Venables J. Introduction to Surface and Thin Film Processes

Подождите немного. Документ загружается.

lasers. Laser ablation, sometimes called pulsed laser deposition (PLD), has typically

been used to deposit ceramic materials, including high temperature superconductors

(Dijkamp et al. 1987, Venkatasan et al. 1988, Morimoto & Shimizu 1995); it produces

very rapid deposition in which whole chunks of material can break oﬀ and be depos-

ited during the immense peak powers which typically last for 10–20 ns. A particular

advantage of rapid evaporation is the control of stoichiometry, since the diﬀerent

species do not have time to segregate to the surface during the evaporation phase.

2.5.3 Molecular beam epitaxy and related methods

The large scale use of such evaporation sources, especially for depositing semiconduc-

tor or metallic multilayers, has become known as molecular beam epitaxy (MBE). This

acroymn has spawned several sub-cases, such as GSMBE (Gas Source MBE), and

MOMBE (Metal–Organic MBE) which is sometimes called CBE (chemical beam

epitaxy). Thus MBE really spans a range of techniques which range from being merely

a fancy name for thermal evaporation, to much more chemically oriented ﬂow process

techniques such as chemical vapor deposition (CVD) and MOCVD which we describe

brieﬂy in section 2.5.5. MBE is covered in several books (e.g. Parker 1985, Tsao 1993,

Glocker & Shah 1995) and in many review articles and ongoing conference series. Only

a few general points can usefully be made here, but this preparation method lies behind

much of the science discussed later in chapters 5–8.

The growth end of such a chamber used for GSMBE or MOMBE is shown sche-

matically in ﬁgure 2.8 (Abernathy 1995). The halfspace in front of the substrate is a

bank of eﬀusion cells and various forms of gas injector cells, which transport reacting

chemicals to the substrate. The growth of III–V compounds such as GaAs or AlGaAs

has been pursued using such techniques, starting from metal–alkyl compounds such as

triethylaluminum (TEA) and triethylgallium (TEG) and hydrides such as AsH

. The

group ﬁve hydrides are injected via cracker cells which convert them catalytically into

and hydrogen, which then impinge on the substrate to react with the alkyl beam to

produce the growing ﬁlm (Panish & Sumski 1984, Abernathy 1995).

These cells are surrounded by liquid nitrogen cooled shrouds, which both condense

unwanted evaporants and improve the vacuum in the sample region, which is monitored

by the mass spectrometer. One of the advantages of MBE and related methods is the rel-

ative ease by which in situ diagnostic tools can be incorporated into the vacuum system.

The most widely used are RHEED, shown in ﬁgure 2.8 and described in section 3.2.2,

plus various optical techniques, some of which can also work in higher pressure environ-

ments. A compendium of these ‘real-time diagnostics’, particularly in the context of

semiconductor growth, is given by several authors in Glocker & Shah (1995, part D).

2.5.4 Sputtering and ion beam assisted deposition

There are many uses for ions in connection with the production of thin ﬁlms.

Sputtering using relatively low energy (100 eV–2 keV) ions is often used for cleaning

samples, while higher energy (5–200 keV) ions can be used for doping layers with

2.5 Thin ﬁlm deposition procedures 57

electrically active impurities. Ions can also be used for deposition, where individual

ions, or charged clusters can be deposited at a range of energies. Clearly, the fact that

the ions are charged allows extra control, and there are various methods by which this

can be done. Directed ion beams from an accelerator form the most obvious possibil-

ity, but plasma and magnetron sources are also widely used (Bunshah 1991, Vossen &

Kern 1991, Shah 1995, Smith 1995).

One of the main claims for IBAD (Ion Beam Assisted Deposition) or IBSD (Ion

Beam Sputter Deposition) is that better quality deposits can be obtained at lower sub-

strate temperatures, thus avoiding large scale interdiﬀusion which results from high tem-

perature processing. However, the deposited ﬁlms adhere well to the substrate because

of the localized limited mixing caused by the ion impacts (Itoh 1989, 1995, Marton

1994). This is one example of device engineers trying to reduce the ‘thermal budget’ and

so produce sharper interfaces between dissimilar components. Another possibilty is to

produce clusters in a supersonic expansion source, and to ionize them, controlling their

ﬂight during cluster depostion with applied voltages. This has been termed ICB (Ionized

Cluster Beam) technology by the inventers (Takagi 1986, Takagi & Yamada 1989). Yet

another possibility is to use ion beams to react with the substrate and grow compounds

such as oxides or nitrides at and near surfaces (Herbots et al. 1994).

