Venables J. Introduction to Surface and Thin Film Processes

Подождите немного. Документ загружается.

good enough for magnetic ﬁeld sensors, and read/record heads using GMR multilay-

ers are already commercially available.

In the above example, the nonmagnetic spacer layers (of Cu) were suﬃciently thick

to ensure that the magnetism of the Co and NiFe layers are not coupled directly via

the conduction electrons. This is in contrast to the much thinner spacer layers discussed

in section 6.3.4 and illustrated in ﬁgure 6.24. At intermediate thicknesses, this spin-

dependent coupling leads to complex magnetization curves, and the transfer of spin

‘information’ from one part of a device to another, which can be used in spin valves or

spin transistors (Parkin et al. 1991, Johnson 1993, 1996, Parkin 1994, Monsma et al.

1995). The next stage may be to make use of spin polarized currents induced by mag-

netic elements into the substrate itself, and then to use these currents as the injector for

a hot electron device (Gregg et al. 1996). The acronym for this UK-based development,

SPICE (spin polarized injection current emitter), may have something to do with the

existence of a popular all-female vocal group of the same name at around the same

time. We shall see what becomes of both.

Another possible way forward, not involving magnetic coupling through the sub-

strate, is to use magnetic wires grown on insulators; these wires have anisotropic mag-

netic properties, as well as being a more favorable geometry for CPP GMR

measurements. NaCl(110) is a substrate with a high surface energy, and self-organizes

into facets on (001) planes at 45° to the substrate plane. By deposition at a shallow

angle, narrow wires will be produced at the tops of the ridges shown in ﬁgure 8.18(a),

and these wires can then be capped to prevent oxidation, etc. In a series of experiments,

Sugawara et al. (1997) ﬁrst deposited a thin SiO layer on either NaCl(110) or (111), fol-

lowed by Fe deposition at a shallow angle, followed by a further SiO layer.

This procedure allowed them to produce isolated islands aligned in one dimension,

continuous parallel wires, or isolated Fe dots of various sizes. They could then remove

the SiO/Fe/SiO assembly by dissolution of the substrate for TEM examination, and to

make particle and wire density observations, exactly as described here in section 5.3, as

shown in ﬁgure 8.18(b). The new feature is of course the ability to perform magnetic

and magneto-optical (Kerr) measurements before this stage, similar to that shown here

in ﬁgure 6.18, and hence explore magnetic anisotropy, dipole couplings and the para-

magnetic to ferromagnetic transition as a function of particle size (Sugawara &

Scheinfein 1997). There are a large range of parameters to explore, just within this one

system, if anyone wants to take these results to the next stage of implementation as a

working device.

Another system which clearly shows promise as a magneto-optical device is based

on the nucleation of Co dots on Au(111), at the position of the surface vacancies

which occur at the intersection of the (2331) herringbone reconstruction, ﬁrst

observed by Voigtländer et al. (1991) and described here in section 5.5.3. A strong

Kerr eﬀect signal from ML deposits in these dots has been observed by Takashita et

al. (1996), and Fruchart et al. (1999) have constructed well ordered Co pillars in

Co/Au (111) multilayers, with improved magnetic properties. Whether or not this

system will end up in a real device is not clear: are we ready to use surface reconstruc-

tions and surface point defects so directly in a manufacturing process? Whatever the

8.3 Conduction processes in thin ﬁlm devices 287

288 8 Surface processes in thin ﬁlm devices

Figure 8.18. (a) Shadow deposition at glancing angle onto ridged substrates to produce Fe

dots and, at higher coverage, nanowires; TEM pictures at lower (b) and higher (c)

magniﬁcation, showing the length and width of the Fe wires in the SiO/Fe/SiO assembly (from

Sugawara et al. 1997, reproduced with permission).

(100)

(010)

NaCl (110)

substrate

[110]

[001]

[110]

a-SiO

SiO deposition

20º

Fe flux

Fe dots

(a)

(b)

(c)

answer, these surface processes can be responsible for observed structures with lateral

length scales in the 10 nm range.

8.4 Chemical routes to manufacturing

Although this book, including this chapter, has been written largely from the perspec-

tive of materials physics, materials-oriented chemistry plays an enormous role in device

development. Chemists are also being drawn into the ﬁeld of combinatorial materials

discovery, a practically based mix of combinatorial chemistry and thin ﬁlm deposition

techniques, used to search for new compounds and compositions. Such developments

are discussed brieﬂy in this section.

