novel concepts for mass storage of archival data

Nuclear Instruments and Methods in Physics Research B 218 (2004) 343–354

www.elsevier.com/locate/nimb

Novel concepts for mass storage of archival data

S. Kalbitzer

Ion Beam Technology – Consulting, Bahofweg 2, D-69121 Heidelberg, Germany

Abstract

Energetic beams of heavy ions are an excellent tool for recording digital information in insulating materials. This

ionographic process is based on the introduction of radiation damage up to the amorphisation level in thin layers of

insulators, preferentially of elemental semiconductors such as silicon or diamond. Due to the greatly enhanced optical

absorption in locally irradiated areas strong optical contrast with reference to the non-damaged crystalline matrix is

realised. The generated patterns are sufficiently stable as to guarantee a practically unlimited lifetime. Focused ion

beams of heavy rare gases, as generated from a gas-field ion source, constitute an effective pencil for writing pixels of

dimensions as small as 10 nm. Thus, by using records of 100 and 10 nm pixel diameter one can store about 1 and 100

Tbit, respectively, on a 100 cm2 disc. Whereas standard optical reading technology is confined to the upper size limit,

near-field technology is required for smaller structures. The feasibilty of recording in some group IV materials will be

demonstrated.

� 2003 Elsevier B.V. All rights reserved.

1. Introduction

The aims of materials research have been to de-

velop new solid media and working tools. Joint ef-

forts by industry and research institutions have

advanced the use of these developments in the very

important social area of information technology

serving many fields of our cultural life. In these

computer related technologies, rapidly advancing

progress has always demanded improved solutionsfor mass data storage. This paper is concerned with a

novel approach to the recording of archival infor-

mation, i.e. permanent storage of the WORM type

(write once read many times). One may define four

basic data processing steps in information handling:

collecting, recording, reading and copying.

E-mail address: [email protected] (S. Kalbitzer).

0168-583X/$ - see front matter � 2003 Elsevier B.V. All rights reser

doi:10.1016/j.nimb.2003.12.029

In particular, after collection of archival data

on a temporary medium, e.g. a magnetic tape,transfer to a permanent storage system is required.

Safe archival data storage has become an urgent

issue. Electronic libraries already constitute an

important issue in this field. Foreseeable total

storage volumes approach the order of 1 Eb, or

1018 b. The presently employed solid memories,

however, suffer from lifetimes of the order of 10

years, so that an enormous effort in continuouslyrewriting library archives onto freshly prepared

media is anticipated. In addition, in view of the

capacious memory volumes, higher storage densi-

ties become necessary.

This is true for both recording techniques

presently in main use: magnetic and optical disks

are affected by thermal, mechanical and chemical

events. In addition, magnetic information can alsobe destroyed by electromagnetic fields. Thus, it is

ved.

mail to: [email protected]

344 S. Kalbitzer / Nucl. Instr. and Meth. in Phys. Res. B 218 (2004) 343–354

of paramount interest to find a truly �permanent�recording technology with the additional condition

of higher storage density.

The main aim of this contribution is to presenta novel approach to this problem. The reading of

these digital data with nanometer dimensions is

also subject to further technical progress. This

problem, however, shall not be dealt with here in

detail. Also, novel copying techniques will be

touched only briefly.

2. A novel approach to recording

The two radical improvements in recording

data are concerned with a better writing instru-

ment, a finer �pencil�, and a better writing material,

a non-corrosive �paper�. So far, focused beams of

electrons and light have been employed to produce

the desired changes of material properties. Scat-tering and diffracting processes pose natural limits

for the size of the written image spot, the pixel. In

order to avoid the deficiencies of the presently

employed recording media the ideal �paper� for

archival storage should combine the properties of

immunity to corrosion, support of nano-structures

and infinite pixel lifetime.

2.1. Immunity to corrosion

Presently, plastic material is a major component

of optical and magnetic recording disks. Due to its

organic composition ageing effects occur within a

decade. Thus, there is insufficient reliability for

truly long-time storage. As a consequence, other

materials of higher stability have to be developed.As compared with multi-layered structures, usu-

ally prepared by coating techniques, monolithic

layers, produced by epitaxy, are much less vul-

nerable to adhesive failures induced by unavoid-

able ambient conditions.