All of these procedures are inherently complex; for almost all purposes the question

is whether they produce ‘better’ ﬁlms for particular applications, i.e. whether they give

a large enough improvement over existing methods to justify the considerable invest-

ment involved. Although our aim here is to give an outline description, we will not

pursue models of how of these methods work in any detail; they are all rather speciﬁc

to the combination of material and ion beam technique, and whether the processes are

reactive, in the sense of involving (ion) chemistry, or physical, meaning that only colli-

sions and clustering are involved. Nonetheless, ion beam assisted methods are very

58 2 Surfaces in vacuum

Figure 2.8. GSMBE or MOMBE growth chamber containing eﬀusion and cracker cells, and

RHEED diagnostics (after Abernathy 1995, reproduced with permission). See text for

discussion.

widespread and are of great economic importance; sputter deposition in particular has

high throughput for the production of (textured) polycrystalline thin ﬁlms of a wide

variety of materials (Bunshah 1991, Vossen & Kern 1991, Shah 1995). The language

used to describe such ion based processes (see Greene 1991) necessarily starts from a

similar vantage point to that used to describe thermal evaporation; the latter topic is

treated here in detail in chapter 5.

2.5.5 Chemical vapor deposition techniques

The growth of semiconductor layers for device production is most frequently done by a

variant of CVD (Chemical Vapor Deposition). This is usually implemented as a ﬂow

technique in which the reacting gases pass over a heated substrate, as indicated in ﬁgure

2.9(a) for growth of silicon from silane and hydrogen. CVD reactors are either of the hot

wall or cold wall type, and are surrounded by an extended gas handling and pumping

system with rigorous control of the (often highly toxic) gases (Carlsson 1991, Vescan

1995). Because the pressure may be up to an atmosphere in certain cases (APCVD:

Atmospheric Pressure CVD) the growth rate can be high, and UHV technology is not

an absolute necessity; but control of impurities is a dominant problem, as pointed out

in section 2.4.4 (O’Hanlon 1994). Thus UHV-CVD is becoming more widespread, where

the total pressure is , 1 mbar. However, most commercial processing corresponds to

2.5 Thin ﬁlm deposition procedures 59

Figure 2.9. (a) Schematic diagram of reactions involved in the CVD of silicon from silane

(SiH

) and hydrogen (H

); (b) typical variation of the growth rate on gas velocity with rate

limiting steps indicated (after Vescan 1995, reproduced with permission).

gas velocitygas velocity

Growth rateGrowth rate

thermodynamics

surface

kinetics

surface

kinetics

gas

transport

gas

transport

(b)

(a)

substrate

reactor wallreactor wall

input

output

SiH

gas phase reactionsgas phase reactions

diffusion

adsorption desorption

surface

processes

surface

processes

LPCVD (Low Pressure CVD) where the pressure is 0.1–1 mbar. MOCVD

(Metal–Organic CVD) or OMVPE (Organo–Metallic Vapor Phase Epitaxy) is also a

widely used technique, as is PECVD (Plasma Enhanced CVD); there are many variants

on this theme (Vossen & Kern 1991).

The question of reaction mechanisms and rate limiting steps in CVD is highly

complex. Under LPCVD conditions diﬀusion processes in the gas are typically not the

dominant eﬀect, so that at the growth temperatures, kinetic processes on the growing

surface are rate limiting, as indicated in ﬁgure 2.9(b) (Vescan 1995). However, we can

see from ﬁgure 2.9(a) that all the reactions, in the gas phase and on the surface of the

growing ﬁlm, are in series and that there is typically very little information on the inter-

mediate states of the reaction. Thus understanding CVD in atomic and molecular terms

is very much an ongoing research project, which we will return to later in section 7.3.

Further reading for chapter 2

Dushman, S. & J. Laﬀerty (1992) Scientiﬁc Foundations of Vacuum Technique (John

Wiley).

Glocker, D.A. & S.I. Shah (Eds) (1995) Handbook of Thin Film Process Technology

(Institute of Physics), especially parts A and B.

Hudson, J.B. (1992) Surface Science: an Introduction (Butterworth-Heinemann) chap-

ters 8 and 9.

Lüth, H. (1993/5) Surfaces and Interfaces of Solid Surfaces (3rd Edn, Springer) chap-

ters 1 and 2.

O’Hanlon, J.F. (1989) A Users Guide to Vacuum Technology (John Wiley).