8.4.1 Synthetic chemistry and manufacturing: the case of Si–Ge–C

The growth of Ge and SiGe alloys on Si(001) was considered in some detail in section

7.3.3. It is the simplest example of strained layer growth, and we described how the low

mismatch alloys can be grown as layers suitable for heterostructures, or the strain can

induce islands to form, which may (or may not) have potential as quantum dots. But

do we want islands, or would layers be better? The answer, of course, depends on the

application envisaged, but in general layers are preferable. So let’s try mixing in another

element, which relieves the strain and so promotes layer growth. This is the motivation

behind research on the Si–Ge–C system; by mixing in the right amount of carbon one

can hope to compensate for the strain introduced by Ge.

But how to do it? CVD is the most widely used manufacturing technique, starting

from silane (SiH

) and germane (GeH

), so mix in some methane (CH

) and see what

happens. But some knowledge of thermochemistry is required, and this tells us that

methane is much more stable than both silane and than germane, which means that

does not want to break up. Moreover, the solubility of C in Si or Ge is very low

(⬃ 1310

% in equilibrium at the Si melting point, and even lower in Ge), so the

carbon which does form is likely to be in the form of SiC, which is not what we want;

in practice up to 1–2% has been incorporated at SiGe CVD process temperatures

⬃ 700 °C. Somehow we have to trick the system, thermodynamically and kinetically.

This has been done by creating custom-built molecular precursors to use as the source

material.

One group in Arizona (Kouvetakis et al. 1998a) has been developing compounds

with an inbuilt tetrahedral arrangement involving one C atom surrounded by four SiH

or GeH

ligands, as illustrated in ﬁgure 8.19(a). These are sizable van der Waals mole-

cules, which are liquids at room temperature and evaporate easily. They are also reac-

tive, in that they lose hydrogen easily, and hence can incorporate the tetrahedral units

C and Ge

C into a growing ﬁlm.

These molecules are crowded, and have bond lengths very diﬀerent from normal, as

indicated in the cluster calculation for Ge

C shown in ﬁgure 8.19(b). This is because

the small C atom is in the middle, and the much larger Si and especially Ge atoms are

8.4 Chemical routes to manufacturing 289

290 8 Surface processes in thin ﬁlm devices

Figure 8.19. (a) The precursor molecule (GeH

)

C, with Ge–C and other bond lengths

determined by gas phase electron diﬀraction; (b) structure of substitutional C in Ge and the

associated Ge–C and Ge–Ge bond lengths in a calculated (Ge)

C cluster (after Kouvetakis et

al. 1998a,b, reproduced with permission).

(a)

(b)

competing for space around it. However, this processing route suppresses the segrega-

tion of the C atoms, because they are always surrounded by Si or Ge, and Kouvetakis

et al. (1998b) have been successful in incorporating 5–6%C into ﬁlms grown at 470°C,

which have relatively few defects (and no SiC precipitates) as seen by TEM. Starting

molecules with carbon on the outside would not be nearly as eﬀective in this respect;

in other words, by using these special molecules and relatively low processing temper-

atures, carbon has been tricked to remain in solution.

Whether this work represents a real breakthrough or just a very interesting develop-

ment depends on the next stage – will the clever molecules be incorporated into actual

manufacturing processes or not? The role of relatively small contract research ﬁrms in

smoothing the path to manufacturing is interesting; such processes will certainly not

be incorporated into large scale Fab lines overnight. Increasingly, it is the equipment

manufacturers who incorporate new processes such as this one, in order to be able to

persuade the large scale producer (of e.g. computer chips) to adopt such technologies

when they reinvest the next few US$ billion.

8.4.2 Chemical routes to opto-electronics and/or nano-magnetics

Optoelectronics, the integration of light sensitive devices with micro-electronics, is

another huge ﬁeld in which surfaces and thin ﬁlms play a major role. One of the main

interests in the growth of III–V and II–VI compounds is related to the ability to inte-

grate such devices with silicon in the form of ‘band gap engineering’ as discussed in

sections 7.3.4 and 8.2.3. Quantum dots are a hot topic touched on in section 8.2.4,

where the intent is that uniform size and spacing can be achieved via self-organization.

But, as always in technology, one has always to bear in mind that the most eﬀective

solution to the original challenge or problem, might come from a completely diﬀerent

route. One such possibility is the assembly and self-organization of ordered arrays of

colloidal particles prepared by more or less traditional wet chemistry methods – a ﬂask

full of goo, a drying oven and a spray-gun may be all that is needed; I exaggerate, of

course, but not by much. Such techniques are not speciﬁc to optoelectronics; magnetic

particles can just as well be prepared and arranged in remarkably uniform arrays, as

shown in ﬁgure 8.20.