2.2. Support of nanostructures

Laser beams have been used for writing optical

patterns into materials of complex composition.

Due to geometrical limits set by light focusing and

the writing process itself, however, minimum

optical features typically extend to several wave

lengths. Thus, a spot size of about 1 lm appears to

be typical for this material modifying process, due

to ablation or melting. Thus, at best a pixel densityof the order of 0.1 Gpix/cm2 can be obtained.

The useful density of magnetic bits depends on

size and stability of the magnetic grains embedded

into some suitable substrate material, usually

plastic. The physical size of a magnetic memory

cell, although greatly reduced during recent years,

amounts to roughly 1 lm2. Thus, the present

recording density reaches a density of the order of1 Gb/cm2. The super-paramagnetic limit, deter-

mining the final minimum size, is expected some-

where below a storage limit of 10 Gb/cm2.

These geometrical limitations do not apply to

focused ion beams obtainable from gas-field and

liquid-metal ion sources, GFIS and LMIS,

respectively. In particular, the optical properties of

a GFIS equipped with a supertip, GFIS�, promisebeam diameters in the nm-range [1]. Fig. 1 shows

calculated image spot diameters for Ar(GFIS�)

and Ga(LMIS) beams using the characteristic

beam figures of Table 1. The resulting curves

indicate that chromatic aberrations, resulting from

relative energy spreads of dE=E ¼ 10, 100 and

1000 ppm, control the minimum image spot

diameter. Thus, a figure of d � 10–100 nm iswithin reach for GFIS� beam energies of 10–100

keV energy. Due to lower figures of merit, the

corresponding LMIS values range up to 100 times

higher. These spot diameters, of d � 100–1000 nm,

can only be rendered useful for high-density data

recording by drastic aperturing with a corre-

sponding marked loss of beam intensity.

Recording speeds with Ne or Ar beams ofI � 10 nA are between 10 and 100 Mpix/s. The

pixel patterns, resulting from lateral range strag-

gling of the trajectories of heavy ions of E � 10–

100 keV, can be tailored to D ¼ 2ðRþ rÞ � 10–100

nm in diameter. An estimated lower limit of

D � 10 nm appears feasible [2]. Figs. 2(a) and (b)

present longitudinal and lateral ion penetration

depths for a variety of input parameters, calcu-lated with the SRIM 2003 computer code [3]. The

longitudinal extension of the implantation profile

Ll ¼ Lþ l, defined as projected range L plus

straggling l, is given as a function of mass ratio of

Fig. 1. Calculated image spot diameters D of Ar(GFIS�) and Ga(LMIS) beams versus convergence half-angle. Relative energy spreads

(indices): dE=E ¼ 10, 100 and 1000 ppm.

Table 1

Ion beam properties

GFIS� LMIS

Ions Hþ2 , rare gases, . . . Gaþ, special alloys

Energy E � 10–100 keV E � 10–100 keV

Energy spread dE � 1 eV dE � 10 eV

Virtual source size dv � 0:1 nm dv � 50 nm

Emission half-angle a � 0:5� a � 0:5�aEmitted current I � 10 nA I � 10 nA

Emittance e � 10�20 cm2 sr e � 10�14 cm2 sr

Spectral brightness b � 1012 A/cm2 srd eV b � 106 A/cm2 sr eV

Range parameters L, l, R, r � 10–100 nm

a Apertured beam; total Ga current into a � 30�: I � 1 lA.

S. Kalbitzer / Nucl. Instr. and Meth. in Phys. Res. B 218 (2004) 343–354 345

target/incident particle, l ¼ M2=M1, for diamond

(CD) and silicon targets (Fig. 2(a)). Due to thehigher atomic density of CD, s ¼ 3:51 g/cm3, as

compared to Si with s ¼ 2:32 g/cm3, all range

parameters are reduced. This is also seen from Fig.

2(b), where the corresponding lateral figures, D,

are shown. Thus, for fixed incident energies, the

highest pixel density can be obtained with high

density recording material and low mass ratios l.

Thus, for image spots of D � 100 nm, a pixeldensity of 10 Gpix/cm2 can be realized. Higher

pixel densities, up to 1 Tpix/cm2, appear physically

possible but pose reading problems. Standard

optical techniques are based on the Abb�e dif-

fraction limit of about one wavelength. Thus,sub-wavelength structures cannot be read with

presently available optical equipment, but require

a different approach, for example the application

of near-field optics.