Matthews, J.W. (Ed.) (1975) Epitaxial Growth, part A (Academic).

Moore, J.H., C.C. Davis & M.A. Coplan (1989) Building Scientiﬁc Apparatus (2nd Edn,

Addison-Wesley) chapters 3 and 5.

Roth, A. (1990) Vacuum Technology (3rd Edn, North-Holland).

Smith, D.L. (1995) Thin-Film Deposition: Principles and Practice (McGraw-Hill).

Tsao, J.Y. (1993) Materials Fundamentals of Molecular Beam Epitaxy (Academic).

Problems for chapter 2

These problems are to practice and test ideas about vacuum systems, design problems

and surface preparation techniques.

Problem 2.1. Design of vacuum systems for speciﬁc purposes

Use your knowledge of (and appendices on) conductances of standard size tubes, and

the characteristics of vacuum pumps, to suggest (and justify semi-quantitatively)

design choices in the following situations.

60 2 Surfaces in vacuum

(a) Pumping an approximately spherical chamber of diameter 0.5 m. The chamber is

to be let up to air infrequently, and we want to achieve as good a pressure as pos-

sible (, 10

210

mbar).

(b) Pumping a cylindrical chamber of length about 10 m and diameter 0.1 m. The

chamber is to be periodically ﬂooded with rare gases up to about 10

mbar pres-

sure, and the important point is to be able to achieve pressures below 10

mbar

quickly and economically.

about 50 m and diameter 0.1 m, with a total length in excess of 50 km (kilometers)

at a pressure of ,10

211

mbar.

Problem 2.2. Design of a Knudsen source for depositing elemental

metal ﬁlms

A Knudsen source is an evaporation furnace which relies on the establishment of the

vapor pressure above a solid or liquid source material. A small hole in the furnace

above the source material, plus collimating holes in front of the source, allow a beam

of the source material to be directed at the sample. Use your knowledge of vapor

pressures and kinetic theory to design a source which will deposit one monolayer per

minute on a sample held 0.15 m away from the exit of the source, will be uniform on

the sample within 1% for the central 0.01 m diameter, and will not deposit any

material on the sample outside a radius of 0.02 m. Do this in stages, with discussion,

as follows.

(a) Consider the formula R5 nv/4 for the number of atoms hitting unit area per second

of an enclosure, and how this formula applies to a Knudsen source. Derive

the formula by considering the relevant integrations over angles and the

Maxwell–Boltzmann velocity distribution.

(b) Consider the geometry of the design, and the constraints on the uniformity and

area of the deposit. Show that this will limit the size of the hole in the furnace, and

suggest a suitable size for holes in both the furnace and the collimator.

vapor pressure as a function of furnace temperature. Using the relationship

between density n and pressure p for this material, coupled with your design from

part (b) work out the temperature at which the source will have to operate to satisfy

the deposition rate requirement, explaining your assumptions.

(d) If you actually want to design a real source for this material, consider carefully the

materials of which the source can be made, whether you should be using a Knudsen

or some other type of source, and how to power the furnace to achieve suﬃcient

temperature uniformity, etc.

Note: short descriptions of most possible deposition techniques are given by Smith (1995)

and by Glocker & Shah (1995); some speciﬁc designs are in Yates (1997).

Problems for chapter 2 61

Problem 2.3. Some questions on surface preparation and related

techniques

Questions about surface preparation are always very speciﬁc to the materials con-

cerned, but here are a few which may be relevant and which spring from the text of this

section.

(a) Why should one either cool the sample slowly through a surface phase transition

(e.g. as in the case of Si(111)), or not anneal the sample above a bulk phase transi-

tion (e.g. in preparing b.c.c. Fe surfaces)?

(b) What is the main reason why Si(111) produces a 231 reconstruction after cleav-

age, when the equilibrium surface structure is the 737?

a device, e.g. by molecular beam epitaxy. Why is this precaution taken, and how

does it improve the quality of the devices grown on such surfaces?

(d) Mass spectrometry shows a range of mass numbers (M/e ratios) for the contents

of the vacuum system, but they don’t seem to be simply related to the molecules,

e.g. O

,CO,H

O, CO

which are present. What range of processes are respon-

sible for this discrepancy?

(e) GaAs often evaporates to leave small liquid Ga droplets on the surface. Why does

this happen, and how can it be prevented?