Here the idea is to prepare II–VI materials such as CdSe in solutions containing

additives which adsorb on the surface of colloidal crystals in the size range 5–10 nm,

thus preventing further growth. These colloids, coated with self-assembled monolayers

(SAMs), form a stable dispersion in solution, which can be made to crystallize out by

gentle evaporation of the lighter component in the solution (e.g. octane from octanol

at 80 °C). Further warming removes all but a few ML of the additives, to the extent that

3D colloidal superlattice crystals can be grown with sizes up to 50 mm (Murray et al.

1995, Heath 1995, Collier et al. 1998).

The size distribution is amazingly narrow, as shown in ﬁgure 8.20, and can be further

controlled by manipulation of the supersaturation as a function of time during growth.

8.4 Chemical routes to manufacturing 291

Financiers should note that the US billion is used here which is only 10

, rather than the UK 10

292 8 Surface processes in thin ﬁlm devices

Figure 8.20. (a) 3D Optical superlattice formed from colloidal crystals of CdSe, spacing

6.5 nm, diameter 5 nm (similar to Murray et al. 1995); (b) 2D magnetic superlattice of Ag

coated Co particles, spacing 13 nm, diameter 8 nm formed by colloidal techniques (after

Murray 1999, both ﬁgures reproduced with permission). In both these cases the 2D hexagonal

lattice extends over much larger distances than can be eﬀectively portrayed here.

(a)

(b)

Other clever tricks are to cap these dots with a compatible material such as CdS having

a higher refractive index or to introduce relevant dye sensitizers onto the surface of the

dots. This latter technique has been a key ingredient in the color photographic pro-

cesses using AgBr crystals for a long time. By the former means the conﬁnement of the

exciton wavefunction can be increased, and so produce stronger photoluminescence,

quantum yields .50% having been demonstrated (Peng et al. 1997, Schlamp et al.

1997). However, until now the overall electro-luminescent energy eﬃciency has been

low, and the devices are not yet stable enough for production (Peng et al. 1997,

Alivisatos 1998, Collier et al. 1998).

A single electron transistor which functions at T54.2 K has been demonstrated,

simply by sprinkling individual colloidal dots of CdSe across the gate region of an FET

(Klein et al. 1997). As a way forward, you might take this demonstration ‘with a pinch

of salt’, but it is certainly spectacular. One of the arguments in favor of these colloidal

crystals is their intrinsic cheapness, yet if there is no way of getting current in or out,

or if they only work at low temperature, then we have just an intriguing demonstration,

not yet an innovation. Another impressive demonstration is an FET which works at

room temperature, made by dropping a single carbon nanotube across the source-drain

gap (Tans et al. 1998a), and the observation of associated Coulomb blockade phenom-

ena (Tans et al. 1998b).

8.4.3 Nanotubes and the future of ﬂat panel TV

A further example centers around chemical routes to the production of ﬁeld emission

sources for computer or TV screens. As discussed in section 6.2.2, various carbides and

nitrides have been researched over the years, but have never really been quite stable

enough to make a serious impression on the market. Yet the conventional TV tube is

the last remaining example of the vacuum triode in production; is it too destined for

oblivion? If ﬁeld emission could be made to work reliably in the planar geometry

shown in ﬁgure 6.13(b), then maybe it could be saved!

These thoughts have gained impetus from the discovery of both multiple and single-

walled carbon nanotubes, cylindrical intermediates between planar graphite and the

closed cage fullerenes (Iijima 1991, Ebbesen & Ajayan 1992, Ajayan & Ebbesen 1997,

Bernholc et al. 1997). The tubes grow as long ﬁlaments sticking out from the substrate,

and emit electrons from the ends. Filaments can be produced in bundles or matrix

arrays (Collins & Zettl 1997), which diﬀer in detail depending on the production

method. Using a plasma arc discharge, Terrones et al. (1998) have produced ﬁlaments

containing a transition metal carbide particle near the end. The carbide particle cata-

lytically converts C-containing compounds into tubes, almost as if it were knitting a

sock, as shown in ﬁgure 8.21. Several diﬀerent refractory carbides have been so encap-

sulated, and progress to date is reviewed by Terrones et al. (1999).