2.3. Data lifetime

Data lifetime of presently employed optical andmagnetic recordings is estimated as of the order of

10 years [4]. Radiation defects, due to the intro-

duction of atomic disorder by heavy ions, can be

Fig. 2. (a), (b) Longitudinal and lateral range parameters, respectively, in CD and Si for heavy ions of 30 keV energy. Distributions,

extending to roughly 10–100 nm, depend on substrate density and mass ratio of target atom to ion species, l ¼ M2=M1.


much more stable. Phase transitions from the

crystalline up to the amorphous state can be in-

duced with strong changes of the optical proper-

ties. Ideally, the irradiated regions of the initially

transparent layer turn opaque. As will be shown insome detail below, these patterns are extremely

stable in a variety of solids because of high binding

energies. These thermodynamic properties guar-

antee a practically unlimited lifetime of the infor-

mation stored in form of optical contrast between

the crystalline and amorphous material modifica-tions.


3. Realisation of permanent archival storage

3.1. Generation of optical contrast by ionography

It is useful to describe the principle of ionog-

raphy, the writing of patterns with ion beams into

solids, in somewhat more detail.

GFIS� systems deliver very intense focused

beams to the target position. GFIS� spectral

brightness, b ¼ B=dE, the key figure of merit, ex-

ceeds those of other sources by many orders ofmagnitude. Thus, at present, a commercially

available GFIS� would be by far the best novel

�pencil�.Optically transparent materials, such as crys-

talline insulators and certain semiconductors,

constitute a novel �paper�. Ideally, they have an

empty band-gap of the order of some eV, so that

visible light can pass thin layers without or withonly little attenuation. Upon irradiation with

massive particles, lattice constituents are displaced

from their regular sites. As a consequence, the

created Frenkel pairs, each consisting of a vacancy

and an interstitial atom, produce electronic energy

levels inside the previously empty energy gap. It is

Fig. 3. Experimental amorphisation-energy densities for a variet

these additional levels which give rise to the

absorption of photons of sub-band-gap energies.

Obviously, with increasing irradiation level more

defects are created. Finally, saturation occurs at alevel of maximum lattice disorder, or amorphicity,

and thus also of opacity.

Particularly, in crystals of covalent bonding this

process is most effective so that a transition from

the crystalline to the amorphous phase can be ef-

fected easily. Elemental semiconductors, such as

group IV materials, are most favourable. Due to

their 100% covalent bonding neighboring atomscan adjust their bonds to the individual displace-

ments within a relatively wide angular range

accompanied by weakening bond energies. The

defect migration energies are relatively high, so

that annealing effects at room temperature can be

tolerated. Irradiation at cryogenic temperature

minimises recombination of defect species. Thus,

as compared to other solids, the required ion flu-ences are lowest, and hence the irradiation times

shortest. Fig. 3 shows this trend as a function of

covalency as derived from literature [5]. Ge, Si and

C (diamond) require the lowest overall energy in-

puts: the specific energies are of the order of some

y of insulating materials of different degrees of covalency.


10 eV/atom, varying with the ion-atom system.

The corresponding displacement efficiency, i.e.

energy/Frenkel pair, employing the fractional en-

ergy spent into nuclear collisions, en, can be ex-tracted from high-resolution ion backscattering

experiments and calculations using the SRIM code

[3]. Typical values are around en � 10 eV [6,7].

A schematic presentation of the intended io-

nographic process is given by Fig. 4 where also a

comparison with standard photo-lithography is

shown. [8] With silicon-on-sapphire, (SOS), com-

mercially available as hetero-epitaxially grown Sifilms 100–1000 nm thick on highly polished mono-

crystalline Al2O3 wafers up to �500 in diameter and

0.5 mm thick, the inscription process takes just one

step to form the optical pattern.

Fig. 4. Comparison of standard photo-lithography with iono-

graphy, the writing with a focused ion beam into a thin semi-

conductor film.

Fig. 5 is a macroscopic experimental demon-

stration of the pattern generating principle. Ar

ions of 50 keV energy were implanted into a thin

film of 400 nm Si on a 300 sapphire substrate [9].The sapphire single-crystal substrate transmits

visible light up to hm � 6 eV with about 80% of the

incident intensity.