62 2 Surfaces in vacuum

3 Electron-based techniques for

examining surface and thin ﬁlm

processes

This book presumes that the reader is interested in experimental techniques for exam-

ining surface and thin ﬁlm processes; however, there are many books devoted to surface

physics and chemistry techniques, some of which are given as further reading at the

end of the chapter. There are even several books which are just about one technique,

such as Pendry (1974) or Clarke (1985), both on low energy electron diﬀraction

(LEED) in relation to surface crystallography. By the mid-1980s it was already stretch-

ing the limits of the review article format to compare the capabilities of the available

surface and thin ﬁlm techniques (Werner & Garten 1984).

Since then, the various sub-ﬁelds have proliferated, so we cannot be comprehensive,

or give all the latest references here. In section 3.1 we discuss ways of classifying the

large number of techniques which exist, and thereafter the chapter is restricted to tech-

niques based on the use of electron beams. Section 3.2 discusses the most widely used

(elastic scattering) diﬀraction techniques used for studying surface structure. Section

3.3 discusses forms of electron spectroscopy based on inelastic scattering, which are

used to obtain composition and chemical state information. Individuals can look in

more detail into a particular technique. Students have been asked to present a talk to

the class, followed by questions and discussion; some of the topics considered in this

way are listed, along with selected problems, at the end of the chapter. As examples,

some case studies are given in section 3.3 on Auger electron spectroscopy, in section 3.4

on quantiﬁcation of AES, and in section 3.5 on the development of secondary and

Auger electron microscopy. Stress is placed on the relationship between microscopy

and analysis: microstructure and microanalysis. The frontier is at nanostructures and

nanometer resolution analysis.

3.1 Classiﬁcation of surface and microscopy techniques

3.1.1 Surface techniques as scattering experiments

Most physics techniques can be classiﬁed as scattering experiments: a particle is inci-

dent on the sample, and another particle is detected after the interaction with the

sample. Surface physics is no exception: we can think of an incident probe and a

response as set out in table 3.1.

The probe will be formed from a particular type of particle, and typically will have

a well deﬁned energy E

, and often a well deﬁned wave vector k

, or equivalently

momentum p

5"k

. The response can either be the same or a diﬀerent particle, and,

depending on the detection system, its energy E and/or its wave vector (momentum)

k(p), and maybe other attributes, can be measured. If we understand the nature of the

scattering process, then we can interpret the experiment and deduce the corresponding

characteristics of the sample. It is easy to see from table 3.1 how the number of tech-

niques, and the corresponding acronyms, can be very large, especially once one realizes

that any probe particle can give rise to several responses, and that we may have diﬀerent

names for essentially the same technique used at diﬀerent energy, and diﬀerent wave-

vector (momentum or angular) regimes.

3.1.2 Reasons for surface sensitivity

The next question is ‘which techniques are actually useful for studying surfaces?’. There

are two cases. In the ﬁrst case, the emergent (i.e. response) particle or the probe parti-

cle has a short mean free path,

. This leads to a useful ‘single surface’ technique.

Examples are Auger electron spectroscopy (AES), where the emerging electrons in

the energy range 100–2000 eV have

for inelastic scattering in solids typically in the

range 0.5–2.5nm. Using an energy analyzer to measure only those Auger electrons

which have not lost energy, attenuates the signal from subsurface layers strongly. A

cruder form of energy discrimination is used in observing LEED patterns, where

both the incident and the emergent electrons have short mean free paths for energy

loss processes. In SIMS (secondary ion mass spectrometry), the emergent ions have

a very high probability of being neutralized if they do not originate very near the

surface. ICISS (impact collision ion scattering spectroscopy) is surface sensitive

because the incident ion will be neutralized, and thereby not detected, if the probe

particle penetrates the solid. An introduction to ion based techniques can be found

in Feldman & Mayer (1986), in Rivière (1990), and in several chapters contained in

Walls (1990); many other books can be unearthed via the web.

In the second case, the sample has a large surface to volume ratio. This condition

allows us to extract surface information from techniques which are not particularly

surface sensitive. We can perform heat capacity or other thermodynamic measure-

ments, or study structures and dynamics by X-ray or neutron scattering. Here we need

to know the signal from the bulk, and maybe subtract it in a diﬀerential measurement.

Much of physical chemistry work on surfaces has been done this way, on powdered, or

exfoliated, samples.