Nanotubes can join up in helical (chiral) as well as cylindrical geometries, p-n junc-

tions along the length have been demonstrated where the chirality changes, and they

also can be doped, or made, with various B–N–C mixtures. The case of BN has led to

some exquisitely delicate analytical microscopy, combining pictures analogous to

8.4 Chemical routes to manufacturing 293

ﬁgure 8.21 with compositional analysis via EELS on the nanometer scale (Golberg et

al. 1996, 1997). With this number of variables to explore, it seems only a matter of time

before practical displays can be devised to rival those already demonstrated for the

Spindt cathode and DLC geometries. But of course that does not mean that such a

display will necessarily be successful in the marketplace, where it has to compete with

all other ﬂat display technologies as discussed earlier in section 6.2.2, and/or ﬁnd its

own particular niche. The advantage is the high intrinsic brightness, which may make

it suitable for projection onto a large screen; the list of potential disadvantages will dis-

tinguish optimists from pessimists.

Note that in the above two examples the distinction between surfaces, thin ﬁlms and

molecules has all but disappeared: the molecule is all surface and is the thin ﬁlm. What

we are seeing here is an important stage in the development of single molecule elec-

tronics. The next stage is again the transition to manufacturing, which may well be a

rocky road, and test to the limit who is serious. As they say on TV: ‘don’t go away’, or

equivalently ‘watch this space’!

8.4.4 Combinatorial materials development and analysis

The combination of step-wise thin ﬁlm deposition processes with analysis on a micro-

scopic scale is coming to be known as combinatorial materials development. This series

of techniques uses automated deposition and annealing sequences to produce a series

of diﬀerent compositions, thicknesses and/or doping levels in a matrix pattern of dots

294 8 Surface processes in thin ﬁlm devices

Figure 8.21. High resolution TEM images of TaC particles in nanotubes where (a) the TaC

particles are encapsulated at tips, and (b) graphite fringes suggest epitaxial growth on the

particles (from Terrones et al. 1998, reproduced with permission).

distributed in x and y.Diﬀerent patterns can then be set up in several areas on the same

chip, and then analyzed by a microscopic technique to screen for the ‘best’ properties.

Typically, such techniques will be useful when one has to span a large range of param-

eters, and when theoretical models are of limited use. Stepper motors controlling stage

movements and multiple deposition shutters, with a resolution of a few micrometers

are now available more or less routinely (if expensively), which means that many thou-

sands of diﬀerent samples can be screened in a single experiment.

An example is the search for improved red phosphor materials to be used in ﬂat panel

displays (Danielson et al. 1997). The starting point materials were polycrystalline ternary

or quaternary oxide layers acting as hosts for ‘activator’ rare earth ions such as Eu or Ce.

Some 25000 individual compositions deposited onto a 3-inch wafer were sampled,

leading to a single best composition of Y

0.845

0.070

0.060

0.025

. This composition

was found to be as good as the existing commercial phosphors. It is clearly just a ques-

tion of time before superior thin ﬁlm materials are found using such techniques, which

originated in the pharmaceutical industry as a means of accelerated drug discovery.

The next stage may well be to combine such approaches with patterned substrates,

so that diﬀerent areas explore diﬀerent surface treatments or diﬀerent growth regimes.

A start in this direction, with emphasis on molecular recognition of metal and semi-

conductor nanocrystals, has been made by Vossmeyer et al. (1998), and patterned

arrays of bio-macromolecules have also been demonstrated. Flexible patterned sub-

strates are also being produced via soft lithography, micro-contact printing and related

techniques (Xia & Whitesides 1998). Such approaches involving microarrays of DNA

are centrally involved in the future of the human genome project (DeRisi et al. 1997).

The huge interest in microbiology means that such experiments, including seletively

tagging the colloidal nanoparticles described in the section 8.4.2, are achieving wide-

spread recognition in the materials community (Mirkin et al. 1996, 1999), even if, or

perhaps because, most of the rest of us are very early on the learning curve. But we can

all recognize the implied potential of the ﬁeld.

All this suggests that we should consider creating patterned nests for microbes, so that

we can then sit back and let them do our work for us. Indeed Richard Feynman was ﬁrst

with this suggestion in a famous lecture in 1959 entitled ‘There’s plenty of room at the

bottom’, republished as Feynman (1992) in a new Journal of Microelectromechanical

Systems. I’ll bet that H.G. Wells had the basic idea well before that – is there anything

really new? But this is getting dangerously close to futurology, the proper business of

the twenty-ﬁrst century, not the twentieth. As this chapter is ﬁnalized, in December

1999, the decoding of chromosome 22 made headline news, special millennium issues

of the journals arrived, and I started to read articles entitled ‘The Net Century’, etc. A

few words on such topics and the educational/training implications in the short ﬁnal

chapter, and I’m done. The twenty-ﬁrst century is essentially yours: good luck!