As Fig. 5 further demonstrates, the optical

contrast between the irradiated and the non-irra-

diated parts of the Si layer sharpens with increased

fluence. Saturation of optical contrast is reached at

a level of about 1 · 1014 Ar/cm2, where a phasetransition to the amorphous state occurs. Evi-

dently, the transmission contrast between crystal-

line and amorphous regions can be used for both

analog and digital information storage. Moreover,

if light detection is sufficiently sensitive, even a

grey scale could be established according to the

applied irradiation level. In this way, even higher

information densities appear feasible.A quantitative investigation of radiation dam-

age on the optical properties of a Si film is pre-

sented by Fig. 6. Here optical absorption is shown

as a function of photon energy with Ne fluence as

parameter [10]. The absorption coefficient between

zero and saturation fluence changes by about 2

orders of magnitude over a wide region of photon

energies. Thus, for example, at hm � 2:5 eV theabsorption for a layer of amorphous Si of 100 nm

thickness amounts to Aa � 95%, whereas the

crystalline absorption is only about Ac � 10%. The

corresponding contrast figure is 90%. We note that

at photon energies of hm > 3 eV, with the onset of

strong direct transitions in c-Si, the contrast van-

ishes. Other heavy ion species, such as Ar and Kr,

yield similar results in terms of optical contrast[9,10].

Finally, we mention that also the reflection

mode can be used for reading ionographic data,

with contrast figures of the order of 10% [11].

Present-day optical disks are read in reflection

mode using similar differences in reflected light

amplitude or polarisation.

As mentioned in context with Fig. 2, the range-parameter set depends on energy and atomic

properties of the incident ion, and on atomic

number and density of the target material,

P ðE;A1;A2; sÞ. In particular, for a given material,

Fig. 6. Absorption coefficient of Si versus photon energy;

parameter: Ne fluence. Amorphisation fluence: 1015 Ne/cm2.

Fig. 5. Optical contrast of ionographic patterns in transmission of a 400 nm crystalline Si film on a 300 sapphire substrate. Fluences (top

to bottom): 3, 1, 0.3, 0.1, 0.03 and 0.01· 1014 Arþ/cm2; energy: 50 keV.


the spot size varies with ion energy and mass ratio

l. Thus, for heavier ions with l 1, at

E ¼ const:, the range parameters decrease consid-

erably and concomitantly the pixel dimensions.

Fig. 7 shows the lateral spread of a 10 keV Ar

beam including the recoil atom distribution in

diamond [3]. Evidently, the outer damage-spot

diameter stays within 20 nm with a smaller half-width around 10 nm. The ion-range parameter set

is: L ¼ 8:1, l ¼ 2:3, R ¼ 2:5, r ¼ 1:5 nm. It follows

that D ¼ 8:0 nm. With an absorption coefficient of

a � 1 · 106 cm�1 the optical transmission contrast

due to Lþ l � 10:4 nm would amount to about

50%. Even smaller spots can be produced by

reduction of the implantation energy to a limit set

by the optical contrast.Fig. 8 presents a model of the density-of-states

distribution for the crystalline and amorphous

phase of an insulator [12]. The radiation effects on

the matrix material are characterised by the

introduction of defect states deep in the gap, due

to displaced lattice constituents, and of disorder

states near the band edges, due to changes in bond

length and angle in neighbouring atoms.Experimental and calculated data for the

absorption of diamond, CD, show the following

features. In the energy interval of hm � 1–5 eV the

absorption coefficient of crystalline CD is

anon-irr � 1 cm�1, whereas that of CD, irradiated

with a sub-amorphous fluence of 3 · 1014 C/cm2,

varies between airr � 3 · 102 and 3 · 104 cm�1 [12].

Absorption coefficients reach maximum values ofa � 1 · 106 cm�1 at hm � 6 eV for fluences around

3 · 1015 C/cm2. Application of heavier ions yields

Fig. 7. SRIM 2003 calculation of the lateral recoil cascades of a beam of 10 keV Ar ! CD.