64 3 Electron-based techniques

Table 3.1. Particle scattering techniques

Electrons (E

, k

, . . . spin s) Electrons (E, k, . . . spin s)

Probe

{

Radiation (

, k

,...polarization)

Response

{

Radiation (

, k, . . . polarization)

Atoms (E

, k

, atomic number Z) Atoms (E, k, atomic number Z9)

Ions (E

, k

,Z, charge state 6n) Ions (E, k,Z9, charge state 6n9)

3.1.3 Microscopic examination of surfaces

Microscopy can be categorized into ﬁxed beam, scanned beam and scanned probe

techniques. A typical ﬁxed beam technique is transmission electron microscopy

(TEM); the same instrument can often be used for reﬂection electron microscopy

(REM). These techniques are illustrated schematically in ﬁgures 3.1(a) and (b).

Examples will be given later which show that it is not essential to have these instru-

ments operating at UHV in order to produce useful surface related information: UHV

experiments followed by ex situ examination can be very informative, provided the ﬁnal

samples are not too reactive in air.

The central element of TEM or REM is the objective lens, a cylindrically symmet-

ric strong magnetic ﬁeld positioned just after the sample. As the equivalent of the ﬁrst

stage of an ampliﬁer in electronics, this element is critical to the performance of the

microscope, and the aberrations and phase transfer characteristics of this lens deter-

mine both the resolution and contrast that are seen in the image. A particularly useful

feature is the use of the aperture situated in the back focal plane of the objective lens

3.1 Classiﬁcation of surface and microscopy techniques 65

Figure 3.1. Schematic geometries for TEM, STEM, SEM and REM (after Venables et al.

1987, redrawn with permission).

to select individual diﬀraction features and to relate these features to the image. In the

case of REM, the image is strongly foreshortened, but this does not mean that the

images are particularly diﬃcult to interpret, as we experience the same sort of image

foreshortening when we look ahead driving a car along the road. Both TEM and REM

have been able to image surface steps, particularly on high atomic number relatively

inert materials such as Au(111) and Pt(111), without requiring that the surfaces were

truly clean.

There are many books on electron microscopy, and TEM in particular has the rep-

utation for being diﬃcult to understand, primarily due to the need for a dynamical

theory of electron diﬀraction to interpret the images of crystalline samples. For an

overview of the ﬁeld, see Buseck et al. (1988), which includes a chapter on surfaces

(Yagi 1988); recent surveys of high resolution (HR)-TEM, describing the approach to

atomic resolution at surfaces and interfaces, are given by Smith (1997) and Spence

(1999), both with extensive references. I have attempted a ten-minute sketch of the

various techniques in Venables et al. (1987).

A few groups have converted their instruments to, or constructed instruments for,

UHV operation, and in situ experiments. These instruments, which can also be used for

transmission high energy electron diﬀraction (THEED) and reﬂection high energy

electron diﬀraction (RHEED), have produced highly valuable information on surface

studies, as reviewed, for example, by Yagi (1988, 1989, 1993). More recently low energy

electron microscopy (LEEM) has been developed, which can be combined with LEED,

and is making a major contribution (Bauer, 1994). This instrument can also be used

for photoemission microscopy (PEEM), which has been developed in several diﬀerent

versions. A specialist form of microscopy with a venerable history is ﬁeld ion micros-

copy (FIM), which is especially useful for studying individual atomic events such as

diﬀusion and cluster formation, as discussed by Bassett (1983), Kellogg (1994), Ehrlich

(1991, 1994, 1995, 1997) and Tsong & Chen (1997).

The great virtue of ﬁxed beam techniques is that the information from each picture

element (pixel) is recorded at the same time, in parallel. This leads to relatively rapid

data acquisition, and the ability to study dynamic events, often in real time, e.g. via

video recording. In contrast, data in a scanned beam technique, such as scanning elec-

tron microscopy (SEM) or scanning transmission electron microscopy (STEM), is col-

lected serially, point by point, with the sample placed after the objective lens as

illustrated in ﬁgures 3.1(c) and (d).

This conﬁguration means that multiple signals (not just electrons at the probe energy

as in TEM or REM) can be used, which makes the instruments very versatile. It also

makes them ideally adapted for computer control and computer-based data collection,

but can have a corresponding disadvantage: the need to concentrate a very high current

density into a small spot means that not all forms of information can be obtained

rapidly, that there will be substantial signal to noise ratio (SNR) problems, and that the

beam can cause damage to sensitive specimens. Nonetheless SEM and STEM form the

basis of a very useful class of techniques; UHV-SEM has been developed in several

laboratories, including the University of Sussex, and UHV-STEM especially at

Arizona State University. We examine particular developments in section 3.5.

The above techniques have been available for several decades, and have been

66 3 Electron-based techniques