attention.

1. Signal-to-noise Ratio

In magneto-optic recording systems, the most

common detection method is one that relies on dif-

ferential detection (Egashira and Yamada 1974)

employing two balanced detectors, as shown sche-

matically in Fig. 1. For simplicity the incident radi-

ation is not shown in this diagram. At any given

moment the two detectors are, to a very good ap-

proximation, balanced in their output by having al-

most equal light intensities falling on them. The

advantage of such a system is that variations in the

intensity of the radiation source (often a laser diode)

are effectively removed from the differentially ampli-

ﬁed output of the two detectors. After traversing the

polarizing beam-splitting optics of the system, mag-

neto-optical rotations, induced by magnetization

changes in the sample surface, cause small imbalan-

ces in the ampliﬁed output and this becomes the de-

tected signal. Accompanying this is a series of noise

levels whose origins may be fundamental (shot noise),

electronic (detector and ampliﬁer noise), or associ-

ated with ﬂuctuations of the laser light source. It

is the ratio of the signal to the noise level that de-

termines the degree of precision with which the mag-

neto-optical effects can be detected. Atkinson et al.

(1991) have shown that for such a system the signal-

Figure 1

Schematic diagram of a magneto-optic differential

readout system.

807

Magneto-optical Effects, Enhancement of

to-noise power ratio is given by

SNR ¼

Rð1  RÞ

cos

e þ sR þðl þ mÞc

Rð1  RÞ

cos

ð2Þ

where

7k7 ¼ c

ð1  RÞð3Þ

P is a constant related to the optical read power and

quantum efﬁciency and gain of the detectors, e is the

electronic noise, s determines the shot-noise level,

proportional to disk reﬂectance, and l and m are the

laser feedback and medium noise, both of which are

assumed to come through the differential detection

system in proportion to the Kerr signal.

An important parameter in Eqns. (2) and (3) is c

known as the potential Kerr coefﬁcient. This para-

meter, introduced by Gamble et al. (1985), may be

understood through Eqn. (3) as being the maximum

possible amplitude of the Kerr coefﬁcient, produced

when the overall system reﬂectance is reduced to zero.

The principles of increasing SNR through magneto-

optic enhancement now reduce to three simple re-

quirements. These are to maximize the value of c

for the particular magnetic medium; to ensure zero

ellipticity (cos D

¼71); and to do this for a system

reﬂectance that maximizes Eqn. (2) or is compatible

with the practical needs of providing sufﬁcient optical

feedback for the purposes of tracking. Depending on

the levels of the various noise sources, the intensity

reﬂectance can lie between the limits of zero and 33%

and is typically 10–20%.

2. Maximizing the Kerr Effect

The most important aspect of magneto-optic en-

hancement is the maximization of the potential Kerr

coefﬁcient and this may be accomplished by the

use of a quadrilayer system, as shown in Fig. 2.

Enhancement is produced through the use of a high-

reﬂectance mirror placed behind a ﬁlm of the

magnetic medium that is thin enough to allow trans-

mission of the incident radiation. The thickness,

, of the magnetic ﬁlm must therefore be much less

than the optical skin depth determined by the ab-

sorption coefﬁcient of the material at the design

wavelength, l

. It is the thickness; the refractive

index, n ( ¼n

þin

), and magneto-optic parameter,

Q ( ¼Q

þiQ

), of the magnetic layer; and the re-

ﬂectance of the mirror that are important in maxi-

mizing c

. Indeed its value is unaffected by any

optical layers in the system that lie on the side of

incidence (e.g., the incident layer).

The mirror itself must satisfy two important re-

quirements: it must have a high-amplitude reﬂectiv-

ity, 7r

7, into air; and the phase change, r

,on

reﬂection from the mirror should be controllable.

This is most easily achieved using an opaque alumi-

num ﬁlm separated from the magnetic layer by a low

refractive index spacer layer. The spacer layer has

little effect on the reﬂectivity but its thickness, d

provides control over r

. It can be shown that the

potential Kerr coefﬁcient for such a system is given

by (Gamble et al. 1985)

¼ð1 þ G

1=2

GQjj=4 ð4Þ

where

G ¼

ð5Þ

and

G ¼ Gðn

; n

; d

; r

; l

Þð6Þ

G is a complex gain function derived by Gamble et al.