Fig. 8. Density-of-states model of irradiated insulators:

vb¼ valence band, vt¼ valence-band tail, Dþ;0;� ¼ positive,

neutral, negative defect centres, ct¼ conduction-band tail,

cb¼ conduction band. Note: in crystalline material only vb and

cb states exist.


lower amorphisation threshold fluences and smal-

ler pixel diameters. The lower atomic number and

the higher density of CD than that of Si reduce

both scattering angles and ranges, so that in this

respect CD is an ideal target material for obtainingsmallest ionographic patterns.

An overview on the changes in the absorption

coefficients of Si, SiC and CD is given in Fig. 9 [13].

Sufficient optical contrast can be realised with thin

Fig. 9. Amorphous and crystalline absorption coefficients of Si,

SiC and CD. The working range due to optical contrast between

the c- and a-phases is below the merger of the respective curves.


layers of all these semiconductors. We emphasise,

however, that with increasing band-gap the

working range of the respective storage medium is

shifted to higher photon energies. This latterproperty allows a better optical resolution with

CD, and thus a recording density increased by a

factor of about 4 for photons of 5 eV over photons

of 2.5 eV.

Fig. 10. Recrystallisation of ion-bombarded Si with annealing

temperature monitored by RBS/channelling. Parameter: ion

fluence.

3.2. Stability of ionographic patterns

Crystalline Si, SiC and CD are quite immuneagainst the usual environmental agencies. CD and

SiC are considered even for use under adverse

chemical and thermal conditions. Studies of their

amorphous phases have shown that these are also

quite stable. Fig. 10 shows, by example of Si, that

the respective activation energies are high enough

as to prevent any recrystallisation for millions or

billions of years at normal ambient temperatures[14]. Thus, contrary to the presently used archival

recording techniques involving corrosive materi-

als, no rewriting will become necessary as data

lifetime is practically unlimited.

Fig. 11. Atomic-force microscopy of irradiated and non-irradiated

energy.

CD in particular can be rendered graphitic,

which is thermodynamically even more stable than

diamond, so that a truly irreversible phase transi-

tion can be achieved. This transition, however, isaccompanied by swelling effects as is seen from

Fig. 11, where atomic force microscopy was used

areas of diamond samples. Fluence: 3 · 1015 C/cm2 of 75 keV

Fig. 12. Ionographic generation of optical contrast in a

monocrystalline Si film on sapphire. Opaque/irradiated and

transparent/non-irradiated regions form the pattern. Irradia-

tion: 20 keV Ne beam.

Fig. 13. Crystalline-amorphous optical contrast in polycrys-

talline SiC on sapphire. Irradiation: 20 keV Ne beam.


to demonstrate the expansion of the irradiated

regions of CD [7]. The difference in specific weight

of sðCDÞ ¼ 3:51 g/cm3 and sðCGÞ ¼ 2:25 g/cm3

explains the magnitude of the density change of a100 nm thick layer by about 15–20 nm for fluences

of (3–10) · 1015 C/cm2 and elevated annealing

temperatures. This transition should be avoided, if

optical near-field technology will be involved in

printing or reading. On the other hand, these

topological features would be useful for reading

techniques based on the atomic force principle

allowing to resolve pixels of the atomic size level.Since solid diamonds, small and expensive, cannot

be used in practice, the design task would be to

produce a thin high quality diamond thin film on a

thick insulating material, such as sapphire or

quartz, with strong adherence in order to avoid

separation of the two-layer system over long time

periods.

3.3. Crystalline/amorphous patterns

In order to demonstrate the feasibility of iono-

graphic recording in group IV materials we show

the following patterns.

Fig. 12 is a micrograph of the EC logo written

into a 100 nm Si film hetero-epitaxially grown on a

sapphire single-crystal substrate [14]. One sees thatfeatures of less than 1 lm are realised. The Ne

beam of 20 keV energy is broadened by mechani-

cal vibrations to the order of 100 nm.

The checkerboard pattern of Fig. 13 was pro-

duced by exposing a poly-crystalline layer of SiC of

100 nm thickness to a beam of 20 keV Ne ions [14].

The continuous line at the bottom of the graph, a

fraction of 1 lm thick, represents the multiplyscanned beam. The fine structure at the peripheries

of the pattern segment are about 100 nm wide.