(1985). The form of the function can be described

graphically by a three-dimensional plot showing how

G varies with 7r

7 and r

. A typical plot for an ab-

sorbing magnetic layer is shown in Fig. 3. It is im-

portant to note that this function does not change

much with d

, providing d

oskin depth, and that

many optically absorbing media show similar plots.

Even more important to note is that G is a maximum

for the largest value of 7r

7 but that there is quite a

range of values of r

over which G remains relatively

high. It is this ﬂexibility in r

that allows us to pro-

ceed with simple designs for enhancement.

3. Quadrilayer

At this point there are two ways of proceeding. In the

ﬁrst, called a quadrilayer (Atkinson et al. 1991), the

thicknesses of the magnetic and spacer layers are ad-

justed to give a zero reﬂectance into the medium of

the incident layer. This almost always requires a

magnetic layer thickness that is less than the skin

Figure 2

Quadrilayer structure for a magneto-optic disk, showing

the reﬂectivity, r

, of the mirror system into a ﬁctitious

air gap.

808

Magneto-optical Effects, Enhancement of

depth and since there is considerable ﬂexibility in the

required value of r

, the maximum value of the po-

tential Kerr coefﬁcient is usually achieved quite eas-

ily. Having reduced the reﬂectance from the top of

the magnetic layer/mirror system to zero, only the

magneto-optical Kerr coefﬁcient, k, remains. The de-

sired ﬁnite reﬂectivity, r, is returned to the system

from the single upper surface of the incident layer

and is determined by the refractive index of the ma-

terial of this layer. This combines with the Kerr co-

efﬁcient to produce a complex Kerr rotation in the

usual way. However, the phase, D

, between k and r

can be adjusted by varying the thickness of the in-

cident layer and in this way can be set equal to zero

or p. This may be done without affecting the value of

either k or r.

4. Trilayer

An alternative to the quadrilayer and one that is

more practical and adequate for most purposes is the

trilayer (Atkinson 1993), where the incident layer

shown in Fig. 2 is omitted. In this case the overall

reﬂectivity is controlled by the thicknesses of the

magnetic and spacer layers. Thus, there are only two

variables of the system that can be adjusted to max-

imize c

, obtain a particular reﬂectance, and main-

tain a zero ellipticity. Theoretically it is rarely

possible to satisfy all three conditions simultaneous-

ly. However, by sacriﬁcing a small percentage in c

and hence maximum Kerr rotation, zero ellipticity

can be obtained for a range of reﬂectances.

The design procedure, based on the use of CoPt,

is illustrated graphically in Fig. 4. This shows how

the intensity reﬂectance, R, of the three-layer system

varies with the thicknesses d

and d

. Also shown in

Fig. 4 is a single contour (dashed line) within which

7cos D

740.99. The optical and magneto-optical

constants used to generate these contour maps are

given in Table 1.

Figure 4 demonstrates how the reﬂectance of the

system can be controlled and that, as required for the

quadrilayer design, an antireﬂection condition can be

produced at the point marked AR. It is worth point-

ing out that the Kerr rotation for such a point would

be 7901, although the reﬂectivity of the system

would be very low indeed. More important is the area

within the dashed contour corresponding to cos

¼0.99. Within this region three designs have

been marked (þ) that would give zero ellipticity for

reﬂectances of 10%, 20%, and 30%.

What is not clear from these ﬁgures is the extent

to which the designs will achieve optimization of

the Kerr rotation through the maximization of the

potential Kerr coefﬁcient. In order to check this as-

pect of the design one must compare the resulting

Figure 3

Typical gain function for an absorbing material with

oskin depth.

Figure 4

Reﬂectance (—) and cos D

(- - -) contours as a function

of the thicknesses of magnetic and spacer layers

¼350 nm).

Table 1

Optical and magneto-optical parameters of materials at

¼350 nm.

Material n

CoPt 1.38 2.84 0.0138 0.008

SiO

1.48 0 0 0

Al 0.38 4.23 0 0

809

Magneto-optical Effects, Enhancement of