Fig. 14 was obtained by irradiating polycrys-

talline CD with C ions through a metal grid of 12.6

lm periodicity [14]. The pattern, produced by a

fluence of 1 · 1015 C/cm2 at 65 keV, clearly indi-

cates that also CD is a useful choice for optical

recording purposes. It remains to be seen, how-ever, whether large-area diamond thin films

deposited by glow-discharge plasma techniques

would be preferred over the already existing high-

quality SOS material.

3.4. Ionographic data recording/copying

A simple digital recording pattern is shown in

Fig. 15 [14]. Here crystalline spots denote the digit

Fig. 14. Pixels generated by irradiating a crystalline diamond

through a metal grid. Fluence: 1· 1015 C/cm2 at 65 keV. Pho-

tograph taken with an optical microscope under incandescent

illumination. Magnification: �1400x.

Fig. 15. Principle of ionographic digital recording: irradiated/

opaque and non-irradiated/transparent regions form optical

contrast for encoding of ‘‘0’’ and ‘‘1’’.


0, amorphous spots the digit 1. Using a conser-

vative pixel size of 100 nm, accessible to opti-

cal reading with UV light, 1010 pix/cm2 can

be recorded so that a 100 cm2 disk carries 1 Tbit

data.

As mentioned before, pixel sizes around 10 nm

may mark the borderline for ionographic record-ing. This would allow data storage densities

around 1 Tbit/cm2 or a total amount of 100 Tbit

on a single disk. Reading would require proximity

techniques, be it optical near-field microscopy for

flat surfaces, e.g. as obtained with silicon films, or

atomic force microscopy in case of a structuredtopology, e.g. as obtained with diamond films.

Copying can be realised by contact lithography

which is an inverse process to optical near-field

reading [14]. Since, in contrast to reading, it is a

parallel process, copying would be very fast.

4. Conclusions

The proposed ionographic recording technol-

ogy is based on group IV insulators. High-quality

silicon-on-sapphire material has been available

with diameters of up to about 500, the size of

CD-ROM�s. The data life-time is estimated as

practically infinite. Diamond films on sapphire,

promising ultra-high recording densities, are underinvestigation.

The writing speed with intense focused beams

from a gas field ion source is at least 10 Mpix/s,

but may be increased to 100 Mpix/s. Thus, the

writing of 10 Gpix/cm2 will take about 103 s.

The problem with ionographic features of sub-

wavelength size is of course the reading process.

Although proximity techniques look promis-ing, there is not yet a viable technology available.

References

[1] S. Kalbitzer, Nucl. Instr. and Meth. B 150 (1999) 53.

[2] S. Kalbitzer, Appl. Phys. A 72 (2001) 639;

S. Kalbitzer, Ch. Klatt, Section II, p. 641.

[3] J.F. Ziegler, J.P. Biersack, Computer Code SRIM 2003.

[4] www.interpares.org/documents/ptf_draft_final_report.pdf–

Last Updated 1 Feb 02.

[5] P.D. Townsend, P.J. Chandler, L. Zhang, Optical Effects of

Ion Implantation, Cambridge 1994, p. 24.

[6] S. Kalbitzer, Appl. Phys. A 72 (2001) 639, in: S. Kalbitzer,

E. Friedland, Section V, p. 668.


E. Friedland, T. Hauser, Ch. Klatt, C.M. Demanet,

Section IV.D, p. 664.

[8] S. Kalbitzer, VDI Berichte 666 (1987) 71.

[9] S. Kalbitzer, Appl. Phys. A 44 (1987) 153.

[10] K.L. Bhatia, W. Kr€atschmer, S. Kalbitzer, Appl. Phys. A

45 (1987) 69.

http://www.interpares.org/documents/ptf_draft_final_report.pdf


[11] D.J. Brink, Physics Department, University of Pretoria,

private communication.


S. Kalbitzer, G. M€uller, L. Voglmeier, H. M€oller, Section

IV.B, p. 653.

[13] S. Kalbitzer, 9th CIMTEC – World Forum on New

Materials, in: P. Vincencini, S. Valeri (Eds.), Symposium

II: Surface and Near-Surface Analysis of Materials, Techna

Srl, 1999, p. 139.

[14] S. Kalbitzer, COSSMS 6 (2002) 271.

novel concepts for mass storage of archival data

Documents